id stringlengths 9 10 | submitter stringlengths 1 64 ⌀ | authors stringlengths 4 20.7k | title stringlengths 4 246 | comments stringlengths 1 523 ⌀ | journal-ref stringlengths 4 404 ⌀ | doi stringlengths 11 153 ⌀ | report-no stringlengths 2 254 ⌀ | categories stringlengths 5 98 | license stringclasses 9 values | orig_abstract stringlengths 14 3.35k | versions listlengths 1 60 | update_date stringlengths 10 10 | authors_parsed listlengths 1 1.35k | abstract stringlengths 11 3.34k |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
2307.13834 | Gabriel Klasson Landin | Gabriel Klasson Landin and Truls Jilborg | Determining the Optimal Frequencies for a Duplicated Randomized Clock
SCA Countermeasure | null | null | null | null | cs.CR | http://creativecommons.org/licenses/by/4.0/ | Side-channel attacks pose significant challenges to the security of embedded
systems, often allowing attackers to circumvent encryption algorithms in
minutes compared to the trillions of years required for brute-force attacks. To
mitigate these vulnerabilities, various countermeasures have been developed.
This study focuses on two specific countermeasures: randomization of the
encryption algorithm's clock and the incorporation of a dummy core to disguise
power traces.
The objective of this research is to identify the optimal frequencies that
yield the highest level of randomness when these two countermeasures are
combined. By investigating the interplay between clock randomization and the
presence of dummy cores, we aim to enhance the overall security of embedded
systems. The insights gained from this study will contribute to the development
of more robust countermeasures against side-channel attacks, bolstering the
protection of sensitive information and systems.
To achieve this, we conduct simulations and perform side-channel attacks on
an FPGA to establish the relationship between frequencies and the resulting
protection. We break the encryption on a non-duplicated circuit and note the
least amount of measured power traces necessary and the timing overhead. We do
this for all sets of frequencies considered which gives a good indication of
which sets of frequencies give good protection. By comparing the frequencies
generated with those from the duplicated circuit we use similar conclusions to
prove whether a frequency set is secure or not.
Based on our results we argue that having one frequency lower than half of
the base frequency and the other frequencies being close but not higher than
the base gives the highest security compared to the timing overhead measured.
| [
{
"created": "Tue, 25 Jul 2023 22:07:41 GMT",
"version": "v1"
}
] | 2023-07-27 | [
[
"Landin",
"Gabriel Klasson",
""
],
[
"Jilborg",
"Truls",
""
]
] | Side-channel attacks pose significant challenges to the security of embedded systems, often allowing attackers to circumvent encryption algorithms in minutes compared to the trillions of years required for brute-force attacks. To mitigate these vulnerabilities, various countermeasures have been developed. This study focuses on two specific countermeasures: randomization of the encryption algorithm's clock and the incorporation of a dummy core to disguise power traces. The objective of this research is to identify the optimal frequencies that yield the highest level of randomness when these two countermeasures are combined. By investigating the interplay between clock randomization and the presence of dummy cores, we aim to enhance the overall security of embedded systems. The insights gained from this study will contribute to the development of more robust countermeasures against side-channel attacks, bolstering the protection of sensitive information and systems. To achieve this, we conduct simulations and perform side-channel attacks on an FPGA to establish the relationship between frequencies and the resulting protection. We break the encryption on a non-duplicated circuit and note the least amount of measured power traces necessary and the timing overhead. We do this for all sets of frequencies considered which gives a good indication of which sets of frequencies give good protection. By comparing the frequencies generated with those from the duplicated circuit we use similar conclusions to prove whether a frequency set is secure or not. Based on our results we argue that having one frequency lower than half of the base frequency and the other frequencies being close but not higher than the base gives the highest security compared to the timing overhead measured. |
2405.08345 | Kun Song | Kun Song, Gaoming Chen, Wenhang Liu, Zhenhua Xiong | Multi-Robot Rendezvous in Unknown Environment with Limited Communication | Submit to RAL. 8 pages, 6 figures | null | null | null | cs.RO | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Rendezvous aims at gathering all robots at a specific location, which is an
important collaborative behavior for multirobot systems. However, in an unknown
environment, it is challenging to achieve rendezvous. Previous researches
mainly focus on special scenarios where communication is not allowed and each
robot executes a random searching strategy, which is highly time-consuming,
especially in large-scale environments. In this work, we focus on rendezvous in
unknown environments where communication is available. We divide this task into
two steps: rendezvous based environment exploration with relative pose (RP)
estimation and rendezvous point election. A new strategy called partitioned and
incomplete exploration for rendezvous (PIER) is proposed to efficiently explore
the unknown environment, where lightweight topological maps are constructed and
shared among robots for RP estimation with very few communications. Then, a
rendezvous point selection algorithm based on the merged topological map is
proposed for efficient rendezvous for multi-robot systems. The effectiveness of
the proposed methods is validated in both simulations and real-world
experiments.
| [
{
"created": "Tue, 14 May 2024 06:33:56 GMT",
"version": "v1"
}
] | 2024-05-15 | [
[
"Song",
"Kun",
""
],
[
"Chen",
"Gaoming",
""
],
[
"Liu",
"Wenhang",
""
],
[
"Xiong",
"Zhenhua",
""
]
] | Rendezvous aims at gathering all robots at a specific location, which is an important collaborative behavior for multirobot systems. However, in an unknown environment, it is challenging to achieve rendezvous. Previous researches mainly focus on special scenarios where communication is not allowed and each robot executes a random searching strategy, which is highly time-consuming, especially in large-scale environments. In this work, we focus on rendezvous in unknown environments where communication is available. We divide this task into two steps: rendezvous based environment exploration with relative pose (RP) estimation and rendezvous point election. A new strategy called partitioned and incomplete exploration for rendezvous (PIER) is proposed to efficiently explore the unknown environment, where lightweight topological maps are constructed and shared among robots for RP estimation with very few communications. Then, a rendezvous point selection algorithm based on the merged topological map is proposed for efficient rendezvous for multi-robot systems. The effectiveness of the proposed methods is validated in both simulations and real-world experiments. |
1501.01495 | Tobias Fehenberger | Tobias Fehenberger, Alex Alvarado, Polina Bayvel, Norbert Hanik | On Achievable Rates for Long-Haul Fiber-Optic Communications | Hard decision mutual information analysis added, two typos corrected | null | 10.1364/OE.23.009183 | null | cs.IT math.IT | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Lower bounds on mutual information (MI) of long-haul optical fiber systems
for hard-decision and soft-decision decoding are studied. Ready-to-use
expressions to calculate the MI are presented. Extensive numerical simulations
are used to quantify how changes in the optical transmitter, receiver, and
channel affect the achievable transmission rates of the system. Special
emphasis is put to the use of different quadrature amplitude modulation
formats, channel spacings, digital back-propagation schemes and probabilistic
shaping. The advantages of using MI over the prevailing $Q$-factor as a figure
of merit of coded optical systems are also highlighted.
| [
{
"created": "Wed, 7 Jan 2015 13:57:07 GMT",
"version": "v1"
},
{
"created": "Fri, 3 Apr 2015 04:36:38 GMT",
"version": "v2"
}
] | 2015-04-06 | [
[
"Fehenberger",
"Tobias",
""
],
[
"Alvarado",
"Alex",
""
],
[
"Bayvel",
"Polina",
""
],
[
"Hanik",
"Norbert",
""
]
] | Lower bounds on mutual information (MI) of long-haul optical fiber systems for hard-decision and soft-decision decoding are studied. Ready-to-use expressions to calculate the MI are presented. Extensive numerical simulations are used to quantify how changes in the optical transmitter, receiver, and channel affect the achievable transmission rates of the system. Special emphasis is put to the use of different quadrature amplitude modulation formats, channel spacings, digital back-propagation schemes and probabilistic shaping. The advantages of using MI over the prevailing $Q$-factor as a figure of merit of coded optical systems are also highlighted. |
2007.06653 | Alex Berke | Alex Berke, Jason Nawyn, Thomas Sanchez Lengeling, Kent Larson | Urban Mobility Swarms: A Scalable Implementation | null | null | 10.1007/978-3-030-52246-9_1 | null | cs.MA cs.CY | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We present a system to coordinate 'urban mobility swarms' in order to promote
the use and safety of lightweight, sustainable transit, while enhancing the
vibrancy and community fabric of cities. This work draws from behavior
exhibited by swarms of nocturnal insects, such as crickets and fireflies,
whereby synchrony unifies individuals in a decentralized network. Coordination
naturally emerges in these cases and provides a compelling demonstration of
'strength in numbers'. Our work is applied to coordinating lightweight
vehicles, such as bicycles, which are automatically inducted into ad-hoc
'swarms', united by the synchronous pulsation of light. We model individual
riders as nodes in a decentralized network and synchronize their behavior via a
peer-to-peer message protocol and algorithm, which preserves individual
privacy. Nodes broadcast over radio with a transmission range tuned to localize
swarm membership. Nodes then join or disconnect from others based on proximity,
accommodating the dynamically changing topology of urban mobility networks.
This paper provides a technical description of our system, including the
protocol and algorithm to coordinate the swarming behavior that emerges from
it. We also demonstrate its implementation in code, circuity, and hardware,
with a system prototype tested on a city bike-share. In doing so, we evince the
scalability of our system. Our prototype uses low-cost components, and
bike-share programs, which manage bicycle fleets distributed across cities,
could deploy the system at city-scale. Our flexible, decentralized design
allows additional bikes to then connect with the network, enhancing its scale
and impact.
| [
{
"created": "Mon, 13 Jul 2020 19:44:16 GMT",
"version": "v1"
}
] | 2020-07-15 | [
[
"Berke",
"Alex",
""
],
[
"Nawyn",
"Jason",
""
],
[
"Lengeling",
"Thomas Sanchez",
""
],
[
"Larson",
"Kent",
""
]
] | We present a system to coordinate 'urban mobility swarms' in order to promote the use and safety of lightweight, sustainable transit, while enhancing the vibrancy and community fabric of cities. This work draws from behavior exhibited by swarms of nocturnal insects, such as crickets and fireflies, whereby synchrony unifies individuals in a decentralized network. Coordination naturally emerges in these cases and provides a compelling demonstration of 'strength in numbers'. Our work is applied to coordinating lightweight vehicles, such as bicycles, which are automatically inducted into ad-hoc 'swarms', united by the synchronous pulsation of light. We model individual riders as nodes in a decentralized network and synchronize their behavior via a peer-to-peer message protocol and algorithm, which preserves individual privacy. Nodes broadcast over radio with a transmission range tuned to localize swarm membership. Nodes then join or disconnect from others based on proximity, accommodating the dynamically changing topology of urban mobility networks. This paper provides a technical description of our system, including the protocol and algorithm to coordinate the swarming behavior that emerges from it. We also demonstrate its implementation in code, circuity, and hardware, with a system prototype tested on a city bike-share. In doing so, we evince the scalability of our system. Our prototype uses low-cost components, and bike-share programs, which manage bicycle fleets distributed across cities, could deploy the system at city-scale. Our flexible, decentralized design allows additional bikes to then connect with the network, enhancing its scale and impact. |
2207.05624 | Ali Munir | Sepehr Abbasi, Shiva Ketabi, Ali Munir, Mahmoud Bahnasy, Yashar
Ganjali | DWTCP: Ultra Low Latency Congestion Control Protocol for Data Centers | 19 pages, 17 figures | null | null | null | cs.NI cs.DC | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Congestion control algorithms rely on a variety of congestion signals (packet
loss, Explicit Congestion Notification, delay, etc.) to achieve fast
convergence, high utilization, and fairness among flows. A key limitation of
these congestion signals is that they are either late in feedback or they incur
significant overheads. An ideal congestion control must discover any available
bandwidth in the network, detect congestion as soon as link utilization
approaches full capacity, and react timely to avoid queuing and packet drops,
without significant overheads. To this end, this work proposes Scout service
that leverages priority queues to infer bandwidth availability and link
busyness at the host. The key observation here is that as the high priority
queue (HPQ) gets busier, the low priority queue (LPQ) is served less.
Therefore, the state of the link can be observed from the LPQ and any
congestion can be detected several RTTs earlier than observing the HPQ. We
propose a new transport protocol, Double-Window Transmission Control Protocol
(DWTCP) that builds upon the Scout service to dynamically adjust its congestion
window. Our testbed and simulation-based evaluation demonstrates that Scout
enables a data center transport to achieve high throughput, near-zero queues,
lower latency, and high fairness.
| [
{
"created": "Tue, 12 Jul 2022 15:46:19 GMT",
"version": "v1"
}
] | 2022-07-13 | [
[
"Abbasi",
"Sepehr",
""
],
[
"Ketabi",
"Shiva",
""
],
[
"Munir",
"Ali",
""
],
[
"Bahnasy",
"Mahmoud",
""
],
[
"Ganjali",
"Yashar",
""
]
] | Congestion control algorithms rely on a variety of congestion signals (packet loss, Explicit Congestion Notification, delay, etc.) to achieve fast convergence, high utilization, and fairness among flows. A key limitation of these congestion signals is that they are either late in feedback or they incur significant overheads. An ideal congestion control must discover any available bandwidth in the network, detect congestion as soon as link utilization approaches full capacity, and react timely to avoid queuing and packet drops, without significant overheads. To this end, this work proposes Scout service that leverages priority queues to infer bandwidth availability and link busyness at the host. The key observation here is that as the high priority queue (HPQ) gets busier, the low priority queue (LPQ) is served less. Therefore, the state of the link can be observed from the LPQ and any congestion can be detected several RTTs earlier than observing the HPQ. We propose a new transport protocol, Double-Window Transmission Control Protocol (DWTCP) that builds upon the Scout service to dynamically adjust its congestion window. Our testbed and simulation-based evaluation demonstrates that Scout enables a data center transport to achieve high throughput, near-zero queues, lower latency, and high fairness. |
1804.07237 | Jiamiao Xu | Jiamiao Xu, Shujian Yu, Xinge You, Mengjun Leng, Xiao-Yuan Jing and C.
L. Philip Chen | Multi-view Hybrid Embedding: A Divide-and-Conquer Approach | This paper has been accepted by IEEE Transactions on Cybernetics | null | null | null | cs.CV cs.LG stat.ML | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We present a novel cross-view classification algorithm where the gallery and
probe data come from different views. A popular approach to tackle this problem
is the multi-view subspace learning (MvSL) that aims to learn a latent subspace
shared by multi-view data. Despite promising results obtained on some
applications, the performance of existing methods deteriorates dramatically
when the multi-view data is sampled from nonlinear manifolds or suffers from
heavy outliers. To circumvent this drawback, motivated by the
Divide-and-Conquer strategy, we propose Multi-view Hybrid Embedding (MvHE), a
unique method of dividing the problem of cross-view classification into three
subproblems and building one model for each subproblem. Specifically, the first
model is designed to remove view discrepancy, whereas the second and third
models attempt to discover the intrinsic nonlinear structure and to increase
discriminability in intra-view and inter-view samples respectively. The kernel
extension is conducted to further boost the representation power of MvHE.
Extensive experiments are conducted on four benchmark datasets. Our methods
demonstrate overwhelming advantages against the state-of-the-art MvSL based
cross-view classification approaches in terms of classification accuracy and
robustness.
| [
{
"created": "Thu, 19 Apr 2018 15:38:15 GMT",
"version": "v1"
},
{
"created": "Mon, 21 Jan 2019 10:46:35 GMT",
"version": "v2"
}
] | 2019-01-23 | [
[
"Xu",
"Jiamiao",
""
],
[
"Yu",
"Shujian",
""
],
[
"You",
"Xinge",
""
],
[
"Leng",
"Mengjun",
""
],
[
"Jing",
"Xiao-Yuan",
""
],
[
"Chen",
"C. L. Philip",
""
]
] | We present a novel cross-view classification algorithm where the gallery and probe data come from different views. A popular approach to tackle this problem is the multi-view subspace learning (MvSL) that aims to learn a latent subspace shared by multi-view data. Despite promising results obtained on some applications, the performance of existing methods deteriorates dramatically when the multi-view data is sampled from nonlinear manifolds or suffers from heavy outliers. To circumvent this drawback, motivated by the Divide-and-Conquer strategy, we propose Multi-view Hybrid Embedding (MvHE), a unique method of dividing the problem of cross-view classification into three subproblems and building one model for each subproblem. Specifically, the first model is designed to remove view discrepancy, whereas the second and third models attempt to discover the intrinsic nonlinear structure and to increase discriminability in intra-view and inter-view samples respectively. The kernel extension is conducted to further boost the representation power of MvHE. Extensive experiments are conducted on four benchmark datasets. Our methods demonstrate overwhelming advantages against the state-of-the-art MvSL based cross-view classification approaches in terms of classification accuracy and robustness. |
2203.16110 | Sean Bin Yang | Sean Bin Yang, Chenjuan Guo, Jilin Hu, Bin Yang, Jian Tang, and
Christian S. Jensen | Weakly-supervised Temporal Path Representation Learning with Contrastive
Curriculum Learning -- Extended Version | This paper has been accepted by IEEE ICDE-22 | null | null | null | cs.LG cs.AI | http://creativecommons.org/licenses/by-nc-sa/4.0/ | In step with the digitalization of transportation, we are witnessing a
growing range of path-based smart-city applications, e.g., travel-time
estimation and travel path ranking. A temporal path(TP) that includes temporal
information, e.g., departure time, into the path is fundamental to enable such
applications. In this setting, it is essential to learn generic temporal path
representations(TPRs) that consider spatial and temporal correlations
simultaneously and that can be used in different applications, i.e., downstream
tasks. Existing methods fail to achieve the goal since (i) supervised methods
require large amounts of task-specific labels when training and thus fail to
generalize the obtained TPRs to other tasks; (ii) through unsupervised methods
can learn generic representations, they disregard the temporal aspect, leading
to sub-optimal results. To contend with the limitations of existing solutions,
we propose a Weakly-Supervised Contrastive (WSC) learning model. We first
propose a temporal path encoder that encodes both the spatial and temporal
information of a temporal path into a TPR. To train the encoder, we introduce
weak labels that are easy and inexpensive to obtain and are relevant to
different tasks, e.g., temporal labels indicating peak vs. off-peak hours from
departure times. Based on the weak labels, we construct meaningful positive and
negative temporal path samples by considering both spatial and temporal
information, which facilities training the encoder using contrastive learning
by pulling closer to the positive samples' representations while pushing away
the negative samples' representations. To better guide contrastive learning, we
propose a learning strategy based on Curriculum Learning such that the learning
performs from easy to hard training instances. Experiments studies verify the
effectiveness of the proposed method.
| [
{
"created": "Wed, 30 Mar 2022 07:36:20 GMT",
"version": "v1"
},
{
"created": "Fri, 1 Apr 2022 15:37:26 GMT",
"version": "v2"
},
{
"created": "Fri, 15 Apr 2022 15:41:04 GMT",
"version": "v3"
}
] | 2022-04-18 | [
[
"Yang",
"Sean Bin",
""
],
[
"Guo",
"Chenjuan",
""
],
[
"Hu",
"Jilin",
""
],
[
"Yang",
"Bin",
""
],
[
"Tang",
"Jian",
""
],
[
"Jensen",
"Christian S.",
""
]
] | In step with the digitalization of transportation, we are witnessing a growing range of path-based smart-city applications, e.g., travel-time estimation and travel path ranking. A temporal path(TP) that includes temporal information, e.g., departure time, into the path is fundamental to enable such applications. In this setting, it is essential to learn generic temporal path representations(TPRs) that consider spatial and temporal correlations simultaneously and that can be used in different applications, i.e., downstream tasks. Existing methods fail to achieve the goal since (i) supervised methods require large amounts of task-specific labels when training and thus fail to generalize the obtained TPRs to other tasks; (ii) through unsupervised methods can learn generic representations, they disregard the temporal aspect, leading to sub-optimal results. To contend with the limitations of existing solutions, we propose a Weakly-Supervised Contrastive (WSC) learning model. We first propose a temporal path encoder that encodes both the spatial and temporal information of a temporal path into a TPR. To train the encoder, we introduce weak labels that are easy and inexpensive to obtain and are relevant to different tasks, e.g., temporal labels indicating peak vs. off-peak hours from departure times. Based on the weak labels, we construct meaningful positive and negative temporal path samples by considering both spatial and temporal information, which facilities training the encoder using contrastive learning by pulling closer to the positive samples' representations while pushing away the negative samples' representations. To better guide contrastive learning, we propose a learning strategy based on Curriculum Learning such that the learning performs from easy to hard training instances. Experiments studies verify the effectiveness of the proposed method. |
1106.3694 | Hugo Jim\'enez-P\'erez | Hugo Jim\'enez-P\'erez and Jacques Laskar | A time-parallel algorithm for almost integrable Hamiltonian systems | 19 pages, 6 figures | null | null | null | cs.NA math.NA | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We introduce a time-parallel algorithm for solving numerically almost
integrable Hamiltonian systems in action-angle coordinates. This algorithm is a
refinement of that introduced by Saha, Stadel and Tremaine in 1997 (SST97) for
the same type of problems. Our refined algorithm has a better convergence
obtained from the use of derivatives of the perturbing term not considered in
the original SST97 algorithm. An advantage of this algorithm is its
independence of the step-size for the parallelized procedures which can be
consider as a particular case of the parareal scheme.
| [
{
"created": "Sat, 18 Jun 2011 22:37:28 GMT",
"version": "v1"
}
] | 2011-06-21 | [
[
"Jiménez-Pérez",
"Hugo",
""
],
[
"Laskar",
"Jacques",
""
]
] | We introduce a time-parallel algorithm for solving numerically almost integrable Hamiltonian systems in action-angle coordinates. This algorithm is a refinement of that introduced by Saha, Stadel and Tremaine in 1997 (SST97) for the same type of problems. Our refined algorithm has a better convergence obtained from the use of derivatives of the perturbing term not considered in the original SST97 algorithm. An advantage of this algorithm is its independence of the step-size for the parallelized procedures which can be consider as a particular case of the parareal scheme. |
2204.07026 | Guilherme Maeda | Guilherme Maeda | Blending Primitive Policies in Shared Control for Assisted Teleoperation | null | null | null | null | cs.RO | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Movement primitives have the property to accommodate changes in the robot
state while maintaining attraction to the original policy. As such, we
investigate the use of primitives as a blending mechanism by considering that
state deviations from the original policy are caused by user inputs. As the
primitive recovers from the user input, it implicitly blends human and robot
policies without requiring their weightings -- referred to as arbitration. In
this paper, we adopt Dynamical Movement Primitives (DMPs), which allow us to
avoid the need for multiple demonstrations, and are fast enough to enable
numerous instantiations, one for each hypothesis of the human intent. User
studies are presented on assisted teleoperation tasks of reaching multiple
goals and dynamic obstacle avoidance. Comparable performance to conventional
teleoperation was achieved while significantly decreasing human intervention,
often by more than 60%.
| [
{
"created": "Thu, 14 Apr 2022 15:28:00 GMT",
"version": "v1"
}
] | 2022-04-15 | [
[
"Maeda",
"Guilherme",
""
]
] | Movement primitives have the property to accommodate changes in the robot state while maintaining attraction to the original policy. As such, we investigate the use of primitives as a blending mechanism by considering that state deviations from the original policy are caused by user inputs. As the primitive recovers from the user input, it implicitly blends human and robot policies without requiring their weightings -- referred to as arbitration. In this paper, we adopt Dynamical Movement Primitives (DMPs), which allow us to avoid the need for multiple demonstrations, and are fast enough to enable numerous instantiations, one for each hypothesis of the human intent. User studies are presented on assisted teleoperation tasks of reaching multiple goals and dynamic obstacle avoidance. Comparable performance to conventional teleoperation was achieved while significantly decreasing human intervention, often by more than 60%. |
2205.01626 | Geoffrey Bomarito | G.F. Bomarito and P.E. Leser and N.C.M Strauss and K.M. Garbrecht and
J.D. Hochhalter | Automated Learning of Interpretable Models with Quantified Uncertainty | null | null | 10.1016/j.cma.2022.115732 | null | cs.NE cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Interpretability and uncertainty quantification in machine learning can
provide justification for decisions, promote scientific discovery and lead to a
better understanding of model behavior. Symbolic regression provides inherently
interpretable machine learning, but relatively little work has focused on the
use of symbolic regression on noisy data and the accompanying necessity to
quantify uncertainty. A new Bayesian framework for genetic-programming-based
symbolic regression (GPSR) is introduced that uses model evidence (i.e.,
marginal likelihood) to formulate replacement probability during the selection
phase of evolution. Model parameter uncertainty is automatically quantified,
enabling probabilistic predictions with each equation produced by the GPSR
algorithm. Model evidence is also quantified in this process, and its use is
shown to increase interpretability, improve robustness to noise, and reduce
overfitting when compared to a conventional GPSR implementation on both
numerical and physical experiments.
| [
{
"created": "Tue, 12 Apr 2022 19:56:42 GMT",
"version": "v1"
}
] | 2022-11-23 | [
[
"Bomarito",
"G. F.",
""
],
[
"Leser",
"P. E.",
""
],
[
"Strauss",
"N. C. M",
""
],
[
"Garbrecht",
"K. M.",
""
],
[
"Hochhalter",
"J. D.",
""
]
] | Interpretability and uncertainty quantification in machine learning can provide justification for decisions, promote scientific discovery and lead to a better understanding of model behavior. Symbolic regression provides inherently interpretable machine learning, but relatively little work has focused on the use of symbolic regression on noisy data and the accompanying necessity to quantify uncertainty. A new Bayesian framework for genetic-programming-based symbolic regression (GPSR) is introduced that uses model evidence (i.e., marginal likelihood) to formulate replacement probability during the selection phase of evolution. Model parameter uncertainty is automatically quantified, enabling probabilistic predictions with each equation produced by the GPSR algorithm. Model evidence is also quantified in this process, and its use is shown to increase interpretability, improve robustness to noise, and reduce overfitting when compared to a conventional GPSR implementation on both numerical and physical experiments. |
2010.09921 | Jun Yu | Cheng Meng and Jun Yu and Jingyi Zhang and Ping Ma and Wenxuan Zhong | Sufficient dimension reduction for classification using principal
optimal transport direction | 18 pages, 4 figures, to be published in 34th Conference on Neural
Information Processing Systems (NeurIPS 2020), add the supplementary material | null | null | null | cs.LG stat.ME stat.ML | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Sufficient dimension reduction is used pervasively as a supervised dimension
reduction approach. Most existing sufficient dimension reduction methods are
developed for data with a continuous response and may have an unsatisfactory
performance for the categorical response, especially for the binary-response.
To address this issue, we propose a novel estimation method of sufficient
dimension reduction subspace (SDR subspace) using optimal transport. The
proposed method, named principal optimal transport direction (POTD), estimates
the basis of the SDR subspace using the principal directions of the optimal
transport coupling between the data respecting different response categories.
The proposed method also reveals the relationship among three seemingly
irrelevant topics, i.e., sufficient dimension reduction, support vector
machine, and optimal transport. We study the asymptotic properties of POTD and
show that in the cases when the class labels contain no error, POTD estimates
the SDR subspace exclusively. Empirical studies show POTD outperforms most of
the state-of-the-art linear dimension reduction methods.
| [
{
"created": "Mon, 19 Oct 2020 23:38:31 GMT",
"version": "v1"
},
{
"created": "Wed, 21 Oct 2020 01:48:14 GMT",
"version": "v2"
},
{
"created": "Tue, 24 Nov 2020 04:34:24 GMT",
"version": "v3"
},
{
"created": "Tue, 2 Feb 2021 04:10:15 GMT",
"version": "v4"
}
] | 2021-02-03 | [
[
"Meng",
"Cheng",
""
],
[
"Yu",
"Jun",
""
],
[
"Zhang",
"Jingyi",
""
],
[
"Ma",
"Ping",
""
],
[
"Zhong",
"Wenxuan",
""
]
] | Sufficient dimension reduction is used pervasively as a supervised dimension reduction approach. Most existing sufficient dimension reduction methods are developed for data with a continuous response and may have an unsatisfactory performance for the categorical response, especially for the binary-response. To address this issue, we propose a novel estimation method of sufficient dimension reduction subspace (SDR subspace) using optimal transport. The proposed method, named principal optimal transport direction (POTD), estimates the basis of the SDR subspace using the principal directions of the optimal transport coupling between the data respecting different response categories. The proposed method also reveals the relationship among three seemingly irrelevant topics, i.e., sufficient dimension reduction, support vector machine, and optimal transport. We study the asymptotic properties of POTD and show that in the cases when the class labels contain no error, POTD estimates the SDR subspace exclusively. Empirical studies show POTD outperforms most of the state-of-the-art linear dimension reduction methods. |
1705.07771 | Kang Wang | Kang Wang, Xueqian Wang, Gang Li | Simulation Experiment of BCI Based on Imagined Speech EEG Decoding | null | null | null | null | cs.HC | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Brain Computer Interface (BCI) can help patients of neuromuscular diseases
restore parts of the movement and communication abilities that they have lost.
Most of BCIs rely on mapping brain activities to device instructions, but
limited number of brain activities decides the limited abilities of BCIs. To
deal with the problem of limited ablility of BCI, this paper verified the
feasibility of constructing BCI based on decoding imagined speech
electroencephalography (EEG). As sentences decoded from EEG can have rich
meanings, BCIs based on EEG decoding can achieve numerous control instructions.
By combining a modified EEG feature extraction mehtod with connectionist
temporal classification (CTC), this paper simulated decoding imagined speech
EEG using synthetic EEG data without help of speech signal. The performance of
decoding model over synthetic data to a certain extent demonstrated the
feasibility of constructing BCI based on imagined speech brain signal.
| [
{
"created": "Mon, 22 May 2017 14:34:20 GMT",
"version": "v1"
}
] | 2017-05-23 | [
[
"Wang",
"Kang",
""
],
[
"Wang",
"Xueqian",
""
],
[
"Li",
"Gang",
""
]
] | Brain Computer Interface (BCI) can help patients of neuromuscular diseases restore parts of the movement and communication abilities that they have lost. Most of BCIs rely on mapping brain activities to device instructions, but limited number of brain activities decides the limited abilities of BCIs. To deal with the problem of limited ablility of BCI, this paper verified the feasibility of constructing BCI based on decoding imagined speech electroencephalography (EEG). As sentences decoded from EEG can have rich meanings, BCIs based on EEG decoding can achieve numerous control instructions. By combining a modified EEG feature extraction mehtod with connectionist temporal classification (CTC), this paper simulated decoding imagined speech EEG using synthetic EEG data without help of speech signal. The performance of decoding model over synthetic data to a certain extent demonstrated the feasibility of constructing BCI based on imagined speech brain signal. |
2207.03622 | Hazim Shakhatreh | Hazim Shakhatreh, Ahmad Sawalmeh, Ali H Alenezi, Sharief Abdel-Razeq,
Muhannad Almutiry, Ala Al-Fuqaha | Mobile-IRS Assisted Next Generation UAV Communication Networks | 11 pages, 8 figures | null | null | null | cs.IT math.IT | http://creativecommons.org/licenses/by/4.0/ | Prior research on intelligent reflection surface (IRS)-assisted unmanned
aerial vehicle (UAV) communications has focused on a fixed location for the IRS
or mounted on a UAV. The assumption that the IRS is located at a fixed position
will prohibit mobile users from maximizing many wireless network benefits, such
as data rate and coverage. Furthermore, assuming that the IRS is placed on a
UAV is impractical for various reasons, including the IRS's weight and size and
the speed of wind in severe weather. Unlike previous studies, this study
assumes a single UAV and an IRS mounted on a mobile ground vehicle (M-IRS) to
be deployed in an Internet-of-Things (IoT) 6G wireless network to maximize the
average data rate. Such a methodology for providing wireless coverage using an
M-IRS assisted UAV system is expected in smart cities. In this paper, we
formulate an optimization problem to find an efficient trajectory for the UAV,
an efficient path for the M-IRS, and users' power allocation coefficients that
maximize the average data rate for mobile ground users. Due to its
intractability, we propose efficient techniques that can help in finding the
solution to the optimization problem. First, we show that our dynamic power
allocation technique outperforms the fixed power allocation technique in terms
of network average sum rate. Then we employ the individual movement model
(Random Waypoint Model) in order to represent the users' movements inside the
coverage area. Finally, we propose an efficient approach using a Genetic
Algorithm (GA) for finding an efficient trajectory for the UAV, and an
efficient path for the M-IRS to provide wireless connectivity for mobile users
during their movement. We demonstrate through simulations that our methodology
can enhance the average data rate by 15\% on average compared with the static
IRS and by 25\% on average compared without the IRS system.
| [
{
"created": "Fri, 8 Jul 2022 00:06:06 GMT",
"version": "v1"
}
] | 2022-07-11 | [
[
"Shakhatreh",
"Hazim",
""
],
[
"Sawalmeh",
"Ahmad",
""
],
[
"Alenezi",
"Ali H",
""
],
[
"Abdel-Razeq",
"Sharief",
""
],
[
"Almutiry",
"Muhannad",
""
],
[
"Al-Fuqaha",
"Ala",
""
]
] | Prior research on intelligent reflection surface (IRS)-assisted unmanned aerial vehicle (UAV) communications has focused on a fixed location for the IRS or mounted on a UAV. The assumption that the IRS is located at a fixed position will prohibit mobile users from maximizing many wireless network benefits, such as data rate and coverage. Furthermore, assuming that the IRS is placed on a UAV is impractical for various reasons, including the IRS's weight and size and the speed of wind in severe weather. Unlike previous studies, this study assumes a single UAV and an IRS mounted on a mobile ground vehicle (M-IRS) to be deployed in an Internet-of-Things (IoT) 6G wireless network to maximize the average data rate. Such a methodology for providing wireless coverage using an M-IRS assisted UAV system is expected in smart cities. In this paper, we formulate an optimization problem to find an efficient trajectory for the UAV, an efficient path for the M-IRS, and users' power allocation coefficients that maximize the average data rate for mobile ground users. Due to its intractability, we propose efficient techniques that can help in finding the solution to the optimization problem. First, we show that our dynamic power allocation technique outperforms the fixed power allocation technique in terms of network average sum rate. Then we employ the individual movement model (Random Waypoint Model) in order to represent the users' movements inside the coverage area. Finally, we propose an efficient approach using a Genetic Algorithm (GA) for finding an efficient trajectory for the UAV, and an efficient path for the M-IRS to provide wireless connectivity for mobile users during their movement. We demonstrate through simulations that our methodology can enhance the average data rate by 15\% on average compared with the static IRS and by 25\% on average compared without the IRS system. |
1901.07521 | Ali Baheri | Ali Baheri and Chris Vermillion | Economically Efficient Combined Plant and Controller Design Using Batch
Bayesian Optimization: Mathematical Framework and Airborne Wind Energy Case
Study | null | null | null | null | cs.SY | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We present a novel data-driven nested optimization framework that addresses
the problem of coupling between plant and controller optimization. This
optimization strategy is tailored towards instances where a closed-form
expression for the system dynamic response is unobtainable and simulations or
experiments are necessary. Specifically, Bayesian Optimization, which is a
data-driven technique for finding the optimum of an unknown and
expensive-to-evaluate objective function, is employed to solve a nested
optimization problem. The underlying objective function is modeled by a
Gaussian Process (GP); then, Bayesian Optimization utilizes the predictive
uncertainty information from the GP to determine the best subsequent control or
plant parameters. The proposed framework differs from the majority of co-design
literature where there exists a closed-form model of the system dynamics.
Furthermore, we utilize the idea of Batch Bayesian Optimization at the plant
optimization level to generate a set of plant designs at each iteration of the
overall optimization process, recognizing that there will exist economies of
scale in running multiple experiments in each iteration of the plant design
process. We validate the proposed framework for a Buoyant Airborne Turbine
(BAT). We choose the horizontal stabilizer area, longitudinal center of mass
relative to center of buoyancy (plant parameters), and the pitch angle
set-point (controller parameter) as our decision variables. Our results
demonstrate that these plant and control parameters converge to their
respective optimal values within only a few iterations.
| [
{
"created": "Tue, 22 Jan 2019 18:52:41 GMT",
"version": "v1"
}
] | 2019-01-23 | [
[
"Baheri",
"Ali",
""
],
[
"Vermillion",
"Chris",
""
]
] | We present a novel data-driven nested optimization framework that addresses the problem of coupling between plant and controller optimization. This optimization strategy is tailored towards instances where a closed-form expression for the system dynamic response is unobtainable and simulations or experiments are necessary. Specifically, Bayesian Optimization, which is a data-driven technique for finding the optimum of an unknown and expensive-to-evaluate objective function, is employed to solve a nested optimization problem. The underlying objective function is modeled by a Gaussian Process (GP); then, Bayesian Optimization utilizes the predictive uncertainty information from the GP to determine the best subsequent control or plant parameters. The proposed framework differs from the majority of co-design literature where there exists a closed-form model of the system dynamics. Furthermore, we utilize the idea of Batch Bayesian Optimization at the plant optimization level to generate a set of plant designs at each iteration of the overall optimization process, recognizing that there will exist economies of scale in running multiple experiments in each iteration of the plant design process. We validate the proposed framework for a Buoyant Airborne Turbine (BAT). We choose the horizontal stabilizer area, longitudinal center of mass relative to center of buoyancy (plant parameters), and the pitch angle set-point (controller parameter) as our decision variables. Our results demonstrate that these plant and control parameters converge to their respective optimal values within only a few iterations. |
2210.11987 | Marco Gaido | Marco Gaido, Sara Papi, Matteo Negri, Marco Turchi | Joint Speech Translation and Named Entity Recognition | Accepted at INTERSPEECH 2023 | null | 10.21437/Interspeech.2023-1767 | null | cs.CL | http://creativecommons.org/licenses/by-sa/4.0/ | Modern automatic translation systems aim at place the human at the center by
providing contextual support and knowledge. In this context, a critical task is
enriching the output with information regarding the mentioned entities, which
is currently achieved processing the generated translation with named entity
recognition (NER) and entity linking systems. In light of the recent promising
results shown by direct speech translation (ST) models and the known weaknesses
of cascades (error propagation and additional latency), in this paper we
propose multitask models that jointly perform ST and NER, and compare them with
a cascade baseline. The experimental results show that our models significantly
outperform the cascade on the NER task (by 0.4-1.0 F1), without degradation in
terms of translation quality, and with the same computational efficiency of a
plain direct ST model.
| [
{
"created": "Fri, 21 Oct 2022 14:24:46 GMT",
"version": "v1"
},
{
"created": "Sat, 20 May 2023 14:33:35 GMT",
"version": "v2"
}
] | 2023-10-09 | [
[
"Gaido",
"Marco",
""
],
[
"Papi",
"Sara",
""
],
[
"Negri",
"Matteo",
""
],
[
"Turchi",
"Marco",
""
]
] | Modern automatic translation systems aim at place the human at the center by providing contextual support and knowledge. In this context, a critical task is enriching the output with information regarding the mentioned entities, which is currently achieved processing the generated translation with named entity recognition (NER) and entity linking systems. In light of the recent promising results shown by direct speech translation (ST) models and the known weaknesses of cascades (error propagation and additional latency), in this paper we propose multitask models that jointly perform ST and NER, and compare them with a cascade baseline. The experimental results show that our models significantly outperform the cascade on the NER task (by 0.4-1.0 F1), without degradation in terms of translation quality, and with the same computational efficiency of a plain direct ST model. |
2012.05836 | Tiago Eugenio de Melo | Tiago de Melo | User Questions from Tweets on COVID-19: An Exploratory Study | in Portuguese | null | null | null | cs.SI | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Social media platforms, such as Twitter, provide a suitable avenue for users
(people or patients) concerned on health questions to discuss and share
information with each other. In December 2019, a few coronavirus disease cases
were first reported in China. Soon after, the World Health Organization (WHO)
declared a state of emergency due to the rapid spread of the virus in other
parts of the world. In this work, we used automated extraction of COVID-19
discussion from Twitter and a natural language processing (NLP) method based on
topic modeling to discover the main questions related to COVID-19 from tweets.
Moreover, we created a Named Entity Recognition (NER) model to identify the
main entities of four different categories: disease, drug, person, and
organization. Our findings can help policy makers and health care organizations
to understand the issues of people on COVID-19 and it can be used to address
them appropriately.
| [
{
"created": "Fri, 20 Nov 2020 12:29:55 GMT",
"version": "v1"
}
] | 2020-12-11 | [
[
"de Melo",
"Tiago",
""
]
] | Social media platforms, such as Twitter, provide a suitable avenue for users (people or patients) concerned on health questions to discuss and share information with each other. In December 2019, a few coronavirus disease cases were first reported in China. Soon after, the World Health Organization (WHO) declared a state of emergency due to the rapid spread of the virus in other parts of the world. In this work, we used automated extraction of COVID-19 discussion from Twitter and a natural language processing (NLP) method based on topic modeling to discover the main questions related to COVID-19 from tweets. Moreover, we created a Named Entity Recognition (NER) model to identify the main entities of four different categories: disease, drug, person, and organization. Our findings can help policy makers and health care organizations to understand the issues of people on COVID-19 and it can be used to address them appropriately. |
2308.02357 | Sanju Tiwari Dr | Nandana Mihindukulasooriya, Sanju Tiwari, Carlos F. Enguix, Kusum Lata | Text2KGBench: A Benchmark for Ontology-Driven Knowledge Graph Generation
from Text | 15 pages, 3 figures, 4 tables. Accepted at ISWC 2023 (Resources
Track) | null | null | null | cs.CL cs.AI | http://creativecommons.org/licenses/by/4.0/ | The recent advances in large language models (LLM) and foundation models with
emergent capabilities have been shown to improve the performance of many NLP
tasks. LLMs and Knowledge Graphs (KG) can complement each other such that LLMs
can be used for KG construction or completion while existing KGs can be used
for different tasks such as making LLM outputs explainable or fact-checking in
Neuro-Symbolic manner. In this paper, we present Text2KGBench, a benchmark to
evaluate the capabilities of language models to generate KGs from natural
language text guided by an ontology. Given an input ontology and a set of
sentences, the task is to extract facts from the text while complying with the
given ontology (concepts, relations, domain/range constraints) and being
faithful to the input sentences. We provide two datasets (i) Wikidata-TekGen
with 10 ontologies and 13,474 sentences and (ii) DBpedia-WebNLG with 19
ontologies and 4,860 sentences. We define seven evaluation metrics to measure
fact extraction performance, ontology conformance, and hallucinations by LLMs.
Furthermore, we provide results for two baseline models, Vicuna-13B and
Alpaca-LoRA-13B using automatic prompt generation from test cases. The baseline
results show that there is room for improvement using both Semantic Web and
Natural Language Processing techniques.
| [
{
"created": "Fri, 4 Aug 2023 14:47:15 GMT",
"version": "v1"
}
] | 2023-08-07 | [
[
"Mihindukulasooriya",
"Nandana",
""
],
[
"Tiwari",
"Sanju",
""
],
[
"Enguix",
"Carlos F.",
""
],
[
"Lata",
"Kusum",
""
]
] | The recent advances in large language models (LLM) and foundation models with emergent capabilities have been shown to improve the performance of many NLP tasks. LLMs and Knowledge Graphs (KG) can complement each other such that LLMs can be used for KG construction or completion while existing KGs can be used for different tasks such as making LLM outputs explainable or fact-checking in Neuro-Symbolic manner. In this paper, we present Text2KGBench, a benchmark to evaluate the capabilities of language models to generate KGs from natural language text guided by an ontology. Given an input ontology and a set of sentences, the task is to extract facts from the text while complying with the given ontology (concepts, relations, domain/range constraints) and being faithful to the input sentences. We provide two datasets (i) Wikidata-TekGen with 10 ontologies and 13,474 sentences and (ii) DBpedia-WebNLG with 19 ontologies and 4,860 sentences. We define seven evaluation metrics to measure fact extraction performance, ontology conformance, and hallucinations by LLMs. Furthermore, we provide results for two baseline models, Vicuna-13B and Alpaca-LoRA-13B using automatic prompt generation from test cases. The baseline results show that there is room for improvement using both Semantic Web and Natural Language Processing techniques. |
1906.06606 | Yair Feldman | Yair Feldman, Ran El-Yaniv | Multi-Hop Paragraph Retrieval for Open-Domain Question Answering | ACL 2019 | null | null | null | cs.CL cs.IR | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | This paper is concerned with the task of multi-hop open-domain Question
Answering (QA). This task is particularly challenging since it requires the
simultaneous performance of textual reasoning and efficient searching. We
present a method for retrieving multiple supporting paragraphs, nested amidst a
large knowledge base, which contain the necessary evidence to answer a given
question. Our method iteratively retrieves supporting paragraphs by forming a
joint vector representation of both a question and a paragraph. The retrieval
is performed by considering contextualized sentence-level representations of
the paragraphs in the knowledge source. Our method achieves state-of-the-art
performance over two well-known datasets, SQuAD-Open and HotpotQA, which serve
as our single- and multi-hop open-domain QA benchmarks, respectively.
| [
{
"created": "Sat, 15 Jun 2019 19:17:10 GMT",
"version": "v1"
}
] | 2019-06-18 | [
[
"Feldman",
"Yair",
""
],
[
"El-Yaniv",
"Ran",
""
]
] | This paper is concerned with the task of multi-hop open-domain Question Answering (QA). This task is particularly challenging since it requires the simultaneous performance of textual reasoning and efficient searching. We present a method for retrieving multiple supporting paragraphs, nested amidst a large knowledge base, which contain the necessary evidence to answer a given question. Our method iteratively retrieves supporting paragraphs by forming a joint vector representation of both a question and a paragraph. The retrieval is performed by considering contextualized sentence-level representations of the paragraphs in the knowledge source. Our method achieves state-of-the-art performance over two well-known datasets, SQuAD-Open and HotpotQA, which serve as our single- and multi-hop open-domain QA benchmarks, respectively. |
1707.00587 | Paul Jaeger | Fabian Isensee, Paul Jaeger, Peter M. Full, Ivo Wolf, Sandy Engelhardt
and Klaus H. Maier-Hein | Automatic Cardiac Disease Assessment on cine-MRI via Time-Series
Segmentation and Domain Specific Features | To appear in the STACOM 2017 proceedings | null | 10.1007/978-3-319-75541-0 | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Cardiac magnetic resonance imaging improves on diagnosis of cardiovascular
diseases by providing images at high spatiotemporal resolution. Manual
evaluation of these time-series, however, is expensive and prone to biased and
non-reproducible outcomes. In this paper, we present a method that addresses
named limitations by integrating segmentation and disease classification into a
fully automatic processing pipeline. We use an ensemble of UNet inspired
architectures for segmentation of cardiac structures such as the left and right
ventricular cavity (LVC, RVC) and the left ventricular myocardium (LVM) on each
time instance of the cardiac cycle. For the classification task, information is
extracted from the segmented time-series in form of comprehensive features
handcrafted to reflect diagnostic clinical procedures. Based on these features
we train an ensemble of heavily regularized multilayer perceptrons (MLP) and a
random forest classifier to predict the pathologic target class. We evaluated
our method on the ACDC dataset (4 pathology groups, 1 healthy group) and
achieve dice scores of 0.945 (LVC), 0.908 (RVC) and 0.905 (LVM) in a
cross-validation over the training set (100 cases) and 0.950 (LVC), 0.923 (RVC)
and 0.911 (LVM) on the test set (50 cases). We report a classification accuracy
of 94% on a training set cross-validation and 92% on the test set. Our results
underpin the potential of machine learning methods for accurate, fast and
reproducible segmentation and computer-assisted diagnosis (CAD).
| [
{
"created": "Mon, 3 Jul 2017 15:10:30 GMT",
"version": "v1"
},
{
"created": "Thu, 25 Jan 2018 08:15:24 GMT",
"version": "v2"
}
] | 2018-04-06 | [
[
"Isensee",
"Fabian",
""
],
[
"Jaeger",
"Paul",
""
],
[
"Full",
"Peter M.",
""
],
[
"Wolf",
"Ivo",
""
],
[
"Engelhardt",
"Sandy",
""
],
[
"Maier-Hein",
"Klaus H.",
""
]
] | Cardiac magnetic resonance imaging improves on diagnosis of cardiovascular diseases by providing images at high spatiotemporal resolution. Manual evaluation of these time-series, however, is expensive and prone to biased and non-reproducible outcomes. In this paper, we present a method that addresses named limitations by integrating segmentation and disease classification into a fully automatic processing pipeline. We use an ensemble of UNet inspired architectures for segmentation of cardiac structures such as the left and right ventricular cavity (LVC, RVC) and the left ventricular myocardium (LVM) on each time instance of the cardiac cycle. For the classification task, information is extracted from the segmented time-series in form of comprehensive features handcrafted to reflect diagnostic clinical procedures. Based on these features we train an ensemble of heavily regularized multilayer perceptrons (MLP) and a random forest classifier to predict the pathologic target class. We evaluated our method on the ACDC dataset (4 pathology groups, 1 healthy group) and achieve dice scores of 0.945 (LVC), 0.908 (RVC) and 0.905 (LVM) in a cross-validation over the training set (100 cases) and 0.950 (LVC), 0.923 (RVC) and 0.911 (LVM) on the test set (50 cases). We report a classification accuracy of 94% on a training set cross-validation and 92% on the test set. Our results underpin the potential of machine learning methods for accurate, fast and reproducible segmentation and computer-assisted diagnosis (CAD). |
1503.07376 | Rafael Cisneros | R. Cisneros, M. Pirro, G. Bergna, R. Ortega, G. Ippoliti, M. Molinas | Global Tracking Passivity--based PI Control of Bilinear Systems and its
Application to the Boost and Modular Multilevel Converters | 9 pages, 10 figures | null | null | null | cs.SY | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | This paper deals with the problem of trajectory tracking of a class of
bilinear systems with time--varying measurable disturbance. A set of matrices
{A,B_i} has been identified, via a linear matrix inequality, for which it is
possible to ensure global tracking of (admissible, differentiable) trajectories
with a simple linear time--varying PI controller. Instrumental to establish the
result is the construction of an output signal with respect to which the
incremental model is passive. The result is applied to the boost and the
modular multilevel converter for which experimental results are given.
| [
{
"created": "Wed, 25 Mar 2015 13:53:00 GMT",
"version": "v1"
}
] | 2015-03-26 | [
[
"Cisneros",
"R.",
""
],
[
"Pirro",
"M.",
""
],
[
"Bergna",
"G.",
""
],
[
"Ortega",
"R.",
""
],
[
"Ippoliti",
"G.",
""
],
[
"Molinas",
"M.",
""
]
] | This paper deals with the problem of trajectory tracking of a class of bilinear systems with time--varying measurable disturbance. A set of matrices {A,B_i} has been identified, via a linear matrix inequality, for which it is possible to ensure global tracking of (admissible, differentiable) trajectories with a simple linear time--varying PI controller. Instrumental to establish the result is the construction of an output signal with respect to which the incremental model is passive. The result is applied to the boost and the modular multilevel converter for which experimental results are given. |
2305.12542 | Zachary Yang | Zachary Yang, Yasmine Maricar, MohammadReza Davari, Nicolas
Grenon-Godbout, Reihaneh Rabbany | ToxBuster: In-game Chat Toxicity Buster with BERT | 11 pages, 3 figures | null | null | null | cs.CL cs.CY | http://creativecommons.org/licenses/by-sa/4.0/ | Detecting toxicity in online spaces is challenging and an ever more pressing
problem given the increase in social media and gaming consumption. We introduce
ToxBuster, a simple and scalable model trained on a relatively large dataset of
194k lines of game chat from Rainbow Six Siege and For Honor, carefully
annotated for different kinds of toxicity. Compared to the existing
state-of-the-art, ToxBuster achieves 82.95% (+7) in precision and 83.56% (+57)
in recall. This improvement is obtained by leveraging past chat history and
metadata. We also study the implication towards real-time and post-game
moderation as well as the model transferability from one game to another.
| [
{
"created": "Sun, 21 May 2023 18:53:26 GMT",
"version": "v1"
}
] | 2023-05-24 | [
[
"Yang",
"Zachary",
""
],
[
"Maricar",
"Yasmine",
""
],
[
"Davari",
"MohammadReza",
""
],
[
"Grenon-Godbout",
"Nicolas",
""
],
[
"Rabbany",
"Reihaneh",
""
]
] | Detecting toxicity in online spaces is challenging and an ever more pressing problem given the increase in social media and gaming consumption. We introduce ToxBuster, a simple and scalable model trained on a relatively large dataset of 194k lines of game chat from Rainbow Six Siege and For Honor, carefully annotated for different kinds of toxicity. Compared to the existing state-of-the-art, ToxBuster achieves 82.95% (+7) in precision and 83.56% (+57) in recall. This improvement is obtained by leveraging past chat history and metadata. We also study the implication towards real-time and post-game moderation as well as the model transferability from one game to another. |
2007.10871 | Christian Hesch | Maik Dittman, Jonathan Schult, Felix Schmidt and Christian Hesch | A strain-gradient formulation for fiber reinforced polymers: Hybrid
phase-field model for porous-ductile fracture | null | null | 10.1007/s00466-021-02018-0 | null | cs.CE | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | A novel numerical approach to analyze the mechanical behavior within
composite materials including the inelastic regime up to final failure is
presented. Therefore, a second-gradient theory is combined with phase-field
methods to fracture. In particular, we assume that the polymeric matrix
material undergoes ductile fracture, whereas continuously embedded fibers
undergo brittle fracture as it is typical e.g. for roving glass reinforced
thermoplastics. A hybrid phase-field approach is developed and applied along
with a modified Gurson-Tvergaard-Needelman GTN-type plasticity model accounting
for a temperature-dependent growth of voids on microscale. The mechanical
response of the arising microstructure of the woven fabric gives rise to
additional higher-order terms, representing homogenized bending contributions
of the fibers. Eventually, a series of tests is conducted for this physically
comprehensive multifield formulation to investigate different kinds and
sequences of failure within long fiber reinforced polymers.
| [
{
"created": "Tue, 21 Jul 2020 14:55:50 GMT",
"version": "v1"
},
{
"created": "Mon, 19 Apr 2021 11:11:24 GMT",
"version": "v2"
}
] | 2021-04-20 | [
[
"Dittman",
"Maik",
""
],
[
"Schult",
"Jonathan",
""
],
[
"Schmidt",
"Felix",
""
],
[
"Hesch",
"Christian",
""
]
] | A novel numerical approach to analyze the mechanical behavior within composite materials including the inelastic regime up to final failure is presented. Therefore, a second-gradient theory is combined with phase-field methods to fracture. In particular, we assume that the polymeric matrix material undergoes ductile fracture, whereas continuously embedded fibers undergo brittle fracture as it is typical e.g. for roving glass reinforced thermoplastics. A hybrid phase-field approach is developed and applied along with a modified Gurson-Tvergaard-Needelman GTN-type plasticity model accounting for a temperature-dependent growth of voids on microscale. The mechanical response of the arising microstructure of the woven fabric gives rise to additional higher-order terms, representing homogenized bending contributions of the fibers. Eventually, a series of tests is conducted for this physically comprehensive multifield formulation to investigate different kinds and sequences of failure within long fiber reinforced polymers. |
2406.16501 | Alvaro Lopez Pellicer | Alvaro Lopez Pellicer, Kittipos Giatgong, Yi Li, Neeraj Suri, Plamen
Angelov | UNICAD: A Unified Approach for Attack Detection, Noise Reduction and
Novel Class Identification | null | null | null | null | cs.CV cs.AI cs.LG | http://creativecommons.org/licenses/by/4.0/ | As the use of Deep Neural Networks (DNNs) becomes pervasive, their
vulnerability to adversarial attacks and limitations in handling unseen classes
poses significant challenges. The state-of-the-art offers discrete solutions
aimed to tackle individual issues covering specific adversarial attack
scenarios, classification or evolving learning. However, real-world systems
need to be able to detect and recover from a wide range of adversarial attacks
without sacrificing classification accuracy and to flexibly act in {\bf unseen}
scenarios. In this paper, UNICAD, is proposed as a novel framework that
integrates a variety of techniques to provide an adaptive solution.
For the targeted image classification, UNICAD achieves accurate image
classification, detects unseen classes, and recovers from adversarial attacks
using Prototype and Similarity-based DNNs with denoising autoencoders. Our
experiments performed on the CIFAR-10 dataset highlight UNICAD's effectiveness
in adversarial mitigation and unseen class classification, outperforming
traditional models.
| [
{
"created": "Mon, 24 Jun 2024 10:10:03 GMT",
"version": "v1"
}
] | 2024-06-25 | [
[
"Pellicer",
"Alvaro Lopez",
""
],
[
"Giatgong",
"Kittipos",
""
],
[
"Li",
"Yi",
""
],
[
"Suri",
"Neeraj",
""
],
[
"Angelov",
"Plamen",
""
]
] | As the use of Deep Neural Networks (DNNs) becomes pervasive, their vulnerability to adversarial attacks and limitations in handling unseen classes poses significant challenges. The state-of-the-art offers discrete solutions aimed to tackle individual issues covering specific adversarial attack scenarios, classification or evolving learning. However, real-world systems need to be able to detect and recover from a wide range of adversarial attacks without sacrificing classification accuracy and to flexibly act in {\bf unseen} scenarios. In this paper, UNICAD, is proposed as a novel framework that integrates a variety of techniques to provide an adaptive solution. For the targeted image classification, UNICAD achieves accurate image classification, detects unseen classes, and recovers from adversarial attacks using Prototype and Similarity-based DNNs with denoising autoencoders. Our experiments performed on the CIFAR-10 dataset highlight UNICAD's effectiveness in adversarial mitigation and unseen class classification, outperforming traditional models. |
2401.14255 | Yumnah Hasan | Yumnah Hasan, Allan de Lima, Fatemeh Amerehi, Darian Reyes Fernandez
de Bulnes, Patrick Healy, and Conor Ryan | Interpretable Solutions for Breast Cancer Diagnosis with Grammatical
Evolution and Data Augmentation | null | null | null | null | cs.LG cs.NE | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Medical imaging diagnosis increasingly relies on Machine Learning (ML)
models. This is a task that is often hampered by severely imbalanced datasets,
where positive cases can be quite rare. Their use is further compromised by
their limited interpretability, which is becoming increasingly important. While
post-hoc interpretability techniques such as SHAP and LIME have been used with
some success on so-called black box models, the use of inherently
understandable models makes such endeavors more fruitful. This paper addresses
these issues by demonstrating how a relatively new synthetic data generation
technique, STEM, can be used to produce data to train models produced by
Grammatical Evolution (GE) that are inherently understandable. STEM is a
recently introduced combination of the Synthetic Minority Oversampling
Technique (SMOTE), Edited Nearest Neighbour (ENN), and Mixup; it has previously
been successfully used to tackle both between class and within class imbalance
issues. We test our technique on the Digital Database for Screening Mammography
(DDSM) and the Wisconsin Breast Cancer (WBC) datasets and compare Area Under
the Curve (AUC) results with an ensemble of the top three performing
classifiers from a set of eight standard ML classifiers with varying degrees of
interpretability. We demonstrate that the GE-derived models present the best
AUC while still maintaining interpretable solutions.
| [
{
"created": "Thu, 25 Jan 2024 15:45:28 GMT",
"version": "v1"
}
] | 2024-01-26 | [
[
"Hasan",
"Yumnah",
""
],
[
"de Lima",
"Allan",
""
],
[
"Amerehi",
"Fatemeh",
""
],
[
"de Bulnes",
"Darian Reyes Fernandez",
""
],
[
"Healy",
"Patrick",
""
],
[
"Ryan",
"Conor",
""
]
] | Medical imaging diagnosis increasingly relies on Machine Learning (ML) models. This is a task that is often hampered by severely imbalanced datasets, where positive cases can be quite rare. Their use is further compromised by their limited interpretability, which is becoming increasingly important. While post-hoc interpretability techniques such as SHAP and LIME have been used with some success on so-called black box models, the use of inherently understandable models makes such endeavors more fruitful. This paper addresses these issues by demonstrating how a relatively new synthetic data generation technique, STEM, can be used to produce data to train models produced by Grammatical Evolution (GE) that are inherently understandable. STEM is a recently introduced combination of the Synthetic Minority Oversampling Technique (SMOTE), Edited Nearest Neighbour (ENN), and Mixup; it has previously been successfully used to tackle both between class and within class imbalance issues. We test our technique on the Digital Database for Screening Mammography (DDSM) and the Wisconsin Breast Cancer (WBC) datasets and compare Area Under the Curve (AUC) results with an ensemble of the top three performing classifiers from a set of eight standard ML classifiers with varying degrees of interpretability. We demonstrate that the GE-derived models present the best AUC while still maintaining interpretable solutions. |
2402.19231 | Feng Lu | Feng Lu, Xiangyuan Lan, Lijun Zhang, Dongmei Jiang, Yaowei Wang, Chun
Yuan | CricaVPR: Cross-image Correlation-aware Representation Learning for
Visual Place Recognition | Accepted by CVPR2024 | null | null | null | cs.CV cs.RO | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Over the past decade, most methods in visual place recognition (VPR) have
used neural networks to produce feature representations. These networks
typically produce a global representation of a place image using only this
image itself and neglect the cross-image variations (e.g. viewpoint and
illumination), which limits their robustness in challenging scenes. In this
paper, we propose a robust global representation method with cross-image
correlation awareness for VPR, named CricaVPR. Our method uses the attention
mechanism to correlate multiple images within a batch. These images can be
taken in the same place with different conditions or viewpoints, or even
captured from different places. Therefore, our method can utilize the
cross-image variations as a cue to guide the representation learning, which
ensures more robust features are produced. To further facilitate the
robustness, we propose a multi-scale convolution-enhanced adaptation method to
adapt pre-trained visual foundation models to the VPR task, which introduces
the multi-scale local information to further enhance the cross-image
correlation-aware representation. Experimental results show that our method
outperforms state-of-the-art methods by a large margin with significantly less
training time. The code is released at https://github.com/Lu-Feng/CricaVPR.
| [
{
"created": "Thu, 29 Feb 2024 15:05:11 GMT",
"version": "v1"
},
{
"created": "Mon, 1 Apr 2024 13:16:01 GMT",
"version": "v2"
}
] | 2024-04-02 | [
[
"Lu",
"Feng",
""
],
[
"Lan",
"Xiangyuan",
""
],
[
"Zhang",
"Lijun",
""
],
[
"Jiang",
"Dongmei",
""
],
[
"Wang",
"Yaowei",
""
],
[
"Yuan",
"Chun",
""
]
] | Over the past decade, most methods in visual place recognition (VPR) have used neural networks to produce feature representations. These networks typically produce a global representation of a place image using only this image itself and neglect the cross-image variations (e.g. viewpoint and illumination), which limits their robustness in challenging scenes. In this paper, we propose a robust global representation method with cross-image correlation awareness for VPR, named CricaVPR. Our method uses the attention mechanism to correlate multiple images within a batch. These images can be taken in the same place with different conditions or viewpoints, or even captured from different places. Therefore, our method can utilize the cross-image variations as a cue to guide the representation learning, which ensures more robust features are produced. To further facilitate the robustness, we propose a multi-scale convolution-enhanced adaptation method to adapt pre-trained visual foundation models to the VPR task, which introduces the multi-scale local information to further enhance the cross-image correlation-aware representation. Experimental results show that our method outperforms state-of-the-art methods by a large margin with significantly less training time. The code is released at https://github.com/Lu-Feng/CricaVPR. |
1310.3252 | Robert Krauthgamer | Alexandr Andoni, Anupam Gupta, Robert Krauthgamer | Towards (1+\epsilon)-Approximate Flow Sparsifiers | Full version of a paper accepted to SODA 2014 | null | null | null | cs.DS math.CO | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | A useful approach to "compress" a large network $G$ is to represent it with a
{\em flow-sparsifier}, i.e., a small network $H$ that supports the same flows
as $G$, up to a factor $q \geq 1$ called the quality of sparsifier.
Specifically, we assume the network $G$ contains a set of $k$ terminals $T$,
shared with the network $H$, i.e., $T\subseteq V(G)\cap V(H)$, and we want $H$
to preserve all multicommodity flows that can be routed between the terminals
$T$. The challenge is to construct $H$ that is small.
These questions have received a lot of attention in recent years, leading to
some known tradeoffs between the sparsifier's quality $q$ and its size
$|V(H)|$. Nevertheless, it remains an outstanding question whether every $G$
admits a flow-sparsifier $H$ with quality $q=1+\epsilon$, or even $q=O(1)$, and
size $|V(H)|\leq f(k,\epsilon)$ (in particular, independent of $|V(G)|$ and the
edge capacities). Making a first step in this direction, we present new
constructions for several scenarios:
* Our main result is that for quasi-bipartite networks $G$, one can construct
a $(1+\epsilon)$-flow-sparsifier of size $\poly(k/\eps)$. In contrast, exact
($q=1$) sparsifiers for this family of networks are known to require size
$2^{\Omega(k)}$.
* For networks $G$ of bounded treewidth $w$, we construct a flow-sparsifier
with quality $q=O(\log w / \log\log w)$ and size $O(w\cdot \poly(k))$.
* For general networks $G$, we construct a {\em sketch} $sk(G)$, that stores
all the feasible multicommodity flows up to factor $q=1+\eps$, and its size
(storage requirement) is $f(k,\epsilon)$.
| [
{
"created": "Fri, 11 Oct 2013 19:23:21 GMT",
"version": "v1"
}
] | 2013-10-14 | [
[
"Andoni",
"Alexandr",
""
],
[
"Gupta",
"Anupam",
""
],
[
"Krauthgamer",
"Robert",
""
]
] | A useful approach to "compress" a large network $G$ is to represent it with a {\em flow-sparsifier}, i.e., a small network $H$ that supports the same flows as $G$, up to a factor $q \geq 1$ called the quality of sparsifier. Specifically, we assume the network $G$ contains a set of $k$ terminals $T$, shared with the network $H$, i.e., $T\subseteq V(G)\cap V(H)$, and we want $H$ to preserve all multicommodity flows that can be routed between the terminals $T$. The challenge is to construct $H$ that is small. These questions have received a lot of attention in recent years, leading to some known tradeoffs between the sparsifier's quality $q$ and its size $|V(H)|$. Nevertheless, it remains an outstanding question whether every $G$ admits a flow-sparsifier $H$ with quality $q=1+\epsilon$, or even $q=O(1)$, and size $|V(H)|\leq f(k,\epsilon)$ (in particular, independent of $|V(G)|$ and the edge capacities). Making a first step in this direction, we present new constructions for several scenarios: * Our main result is that for quasi-bipartite networks $G$, one can construct a $(1+\epsilon)$-flow-sparsifier of size $\poly(k/\eps)$. In contrast, exact ($q=1$) sparsifiers for this family of networks are known to require size $2^{\Omega(k)}$. * For networks $G$ of bounded treewidth $w$, we construct a flow-sparsifier with quality $q=O(\log w / \log\log w)$ and size $O(w\cdot \poly(k))$. * For general networks $G$, we construct a {\em sketch} $sk(G)$, that stores all the feasible multicommodity flows up to factor $q=1+\eps$, and its size (storage requirement) is $f(k,\epsilon)$. |
2108.01358 | Jakob Karalus | Jakob Karalus, Felix Lindner | Accelerating the Learning of TAMER with Counterfactual Explanations | null | null | null | null | cs.AI cs.LG | http://creativecommons.org/licenses/by/4.0/ | The capability to interactively learn from human feedback would enable agents
in new settings. For example, even novice users could train service robots in
new tasks naturally and interactively. Human-in-the-loop Reinforcement Learning
(HRL) combines human feedback and Reinforcement Learning (RL) techniques.
State-of-the-art interactive learning techniques suffer from slow learning
speed, thus leading to a frustrating experience for the human. We approach this
problem by extending the HRL framework TAMER for evaluative feedback with the
possibility to enhance human feedback with two different types of
counterfactual explanations (action and state based). We experimentally show
that our extensions improve the speed of learning.
| [
{
"created": "Tue, 3 Aug 2021 08:27:28 GMT",
"version": "v1"
},
{
"created": "Wed, 27 Jul 2022 07:59:22 GMT",
"version": "v2"
}
] | 2022-07-28 | [
[
"Karalus",
"Jakob",
""
],
[
"Lindner",
"Felix",
""
]
] | The capability to interactively learn from human feedback would enable agents in new settings. For example, even novice users could train service robots in new tasks naturally and interactively. Human-in-the-loop Reinforcement Learning (HRL) combines human feedback and Reinforcement Learning (RL) techniques. State-of-the-art interactive learning techniques suffer from slow learning speed, thus leading to a frustrating experience for the human. We approach this problem by extending the HRL framework TAMER for evaluative feedback with the possibility to enhance human feedback with two different types of counterfactual explanations (action and state based). We experimentally show that our extensions improve the speed of learning. |
1803.09179 | Matthias Nie{\ss}ner | Andreas R\"ossler, Davide Cozzolino, Luisa Verdoliva, Christian Riess,
Justus Thies, Matthias Nie{\ss}ner | FaceForensics: A Large-scale Video Dataset for Forgery Detection in
Human Faces | Video: https://youtu.be/Tle7YaPkO_k | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | With recent advances in computer vision and graphics, it is now possible to
generate videos with extremely realistic synthetic faces, even in real time.
Countless applications are possible, some of which raise a legitimate alarm,
calling for reliable detectors of fake videos. In fact, distinguishing between
original and manipulated video can be a challenge for humans and computers
alike, especially when the videos are compressed or have low resolution, as it
often happens on social networks. Research on the detection of face
manipulations has been seriously hampered by the lack of adequate datasets. To
this end, we introduce a novel face manipulation dataset of about half a
million edited images (from over 1000 videos). The manipulations have been
generated with a state-of-the-art face editing approach. It exceeds all
existing video manipulation datasets by at least an order of magnitude. Using
our new dataset, we introduce benchmarks for classical image forensic tasks,
including classification and segmentation, considering videos compressed at
various quality levels. In addition, we introduce a benchmark evaluation for
creating indistinguishable forgeries with known ground truth; for instance with
generative refinement models.
| [
{
"created": "Sat, 24 Mar 2018 23:12:44 GMT",
"version": "v1"
}
] | 2018-03-28 | [
[
"Rössler",
"Andreas",
""
],
[
"Cozzolino",
"Davide",
""
],
[
"Verdoliva",
"Luisa",
""
],
[
"Riess",
"Christian",
""
],
[
"Thies",
"Justus",
""
],
[
"Nießner",
"Matthias",
""
]
] | With recent advances in computer vision and graphics, it is now possible to generate videos with extremely realistic synthetic faces, even in real time. Countless applications are possible, some of which raise a legitimate alarm, calling for reliable detectors of fake videos. In fact, distinguishing between original and manipulated video can be a challenge for humans and computers alike, especially when the videos are compressed or have low resolution, as it often happens on social networks. Research on the detection of face manipulations has been seriously hampered by the lack of adequate datasets. To this end, we introduce a novel face manipulation dataset of about half a million edited images (from over 1000 videos). The manipulations have been generated with a state-of-the-art face editing approach. It exceeds all existing video manipulation datasets by at least an order of magnitude. Using our new dataset, we introduce benchmarks for classical image forensic tasks, including classification and segmentation, considering videos compressed at various quality levels. In addition, we introduce a benchmark evaluation for creating indistinguishable forgeries with known ground truth; for instance with generative refinement models. |
2405.01392 | David Maranto | David Maranto | LLMSat: A Large Language Model-Based Goal-Oriented Agent for Autonomous
Space Exploration | B.A.Sc thesis | null | null | null | cs.RO cs.AI cs.LG cs.MA physics.space-ph | http://creativecommons.org/licenses/by/4.0/ | As spacecraft journey further from Earth with more complex missions, systems
of greater autonomy and onboard intelligence are called for. Reducing reliance
on human-based mission control becomes increasingly critical if we are to
increase our rate of solar-system-wide exploration. Recent work has explored
AI-based goal-oriented systems to increase the level of autonomy in mission
execution. These systems make use of symbolic reasoning managers to make
inferences from the state of a spacecraft and a handcrafted knowledge base,
enabling autonomous generation of tasks and re-planning. Such systems have
proven to be successful in controlled cases, but they are difficult to
implement as they require human-crafted ontological models to allow the
spacecraft to understand the world. Reinforcement learning has been applied to
train robotic agents to pursue a goal. A new architecture for autonomy is
called for. This work explores the application of Large Language Models (LLMs)
as the high-level control system of a spacecraft. Using a systems engineering
approach, this work presents the design and development of an agentic
spacecraft controller by leveraging an LLM as a reasoning engine, to evaluate
the utility of such an architecture in achieving higher levels of spacecraft
autonomy. A series of deep space mission scenarios simulated within the popular
game engine Kerbal Space Program (KSP) are used as case studies to evaluate the
implementation against the requirements. It is shown the reasoning and planning
abilities of present-day LLMs do not scale well as the complexity of a mission
increases, but this can be alleviated with adequate prompting frameworks and
strategic selection of the agent's level of authority over the host spacecraft.
This research evaluates the potential of LLMs in augmenting autonomous
decision-making systems for future robotic space applications.
| [
{
"created": "Sat, 13 Apr 2024 03:33:17 GMT",
"version": "v1"
}
] | 2024-05-03 | [
[
"Maranto",
"David",
""
]
] | As spacecraft journey further from Earth with more complex missions, systems of greater autonomy and onboard intelligence are called for. Reducing reliance on human-based mission control becomes increasingly critical if we are to increase our rate of solar-system-wide exploration. Recent work has explored AI-based goal-oriented systems to increase the level of autonomy in mission execution. These systems make use of symbolic reasoning managers to make inferences from the state of a spacecraft and a handcrafted knowledge base, enabling autonomous generation of tasks and re-planning. Such systems have proven to be successful in controlled cases, but they are difficult to implement as they require human-crafted ontological models to allow the spacecraft to understand the world. Reinforcement learning has been applied to train robotic agents to pursue a goal. A new architecture for autonomy is called for. This work explores the application of Large Language Models (LLMs) as the high-level control system of a spacecraft. Using a systems engineering approach, this work presents the design and development of an agentic spacecraft controller by leveraging an LLM as a reasoning engine, to evaluate the utility of such an architecture in achieving higher levels of spacecraft autonomy. A series of deep space mission scenarios simulated within the popular game engine Kerbal Space Program (KSP) are used as case studies to evaluate the implementation against the requirements. It is shown the reasoning and planning abilities of present-day LLMs do not scale well as the complexity of a mission increases, but this can be alleviated with adequate prompting frameworks and strategic selection of the agent's level of authority over the host spacecraft. This research evaluates the potential of LLMs in augmenting autonomous decision-making systems for future robotic space applications. |
1410.6277 | Sven Banisch | Sven Banisch | The Probabilistic Structure of Discrete Agent-Based Models | null | Discontinuity, Nonlinearity, and Complexity, 3:281--292, 2014 | null | null | cs.MA cs.CY nlin.AO physics.comp-ph physics.soc-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | This paper describes a formalization of agent-based models (ABMs) as random
walks on regular graphs and relates the symmetry group of those graphs to a
coarse-graining of the ABM that is still Markovian. An ABM in which $N$ agents
can be in $\delta$ different states leads to a Markov chain with $\delta^N$
states. In ABMs with a sequential update scheme by which one agent is chosen to
update its state at a time, transitions are only allowed between system
configurations that differ with respect to a single agent. This characterizes
ABMs as random walks on regular graphs. The non-trivial automorphisms of those
graphs make visible the dynamical symmetries that an ABM gives rise to because
sets of micro configurations can be interchanged without changing the
probability structure of the random walk. This allows for a systematic
loss-less reduction of the state space of the model.
| [
{
"created": "Thu, 23 Oct 2014 08:01:35 GMT",
"version": "v1"
}
] | 2014-10-24 | [
[
"Banisch",
"Sven",
""
]
] | This paper describes a formalization of agent-based models (ABMs) as random walks on regular graphs and relates the symmetry group of those graphs to a coarse-graining of the ABM that is still Markovian. An ABM in which $N$ agents can be in $\delta$ different states leads to a Markov chain with $\delta^N$ states. In ABMs with a sequential update scheme by which one agent is chosen to update its state at a time, transitions are only allowed between system configurations that differ with respect to a single agent. This characterizes ABMs as random walks on regular graphs. The non-trivial automorphisms of those graphs make visible the dynamical symmetries that an ABM gives rise to because sets of micro configurations can be interchanged without changing the probability structure of the random walk. This allows for a systematic loss-less reduction of the state space of the model. |
2204.13515 | Abdellah El Mekki | Abdellah El Mekki and Abdelkader El Mahdaouy and Mohammed Akallouch
and Ismail Berrada and Ahmed Khoumsi | UM6P-CS at SemEval-2022 Task 11: Enhancing Multilingual and Code-Mixed
Complex Named Entity Recognition via Pseudo Labels using Multilingual
Transformer | null | null | null | null | cs.CL | http://creativecommons.org/licenses/by/4.0/ | Building real-world complex Named Entity Recognition (NER) systems is a
challenging task. This is due to the complexity and ambiguity of named entities
that appear in various contexts such as short input sentences, emerging
entities, and complex entities. Besides, real-world queries are mostly
malformed, as they can be code-mixed or multilingual, among other scenarios. In
this paper, we introduce our submitted system to the Multilingual Complex Named
Entity Recognition (MultiCoNER) shared task. We approach the complex NER for
multilingual and code-mixed queries, by relying on the contextualized
representation provided by the multilingual Transformer XLM-RoBERTa. In
addition to the CRF-based token classification layer, we incorporate a span
classification loss to recognize named entities spans. Furthermore, we use a
self-training mechanism to generate weakly-annotated data from a large
unlabeled dataset. Our proposed system is ranked 6th and 8th in the
multilingual and code-mixed MultiCoNER's tracks respectively.
| [
{
"created": "Thu, 28 Apr 2022 14:07:06 GMT",
"version": "v1"
}
] | 2022-04-29 | [
[
"Mekki",
"Abdellah El",
""
],
[
"Mahdaouy",
"Abdelkader El",
""
],
[
"Akallouch",
"Mohammed",
""
],
[
"Berrada",
"Ismail",
""
],
[
"Khoumsi",
"Ahmed",
""
]
] | Building real-world complex Named Entity Recognition (NER) systems is a challenging task. This is due to the complexity and ambiguity of named entities that appear in various contexts such as short input sentences, emerging entities, and complex entities. Besides, real-world queries are mostly malformed, as they can be code-mixed or multilingual, among other scenarios. In this paper, we introduce our submitted system to the Multilingual Complex Named Entity Recognition (MultiCoNER) shared task. We approach the complex NER for multilingual and code-mixed queries, by relying on the contextualized representation provided by the multilingual Transformer XLM-RoBERTa. In addition to the CRF-based token classification layer, we incorporate a span classification loss to recognize named entities spans. Furthermore, we use a self-training mechanism to generate weakly-annotated data from a large unlabeled dataset. Our proposed system is ranked 6th and 8th in the multilingual and code-mixed MultiCoNER's tracks respectively. |
1612.05626 | Biplav Srivastava | Biplav Srivastava, Sandeep Sandha, Vaskar Raychoudhury, Sukanya
Randhawa, Viral Kapoor, Anmol Agrawal | An Open, Multi-Sensor, Dataset of Water Pollution of Ganga Basin and its
Application to Understand Impact of Large Religious Gathering | 7 pages | null | null | null | cs.CY | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Water is a crucial pre-requisite for all human activities. Due to growing
demand from population and shrinking supply of potable water, there is an
urgent need to use computational methods to manage available water
intelligently, and especially in developing countries like India where even
basic data to track water availability or physical infrastructure to process
water are inadequate. In this context, we present a dataset of water pollution
containing quantitative and qualitative data from a combination for modalities
- real-time sensors, lab results, and estimates from people using mobile apps.
The data on our API-accessible cloud platform covers more than 60 locations and
consists of both what we have ourselves collected from multiple location
following a novel process, and from others (lab-results) which were open but
hither-to difficult to access. Further, we discuss an application of released
data to understand spatio-temporal pollution impact of a large event with
hundreds of millions of people converging on a river during a religious
gathering (Ardh Khumbh 2016) spread over months. Such unprecedented details can
help authorities manage an ongoing event or plan for future ones. The community
can use the data for any application and also contribute new data to the
platform.
| [
{
"created": "Sun, 20 Nov 2016 01:45:36 GMT",
"version": "v1"
}
] | 2016-12-19 | [
[
"Srivastava",
"Biplav",
""
],
[
"Sandha",
"Sandeep",
""
],
[
"Raychoudhury",
"Vaskar",
""
],
[
"Randhawa",
"Sukanya",
""
],
[
"Kapoor",
"Viral",
""
],
[
"Agrawal",
"Anmol",
""
]
] | Water is a crucial pre-requisite for all human activities. Due to growing demand from population and shrinking supply of potable water, there is an urgent need to use computational methods to manage available water intelligently, and especially in developing countries like India where even basic data to track water availability or physical infrastructure to process water are inadequate. In this context, we present a dataset of water pollution containing quantitative and qualitative data from a combination for modalities - real-time sensors, lab results, and estimates from people using mobile apps. The data on our API-accessible cloud platform covers more than 60 locations and consists of both what we have ourselves collected from multiple location following a novel process, and from others (lab-results) which were open but hither-to difficult to access. Further, we discuss an application of released data to understand spatio-temporal pollution impact of a large event with hundreds of millions of people converging on a river during a religious gathering (Ardh Khumbh 2016) spread over months. Such unprecedented details can help authorities manage an ongoing event or plan for future ones. The community can use the data for any application and also contribute new data to the platform. |
1506.01501 | Nematollah Zarmehi | Nematollah Zarmehi, Morteza Banagar, Mohammad Ali Akhaee | Optimum Decoder for an Additive Video Watermarking with Laplacian Noise
in H.264 | null | 2013 10th International ISC Conference on Information Security and
Cryptology (ISCISC),Aug. 2013, pp. 1-5 | 10.1109/ISCISC.2013.6767352 | null | cs.MM | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In this paper, we investigate an additive video watermarking method in H.264
standard in presence of the Laplacian noise. In some applications, due to the
loss of some pixels or a region of a frame, we resort to Laplacian noise rather
than Gaussian one. The embedding is performed in the transform domain; while an
optimum and a sub-optimum decoder are derived for the proposed Laplacian model.
Simulation results show that the proposed watermarking scheme has suitable
performance with enough transparency required for watermarking applications.
| [
{
"created": "Thu, 4 Jun 2015 08:10:13 GMT",
"version": "v1"
}
] | 2015-06-05 | [
[
"Zarmehi",
"Nematollah",
""
],
[
"Banagar",
"Morteza",
""
],
[
"Akhaee",
"Mohammad Ali",
""
]
] | In this paper, we investigate an additive video watermarking method in H.264 standard in presence of the Laplacian noise. In some applications, due to the loss of some pixels or a region of a frame, we resort to Laplacian noise rather than Gaussian one. The embedding is performed in the transform domain; while an optimum and a sub-optimum decoder are derived for the proposed Laplacian model. Simulation results show that the proposed watermarking scheme has suitable performance with enough transparency required for watermarking applications. |
1507.02674 | David Harris | David G. Harris, Aravind Srinivasan | Algorithmic and enumerative aspects of the Moser-Tardos distribution | null | ACM Transactions on Algorithms 13(3), Article #33 (2017) | null | null | cs.DM math.CO | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Moser & Tardos have developed a powerful algorithmic approach (henceforth
"MT") to the Lovasz Local Lemma (LLL); the basic operation done in MT and its
variants is a search for "bad" events in a current configuration. In the
initial stage of MT, the variables are set independently. We examine the
distributions on these variables which arise during intermediate stages of MT.
We show that these configurations have a more or less "random" form, building
further on the "MT-distribution" concept of Haeupler et al. in understanding
the (intermediate and) output distribution of MT. This has a variety of
algorithmic applications; the most important is that bad events can be found
relatively quickly, improving upon MT across the complexity spectrum: it makes
some polynomial-time algorithms sub-linear (e.g., for Latin transversals, which
are of basic combinatorial interest), gives lower-degree polynomial run-times
in some settings, transforms certain super-polynomial-time algorithms into
polynomial-time ones, and leads to Las Vegas algorithms for some coloring
problems for which only Monte Carlo algorithms were known.
We show that in certain conditions when the LLL condition is violated, a
variant of the MT algorithm can still produce a distribution which avoids most
of the bad events. We show in some cases this MT variant can run faster than
the original MT algorithm itself, and develop the first-known criterion for the
case of the asymmetric LLL. This can be used to find partial Latin transversals
-- improving upon earlier bounds of Stein (1975) -- among other applications.
We furthermore give applications in enumeration, showing that most applications
(where we aim for all or most of the bad events to be avoided) have many more
solutions than known before by proving that the MT-distribution has "large"
min-entropy and hence that its support-size is large.
| [
{
"created": "Thu, 9 Jul 2015 19:54:36 GMT",
"version": "v1"
},
{
"created": "Mon, 12 Sep 2016 17:54:20 GMT",
"version": "v2"
},
{
"created": "Thu, 8 Dec 2016 14:11:34 GMT",
"version": "v3"
},
{
"created": "Thu, 16 Feb 2017 16:02:07 GMT",
"version": "v4"
}
] | 2023-10-13 | [
[
"Harris",
"David G.",
""
],
[
"Srinivasan",
"Aravind",
""
]
] | Moser & Tardos have developed a powerful algorithmic approach (henceforth "MT") to the Lovasz Local Lemma (LLL); the basic operation done in MT and its variants is a search for "bad" events in a current configuration. In the initial stage of MT, the variables are set independently. We examine the distributions on these variables which arise during intermediate stages of MT. We show that these configurations have a more or less "random" form, building further on the "MT-distribution" concept of Haeupler et al. in understanding the (intermediate and) output distribution of MT. This has a variety of algorithmic applications; the most important is that bad events can be found relatively quickly, improving upon MT across the complexity spectrum: it makes some polynomial-time algorithms sub-linear (e.g., for Latin transversals, which are of basic combinatorial interest), gives lower-degree polynomial run-times in some settings, transforms certain super-polynomial-time algorithms into polynomial-time ones, and leads to Las Vegas algorithms for some coloring problems for which only Monte Carlo algorithms were known. We show that in certain conditions when the LLL condition is violated, a variant of the MT algorithm can still produce a distribution which avoids most of the bad events. We show in some cases this MT variant can run faster than the original MT algorithm itself, and develop the first-known criterion for the case of the asymmetric LLL. This can be used to find partial Latin transversals -- improving upon earlier bounds of Stein (1975) -- among other applications. We furthermore give applications in enumeration, showing that most applications (where we aim for all or most of the bad events to be avoided) have many more solutions than known before by proving that the MT-distribution has "large" min-entropy and hence that its support-size is large. |
1705.05344 | Gilwoo Lee | Gilwoo Lee, Siddhartha S. Srinivasa, Matthew T. Mason | GP-ILQG: Data-driven Robust Optimal Control for Uncertain Nonlinear
Dynamical Systems | null | null | null | null | cs.RO cs.SY | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | As we aim to control complex systems, use of a simulator in model-based
reinforcement learning is becoming more common. However, it has been
challenging to overcome the Reality Gap, which comes from nonlinear model bias
and susceptibility to disturbance. To address these problems, we propose a
novel algorithm that combines data-driven system identification approach
(Gaussian Process) with a Differential-Dynamic-Programming-based robust optimal
control method (Iterative Linear Quadratic Control). Our algorithm uses the
simulator's model as the mean function for a Gaussian Process and learns only
the difference between the simulator's prediction and actual observations,
making it a natural hybrid of simulation and real-world observation. We show
that our approach quickly corrects incorrect models, comes up with robust
optimal controllers, and transfers its acquired model knowledge to new tasks
efficiently.
| [
{
"created": "Mon, 15 May 2017 17:29:25 GMT",
"version": "v1"
}
] | 2017-05-16 | [
[
"Lee",
"Gilwoo",
""
],
[
"Srinivasa",
"Siddhartha S.",
""
],
[
"Mason",
"Matthew T.",
""
]
] | As we aim to control complex systems, use of a simulator in model-based reinforcement learning is becoming more common. However, it has been challenging to overcome the Reality Gap, which comes from nonlinear model bias and susceptibility to disturbance. To address these problems, we propose a novel algorithm that combines data-driven system identification approach (Gaussian Process) with a Differential-Dynamic-Programming-based robust optimal control method (Iterative Linear Quadratic Control). Our algorithm uses the simulator's model as the mean function for a Gaussian Process and learns only the difference between the simulator's prediction and actual observations, making it a natural hybrid of simulation and real-world observation. We show that our approach quickly corrects incorrect models, comes up with robust optimal controllers, and transfers its acquired model knowledge to new tasks efficiently. |
1605.09653 | Srikrishna Karanam | Srikrishna Karanam, Mengran Gou, Ziyan Wu, Angels Rates-Borras,
Octavia Camps, Richard J. Radke | A Systematic Evaluation and Benchmark for Person Re-Identification:
Features, Metrics, and Datasets | Preliminary work on person Re-Id benchmark. S. Karanam and M. Gou
contributed equally. 14 pages, 6 figures, 4 tables. For supplementary
material, see
http://robustsystems.coe.neu.edu/sites/robustsystems.coe.neu.edu/files/systems/supmat/ReID_benchmark_supp.zip | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Person re-identification (re-id) is a critical problem in video analytics
applications such as security and surveillance. The public release of several
datasets and code for vision algorithms has facilitated rapid progress in this
area over the last few years. However, directly comparing re-id algorithms
reported in the literature has become difficult since a wide variety of
features, experimental protocols, and evaluation metrics are employed. In order
to address this need, we present an extensive review and performance evaluation
of single- and multi-shot re-id algorithms. The experimental protocol
incorporates the most recent advances in both feature extraction and metric
learning. To ensure a fair comparison, all of the approaches were implemented
using a unified code library that includes 11 feature extraction algorithms and
22 metric learning and ranking techniques. All approaches were evaluated using
a new large-scale dataset that closely mimics a real-world problem setting, in
addition to 16 other publicly available datasets: VIPeR, GRID, CAVIAR,
DukeMTMC4ReID, 3DPeS, PRID, V47, WARD, SAIVT-SoftBio, CUHK01, CHUK02, CUHK03,
RAiD, iLIDSVID, HDA+ and Market1501. The evaluation codebase and results will
be made publicly available for community use.
| [
{
"created": "Tue, 31 May 2016 15:01:46 GMT",
"version": "v1"
},
{
"created": "Wed, 1 Jun 2016 05:55:46 GMT",
"version": "v2"
},
{
"created": "Tue, 29 Nov 2016 19:50:51 GMT",
"version": "v3"
},
{
"created": "Fri, 18 Aug 2017 03:39:58 GMT",
"version": "v4"
},
{
"created": "Wed, 14 Feb 2018 16:27:31 GMT",
"version": "v5"
}
] | 2018-02-15 | [
[
"Karanam",
"Srikrishna",
""
],
[
"Gou",
"Mengran",
""
],
[
"Wu",
"Ziyan",
""
],
[
"Rates-Borras",
"Angels",
""
],
[
"Camps",
"Octavia",
""
],
[
"Radke",
"Richard J.",
""
]
] | Person re-identification (re-id) is a critical problem in video analytics applications such as security and surveillance. The public release of several datasets and code for vision algorithms has facilitated rapid progress in this area over the last few years. However, directly comparing re-id algorithms reported in the literature has become difficult since a wide variety of features, experimental protocols, and evaluation metrics are employed. In order to address this need, we present an extensive review and performance evaluation of single- and multi-shot re-id algorithms. The experimental protocol incorporates the most recent advances in both feature extraction and metric learning. To ensure a fair comparison, all of the approaches were implemented using a unified code library that includes 11 feature extraction algorithms and 22 metric learning and ranking techniques. All approaches were evaluated using a new large-scale dataset that closely mimics a real-world problem setting, in addition to 16 other publicly available datasets: VIPeR, GRID, CAVIAR, DukeMTMC4ReID, 3DPeS, PRID, V47, WARD, SAIVT-SoftBio, CUHK01, CHUK02, CUHK03, RAiD, iLIDSVID, HDA+ and Market1501. The evaluation codebase and results will be made publicly available for community use. |
1903.05879 | Christian Graulund | Patrick Bahr, Christian Graulund, Rasmus M{\o}gelberg | Simply RaTT: A Fitch-style Modal Calculus for Reactive Programming
without Space Leaks | null | null | null | null | cs.PL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Functional reactive programming (FRP) is a paradigm for programming with
signals and events, allowing the user to describe reactive programs on a high
level of abstraction. For this to make sense, an FRP language must ensure that
all programs are causal, and can be implemented without introducing space leaks
and time leaks. To this end, some FRP languages do not give direct access to
signals, but just to signal functions.
Recently, modal types have been suggested as an alternative approach to
ensuring causality in FRP languages in the synchronous case, giving direct
access to the signal and event abstractions. This paper presents Simply RaTT, a
new modal calculus for reactive programming. Unlike prior calculi, Simply RaTT
uses a Fitch-style approach to modal types, which simplifies the type system
and makes programs more concise. Echoing a previous result by Krishnaswami for
a different language, we devise an operational semantics that safely executes
Simply RaTT programs without space leaks.
We also identify a source of time leaks present in other modal FRP languages:
The unfolding of fixed points in delayed computations. These time leaks are
eliminated by the Simply RaTT type system.
| [
{
"created": "Thu, 14 Mar 2019 09:50:15 GMT",
"version": "v1"
},
{
"created": "Tue, 11 Jun 2019 13:23:14 GMT",
"version": "v2"
}
] | 2019-06-12 | [
[
"Bahr",
"Patrick",
""
],
[
"Graulund",
"Christian",
""
],
[
"Møgelberg",
"Rasmus",
""
]
] | Functional reactive programming (FRP) is a paradigm for programming with signals and events, allowing the user to describe reactive programs on a high level of abstraction. For this to make sense, an FRP language must ensure that all programs are causal, and can be implemented without introducing space leaks and time leaks. To this end, some FRP languages do not give direct access to signals, but just to signal functions. Recently, modal types have been suggested as an alternative approach to ensuring causality in FRP languages in the synchronous case, giving direct access to the signal and event abstractions. This paper presents Simply RaTT, a new modal calculus for reactive programming. Unlike prior calculi, Simply RaTT uses a Fitch-style approach to modal types, which simplifies the type system and makes programs more concise. Echoing a previous result by Krishnaswami for a different language, we devise an operational semantics that safely executes Simply RaTT programs without space leaks. We also identify a source of time leaks present in other modal FRP languages: The unfolding of fixed points in delayed computations. These time leaks are eliminated by the Simply RaTT type system. |
2306.15880 | Jianzong Wu | Jianzong Wu, Xiangtai Li, Shilin Xu, Haobo Yuan, Henghui Ding, Yibo
Yang, Xia Li, Jiangning Zhang, Yunhai Tong, Xudong Jiang, Bernard Ghanem,
Dacheng Tao | Towards Open Vocabulary Learning: A Survey | Accepted by IEEE T-PAMI. Project page:
https://github.com/jianzongwu/Awesome-Open-Vocabulary | null | null | null | cs.CV cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In the field of visual scene understanding, deep neural networks have made
impressive advancements in various core tasks like segmentation, tracking, and
detection. However, most approaches operate on the close-set assumption,
meaning that the model can only identify pre-defined categories that are
present in the training set. Recently, open vocabulary settings were proposed
due to the rapid progress of vision language pre-training. These new approaches
seek to locate and recognize categories beyond the annotated label space. The
open vocabulary approach is more general, practical, and effective compared to
weakly supervised and zero-shot settings. This paper provides a thorough review
of open vocabulary learning, summarizing and analyzing recent developments in
the field. In particular, we begin by comparing it to related concepts such as
zero-shot learning, open-set recognition, and out-of-distribution detection.
Then, we review several closely related tasks in the case of segmentation and
detection, including long-tail problems, few-shot, and zero-shot settings. For
the method survey, we first present the basic knowledge of detection and
segmentation in close-set as the preliminary knowledge. Next, we examine
various scenarios in which open vocabulary learning is used, identifying common
design elements and core ideas. Then, we compare the recent detection and
segmentation approaches in commonly used datasets and benchmarks. Finally, we
conclude with insights, issues, and discussions regarding future research
directions. To our knowledge, this is the first comprehensive literature review
of open vocabulary learning. We keep tracing related works at
https://github.com/jianzongwu/Awesome-Open-Vocabulary.
| [
{
"created": "Wed, 28 Jun 2023 02:33:06 GMT",
"version": "v1"
},
{
"created": "Thu, 6 Jul 2023 10:45:39 GMT",
"version": "v2"
},
{
"created": "Sun, 23 Jul 2023 10:50:18 GMT",
"version": "v3"
},
{
"created": "Thu, 1 Feb 2024 08:31:59 GMT",
"version": "v4"
}
] | 2024-02-02 | [
[
"Wu",
"Jianzong",
""
],
[
"Li",
"Xiangtai",
""
],
[
"Xu",
"Shilin",
""
],
[
"Yuan",
"Haobo",
""
],
[
"Ding",
"Henghui",
""
],
[
"Yang",
"Yibo",
""
],
[
"Li",
"Xia",
""
],
[
"Zhang",
"Jiangning",
""
],
[
"Tong",
"Yunhai",
""
],
[
"Jiang",
"Xudong",
""
],
[
"Ghanem",
"Bernard",
""
],
[
"Tao",
"Dacheng",
""
]
] | In the field of visual scene understanding, deep neural networks have made impressive advancements in various core tasks like segmentation, tracking, and detection. However, most approaches operate on the close-set assumption, meaning that the model can only identify pre-defined categories that are present in the training set. Recently, open vocabulary settings were proposed due to the rapid progress of vision language pre-training. These new approaches seek to locate and recognize categories beyond the annotated label space. The open vocabulary approach is more general, practical, and effective compared to weakly supervised and zero-shot settings. This paper provides a thorough review of open vocabulary learning, summarizing and analyzing recent developments in the field. In particular, we begin by comparing it to related concepts such as zero-shot learning, open-set recognition, and out-of-distribution detection. Then, we review several closely related tasks in the case of segmentation and detection, including long-tail problems, few-shot, and zero-shot settings. For the method survey, we first present the basic knowledge of detection and segmentation in close-set as the preliminary knowledge. Next, we examine various scenarios in which open vocabulary learning is used, identifying common design elements and core ideas. Then, we compare the recent detection and segmentation approaches in commonly used datasets and benchmarks. Finally, we conclude with insights, issues, and discussions regarding future research directions. To our knowledge, this is the first comprehensive literature review of open vocabulary learning. We keep tracing related works at https://github.com/jianzongwu/Awesome-Open-Vocabulary. |
1106.2522 | Ersen Ekrem | Ersen Ekrem and Sennur Ulukus | Degrees of Freedom Region of the Gaussian MIMO Broadcast Channel with
Common and Private Messages | Submitted to IEEE Transactions on Information Theory, May 2011 | null | null | null | cs.IT math.IT | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We consider the Gaussian multiple-input multiple-output (MIMO) broadcast
channel with common and private messages. We obtain the degrees of freedom
(DoF) region of this channel. We first show that a parallel Gaussian broadcast
channel with unmatched sub-channels can be constructed from any given Gaussian
MIMO broadcast channel by using the generalized singular value decomposition
(GSVD) and a relaxation on the power constraint for the channel input, in a way
that the capacity region of the constructed parallel channel provides an outer
bound for the capacity region of the original channel. The capacity region of
the parallel Gaussian broadcast channel with unmatched sub-channels is known,
using which we obtain an explicit outer bound for the DoF region of the
Gaussian MIMO broadcast channel. We finally show that this outer bound for the
DoF region can be attained both by the achievable scheme that uses a classical
Gaussian coding for the common message and dirty-paper coding (DPC) for the
private messages, as well as by a variation of the zero-forcing (ZF) scheme.
| [
{
"created": "Mon, 13 Jun 2011 19:06:43 GMT",
"version": "v1"
}
] | 2011-06-14 | [
[
"Ekrem",
"Ersen",
""
],
[
"Ulukus",
"Sennur",
""
]
] | We consider the Gaussian multiple-input multiple-output (MIMO) broadcast channel with common and private messages. We obtain the degrees of freedom (DoF) region of this channel. We first show that a parallel Gaussian broadcast channel with unmatched sub-channels can be constructed from any given Gaussian MIMO broadcast channel by using the generalized singular value decomposition (GSVD) and a relaxation on the power constraint for the channel input, in a way that the capacity region of the constructed parallel channel provides an outer bound for the capacity region of the original channel. The capacity region of the parallel Gaussian broadcast channel with unmatched sub-channels is known, using which we obtain an explicit outer bound for the DoF region of the Gaussian MIMO broadcast channel. We finally show that this outer bound for the DoF region can be attained both by the achievable scheme that uses a classical Gaussian coding for the common message and dirty-paper coding (DPC) for the private messages, as well as by a variation of the zero-forcing (ZF) scheme. |
1808.04107 | Sajad Daei Omshi | Sajad Daei, Farzan Haddadi, Arash Amini | Improved Recovery of Analysis Sparse Vectors in Presence of Prior
Information | null | null | 10.1109/LSP.2018.2886141 | null | cs.IT math.IT | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In this work, we consider the problem of recovering analysis-sparse signals
from under-sampled measurements when some prior information about the support
is available. We incorporate such information in the recovery stage by suitably
tuning the weights in a weighted $\ell_1$ analysis optimization problem.
Indeed, we try to set the weights such that the method succeeds with minimum
number of measurements. For this purpose, we exploit the upper-bound on the
statistical dimension of a certain cone to determine the weights. Our numerical
simulations confirm that the introduced method with tuned weights outperforms
the standard $\ell_1$ analysis technique.
| [
{
"created": "Mon, 13 Aug 2018 08:42:45 GMT",
"version": "v1"
}
] | 2019-01-30 | [
[
"Daei",
"Sajad",
""
],
[
"Haddadi",
"Farzan",
""
],
[
"Amini",
"Arash",
""
]
] | In this work, we consider the problem of recovering analysis-sparse signals from under-sampled measurements when some prior information about the support is available. We incorporate such information in the recovery stage by suitably tuning the weights in a weighted $\ell_1$ analysis optimization problem. Indeed, we try to set the weights such that the method succeeds with minimum number of measurements. For this purpose, we exploit the upper-bound on the statistical dimension of a certain cone to determine the weights. Our numerical simulations confirm that the introduced method with tuned weights outperforms the standard $\ell_1$ analysis technique. |
1708.07188 | Sanjay Goel | Sanjay Goel, Stephen F. Bush, Carlos Gershenson | Self-Organization in Traffic Lights: Evolution of Signal Control with
Advances in Sensors and Communications | null | null | null | null | cs.SY | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Traffic signals are ubiquitous devices that first appeared in 1868. Recent
advances in information and communications technology (ICT) have led to
unprecedented improvements in such areas as mobile handheld devices (i.e.,
smartphones), the electric power industry (i.e., smart grids), transportation
infrastructure, and vehicle area networks. Given the trend towards
interconnectivity, it is only a matter of time before vehicles communicate with
one another and with infrastructure. In fact, several pilots of such
vehicle-to-vehicle and vehicle-to-infrastructure (e.g. traffic lights and
parking spaces) communication systems are already operational. This survey of
autonomous and self-organized traffic signaling control has been undertaken
with these potential developments in mind. Our research results indicate that,
while many sophisticated techniques have attempted to improve the scheduling of
traffic signal control, either real-time sensing of traffic patterns or a
priori knowledge of traffic flow is required to optimize traffic. Once this is
achieved, communication between traffic signals will serve to vastly improve
overall traffic efficiency.
| [
{
"created": "Sun, 18 Jun 2017 21:41:24 GMT",
"version": "v1"
}
] | 2017-08-25 | [
[
"Goel",
"Sanjay",
""
],
[
"Bush",
"Stephen F.",
""
],
[
"Gershenson",
"Carlos",
""
]
] | Traffic signals are ubiquitous devices that first appeared in 1868. Recent advances in information and communications technology (ICT) have led to unprecedented improvements in such areas as mobile handheld devices (i.e., smartphones), the electric power industry (i.e., smart grids), transportation infrastructure, and vehicle area networks. Given the trend towards interconnectivity, it is only a matter of time before vehicles communicate with one another and with infrastructure. In fact, several pilots of such vehicle-to-vehicle and vehicle-to-infrastructure (e.g. traffic lights and parking spaces) communication systems are already operational. This survey of autonomous and self-organized traffic signaling control has been undertaken with these potential developments in mind. Our research results indicate that, while many sophisticated techniques have attempted to improve the scheduling of traffic signal control, either real-time sensing of traffic patterns or a priori knowledge of traffic flow is required to optimize traffic. Once this is achieved, communication between traffic signals will serve to vastly improve overall traffic efficiency. |
2102.05692 | Mollie Bianchi | Mollie Bianchi and Timothy D. Barfoot | UAV Localization Using Autoencoded Satellite Images | Accepted for publication in RA-L 2021 | null | null | null | cs.CV cs.RO | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We propose and demonstrate a fast, robust method for using satellite images
to localize an Unmanned Aerial Vehicle (UAV). Previous work using satellite
images has large storage and computation costs and is unable to run in real
time. In this work, we collect Google Earth (GE) images for a desired flight
path offline and an autoencoder is trained to compress these images to a
low-dimensional vector representation while retaining the key features. This
trained autoencoder is used to compress a real UAV image, which is then
compared to the precollected, nearby, autoencoded GE images using an
inner-product kernel. This results in a distribution of weights over the
corresponding GE image poses and is used to generate a single localization and
associated covariance to represent uncertainty. Our localization is computed in
1% of the time of the current standard and is able to achieve a comparable RMSE
of less than 3m in our experiments, where we robustly matched UAV images from
six runs spanning the lighting conditions of a single day to the same map of
satellite images.
| [
{
"created": "Wed, 10 Feb 2021 19:08:10 GMT",
"version": "v1"
}
] | 2021-02-12 | [
[
"Bianchi",
"Mollie",
""
],
[
"Barfoot",
"Timothy D.",
""
]
] | We propose and demonstrate a fast, robust method for using satellite images to localize an Unmanned Aerial Vehicle (UAV). Previous work using satellite images has large storage and computation costs and is unable to run in real time. In this work, we collect Google Earth (GE) images for a desired flight path offline and an autoencoder is trained to compress these images to a low-dimensional vector representation while retaining the key features. This trained autoencoder is used to compress a real UAV image, which is then compared to the precollected, nearby, autoencoded GE images using an inner-product kernel. This results in a distribution of weights over the corresponding GE image poses and is used to generate a single localization and associated covariance to represent uncertainty. Our localization is computed in 1% of the time of the current standard and is able to achieve a comparable RMSE of less than 3m in our experiments, where we robustly matched UAV images from six runs spanning the lighting conditions of a single day to the same map of satellite images. |
2007.06604 | Amihood Amir | Amihood Amir and Itai Boneh | Update Query Time Trade-off for dynamic Suffix Arrays | 19 pages, 3 figures | null | null | null | cs.DS | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The Suffix Array SA(S) of a string S[1 ... n] is an array containing all the
suffixes of S sorted by lexicographic order. The suffix array is one of the
most well known indexing data structures, and it functions as a key tool in
many string algorithms. In this paper, we present a data structure for
maintaining the Suffix Array of a dynamic string. For every $0 \leq \varepsilon
\leq 1$, our data structure reports SA[i] in $\tilde{O}(n^{\varepsilon})$ time
and handles text modification in $\tilde{O}(n^{1-\varepsilon})$ time.
Additionally, our data structure enables the same query time for reporting
iSA[i], with iSA being the Inverse Suffix Array of S[1 ... n]. Our data
structure can be used to construct sub-linear dynamic variants of static
strings algorithms or data structures that are based on the Suffix Array and
the Inverse Suffix Array.
| [
{
"created": "Mon, 13 Jul 2020 18:11:19 GMT",
"version": "v1"
}
] | 2020-07-15 | [
[
"Amir",
"Amihood",
""
],
[
"Boneh",
"Itai",
""
]
] | The Suffix Array SA(S) of a string S[1 ... n] is an array containing all the suffixes of S sorted by lexicographic order. The suffix array is one of the most well known indexing data structures, and it functions as a key tool in many string algorithms. In this paper, we present a data structure for maintaining the Suffix Array of a dynamic string. For every $0 \leq \varepsilon \leq 1$, our data structure reports SA[i] in $\tilde{O}(n^{\varepsilon})$ time and handles text modification in $\tilde{O}(n^{1-\varepsilon})$ time. Additionally, our data structure enables the same query time for reporting iSA[i], with iSA being the Inverse Suffix Array of S[1 ... n]. Our data structure can be used to construct sub-linear dynamic variants of static strings algorithms or data structures that are based on the Suffix Array and the Inverse Suffix Array. |
2108.12630 | Shuaicheng Li | Shuaicheng Li, Qianggang Cao, Lingbo Liu, Kunlin Yang, Shinan Liu, Jun
Hou and Shuai Yi | GroupFormer: Group Activity Recognition with Clustered Spatial-Temporal
Transformer | Accepted at ICCV2021 | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Group activity recognition is a crucial yet challenging problem, whose core
lies in fully exploring spatial-temporal interactions among individuals and
generating reasonable group representations. However, previous methods either
model spatial and temporal information separately, or directly aggregate
individual features to form group features. To address these issues, we propose
a novel group activity recognition network termed GroupFormer. It captures
spatial-temporal contextual information jointly to augment the individual and
group representations effectively with a clustered spatial-temporal
transformer. Specifically, our GroupFormer has three appealing advantages: (1)
A tailor-modified Transformer, Clustered Spatial-Temporal Transformer, is
proposed to enhance the individual representation and group representation. (2)
It models the spatial and temporal dependencies integrally and utilizes
decoders to build the bridge between the spatial and temporal information. (3)
A clustered attention mechanism is utilized to dynamically divide individuals
into multiple clusters for better learning activity-aware semantic
representations. Moreover, experimental results show that the proposed
framework outperforms state-of-the-art methods on the Volleyball dataset and
Collective Activity dataset. Code is available at
https://github.com/xueyee/GroupFormer.
| [
{
"created": "Sat, 28 Aug 2021 11:24:36 GMT",
"version": "v1"
}
] | 2021-08-31 | [
[
"Li",
"Shuaicheng",
""
],
[
"Cao",
"Qianggang",
""
],
[
"Liu",
"Lingbo",
""
],
[
"Yang",
"Kunlin",
""
],
[
"Liu",
"Shinan",
""
],
[
"Hou",
"Jun",
""
],
[
"Yi",
"Shuai",
""
]
] | Group activity recognition is a crucial yet challenging problem, whose core lies in fully exploring spatial-temporal interactions among individuals and generating reasonable group representations. However, previous methods either model spatial and temporal information separately, or directly aggregate individual features to form group features. To address these issues, we propose a novel group activity recognition network termed GroupFormer. It captures spatial-temporal contextual information jointly to augment the individual and group representations effectively with a clustered spatial-temporal transformer. Specifically, our GroupFormer has three appealing advantages: (1) A tailor-modified Transformer, Clustered Spatial-Temporal Transformer, is proposed to enhance the individual representation and group representation. (2) It models the spatial and temporal dependencies integrally and utilizes decoders to build the bridge between the spatial and temporal information. (3) A clustered attention mechanism is utilized to dynamically divide individuals into multiple clusters for better learning activity-aware semantic representations. Moreover, experimental results show that the proposed framework outperforms state-of-the-art methods on the Volleyball dataset and Collective Activity dataset. Code is available at https://github.com/xueyee/GroupFormer. |
2112.01079 | Shudong Yang | Shudong Yang (1) ((1) Dalian University of Technology) | Who will dropout from university? Academic risk prediction based on
interpretable machine learning | 15 pages,7 figures | null | null | null | cs.LG stat.ML | http://creativecommons.org/licenses/by-nc-nd/4.0/ | In the institutional research mode, in order to explore which characteristics
are the best indicators for predicting academic risk from the student behavior
data sets that have high-dimensional, unbalanced classified small sample, it
transforms the academic risk prediction of college students into a binary
classification task. It predicts academic risk based on the LightGBM model and
the interpretable machine learning method of Shapley value. The simulation
results show that from the global perspective of the prediction model,
characteristics such as the quality of academic partners, the seating position
in classroom, the dormitory study atmosphere, the English scores of the college
entrance examination, the quantity of academic partners, the addiction level of
video games, the mobility of academic partners, and the degree of truancy are
the best 8 predictors for academic risk. It is contrary to intuition that
characteristics such as living in campus or not, work-study, lipstick
addiction, student leader or not, lover amount, and smoking have little
correlation with university academic risk in this experiment. From the local
perspective of the sample, the factors affecting academic risk vary from person
to person. It can perform personalized interpretable analysis through Shapley
values, which cannot be done by traditional mathematical statistical prediction
models. The academic contributions of this research are mainly in two aspects:
First, the learning interaction networks is proposed for the first time, so
that social behavior can be used to compensate for the one-sided individual
behavior and improve the performance of academic risk prediction. Second, the
introduction of Shapley value calculation makes machine learning that lacks a
clear reasoning process visualized, and provides intuitive decision support for
education managers.
| [
{
"created": "Thu, 2 Dec 2021 09:43:31 GMT",
"version": "v1"
}
] | 2021-12-03 | [
[
"Yang",
"Shudong",
"",
"Dalian University of Technology"
]
] | In the institutional research mode, in order to explore which characteristics are the best indicators for predicting academic risk from the student behavior data sets that have high-dimensional, unbalanced classified small sample, it transforms the academic risk prediction of college students into a binary classification task. It predicts academic risk based on the LightGBM model and the interpretable machine learning method of Shapley value. The simulation results show that from the global perspective of the prediction model, characteristics such as the quality of academic partners, the seating position in classroom, the dormitory study atmosphere, the English scores of the college entrance examination, the quantity of academic partners, the addiction level of video games, the mobility of academic partners, and the degree of truancy are the best 8 predictors for academic risk. It is contrary to intuition that characteristics such as living in campus or not, work-study, lipstick addiction, student leader or not, lover amount, and smoking have little correlation with university academic risk in this experiment. From the local perspective of the sample, the factors affecting academic risk vary from person to person. It can perform personalized interpretable analysis through Shapley values, which cannot be done by traditional mathematical statistical prediction models. The academic contributions of this research are mainly in two aspects: First, the learning interaction networks is proposed for the first time, so that social behavior can be used to compensate for the one-sided individual behavior and improve the performance of academic risk prediction. Second, the introduction of Shapley value calculation makes machine learning that lacks a clear reasoning process visualized, and provides intuitive decision support for education managers. |
2009.08044 | Mark Hamilton | Mark Hamilton, Nick Gonsalves, Christina Lee, Anand Raman, Brendan
Walsh, Siddhartha Prasad, Dalitso Banda, Lucy Zhang, Mei Gao, Lei Zhang,
William T. Freeman | Large-Scale Intelligent Microservices | null | null | 10.1109/BigData50022.2020.9378270 | null | cs.AI cs.DB cs.DC cs.LG cs.NI | http://creativecommons.org/licenses/by/4.0/ | Deploying Machine Learning (ML) algorithms within databases is a challenge
due to the varied computational footprints of modern ML algorithms and the
myriad of database technologies each with its own restrictive syntax. We
introduce an Apache Spark-based micro-service orchestration framework that
extends database operations to include web service primitives. Our system can
orchestrate web services across hundreds of machines and takes full advantage
of cluster, thread, and asynchronous parallelism. Using this framework, we
provide large scale clients for intelligent services such as speech, vision,
search, anomaly detection, and text analysis. This allows users to integrate
ready-to-use intelligence into any datastore with an Apache Spark connector. To
eliminate the majority of overhead from network communication, we also
introduce a low-latency containerized version of our architecture. Finally, we
demonstrate that the services we investigate are competitive on a variety of
benchmarks, and present two applications of this framework to create
intelligent search engines, and real-time auto race analytics systems.
| [
{
"created": "Thu, 17 Sep 2020 03:38:28 GMT",
"version": "v1"
},
{
"created": "Thu, 3 Dec 2020 20:51:47 GMT",
"version": "v2"
},
{
"created": "Thu, 2 Dec 2021 20:09:30 GMT",
"version": "v3"
}
] | 2022-03-17 | [
[
"Hamilton",
"Mark",
""
],
[
"Gonsalves",
"Nick",
""
],
[
"Lee",
"Christina",
""
],
[
"Raman",
"Anand",
""
],
[
"Walsh",
"Brendan",
""
],
[
"Prasad",
"Siddhartha",
""
],
[
"Banda",
"Dalitso",
""
],
[
"Zhang",
"Lucy",
""
],
[
"Gao",
"Mei",
""
],
[
"Zhang",
"Lei",
""
],
[
"Freeman",
"William T.",
""
]
] | Deploying Machine Learning (ML) algorithms within databases is a challenge due to the varied computational footprints of modern ML algorithms and the myriad of database technologies each with its own restrictive syntax. We introduce an Apache Spark-based micro-service orchestration framework that extends database operations to include web service primitives. Our system can orchestrate web services across hundreds of machines and takes full advantage of cluster, thread, and asynchronous parallelism. Using this framework, we provide large scale clients for intelligent services such as speech, vision, search, anomaly detection, and text analysis. This allows users to integrate ready-to-use intelligence into any datastore with an Apache Spark connector. To eliminate the majority of overhead from network communication, we also introduce a low-latency containerized version of our architecture. Finally, we demonstrate that the services we investigate are competitive on a variety of benchmarks, and present two applications of this framework to create intelligent search engines, and real-time auto race analytics systems. |
2306.09385 | Rahee Walambe Dr | Rahee Walambe, Pranav Nayak, Ashmit Bhardwaj, Ketan Kotecha | Employing Multimodal Machine Learning for Stress Detection | null | null | 10.1155/2021/9356452 | null | cs.LG cs.AI eess.SP | http://creativecommons.org/licenses/by/4.0/ | In the current age, human lifestyle has become more knowledge oriented
leading to generation of sedentary employment. This has given rise to a number
of health and mental disorders. Mental wellness is one of the most neglected
but crucial aspects of today's world. Mental health issues can, both directly
and indirectly, affect other sections of human physiology and impede an
individual's day-to-day activities and performance. However, identifying the
stress and finding the stress trend for an individual leading to serious mental
ailments is challenging and involves multiple factors. Such identification can
be achieved accurately by fusing these multiple modalities (due to various
factors) arising from behavioral patterns. Certain techniques are identified in
the literature for this purpose; however, very few machine learning-based
methods are proposed for such multimodal fusion tasks. In this work, a
multimodal AI-based framework is proposed to monitor a person's working
behavior and stress levels. We propose a methodology for efficiently detecting
stress due to workload by concatenating heterogeneous raw sensor data streams
(e.g., face expressions, posture, heart rate, computer interaction). This data
can be securely stored and analyzed to understand and discover personalized
unique behavioral patterns leading to mental strain and fatigue. The
contribution of this work is twofold; proposing a multimodal AI-based strategy
for fusion to detect stress and its level and secondly identify a stress
pattern over a period of time. We were able to achieve 96.09% accuracy on the
test set in stress detection and classification. Further, we reduce the stress
scale prediction model loss to 0.036 using these modalities. This work can
prove important for the community at large, specifically those working
sedentary jobs to monitor and identify stress levels, especially in current
times of COVID-19.
| [
{
"created": "Thu, 15 Jun 2023 14:34:16 GMT",
"version": "v1"
}
] | 2023-06-19 | [
[
"Walambe",
"Rahee",
""
],
[
"Nayak",
"Pranav",
""
],
[
"Bhardwaj",
"Ashmit",
""
],
[
"Kotecha",
"Ketan",
""
]
] | In the current age, human lifestyle has become more knowledge oriented leading to generation of sedentary employment. This has given rise to a number of health and mental disorders. Mental wellness is one of the most neglected but crucial aspects of today's world. Mental health issues can, both directly and indirectly, affect other sections of human physiology and impede an individual's day-to-day activities and performance. However, identifying the stress and finding the stress trend for an individual leading to serious mental ailments is challenging and involves multiple factors. Such identification can be achieved accurately by fusing these multiple modalities (due to various factors) arising from behavioral patterns. Certain techniques are identified in the literature for this purpose; however, very few machine learning-based methods are proposed for such multimodal fusion tasks. In this work, a multimodal AI-based framework is proposed to monitor a person's working behavior and stress levels. We propose a methodology for efficiently detecting stress due to workload by concatenating heterogeneous raw sensor data streams (e.g., face expressions, posture, heart rate, computer interaction). This data can be securely stored and analyzed to understand and discover personalized unique behavioral patterns leading to mental strain and fatigue. The contribution of this work is twofold; proposing a multimodal AI-based strategy for fusion to detect stress and its level and secondly identify a stress pattern over a period of time. We were able to achieve 96.09% accuracy on the test set in stress detection and classification. Further, we reduce the stress scale prediction model loss to 0.036 using these modalities. This work can prove important for the community at large, specifically those working sedentary jobs to monitor and identify stress levels, especially in current times of COVID-19. |
1906.02134 | Jie Wang | Jie Wang, Xinyan Zhao | Theme-aware generation model for chinese lyrics | null | null | null | null | cs.CL cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | With rapid development of neural networks, deep-learning has been extended to
various natural language generation fields, such as machine translation,
dialogue generation and even literature creation. In this paper, we propose a
theme-aware language generation model for Chinese music lyrics, which improves
the theme-connectivity and coherence of generated paragraphs greatly. A
multi-channel sequence-to-sequence (seq2seq) model encodes themes and previous
sentences as global and local contextual information. Moreover, attention
mechanism is incorporated for sequence decoding, enabling to fuse context into
predicted next texts. To prepare appropriate train corpus, LDA (Latent
Dirichlet Allocation) is applied for theme extraction. Generated lyrics is
grammatically correct and semantically coherent with selected themes, which
offers a valuable modelling method in other fields including multi-turn
chatbots, long paragraph generation and etc.
| [
{
"created": "Thu, 23 May 2019 08:50:15 GMT",
"version": "v1"
}
] | 2019-06-06 | [
[
"Wang",
"Jie",
""
],
[
"Zhao",
"Xinyan",
""
]
] | With rapid development of neural networks, deep-learning has been extended to various natural language generation fields, such as machine translation, dialogue generation and even literature creation. In this paper, we propose a theme-aware language generation model for Chinese music lyrics, which improves the theme-connectivity and coherence of generated paragraphs greatly. A multi-channel sequence-to-sequence (seq2seq) model encodes themes and previous sentences as global and local contextual information. Moreover, attention mechanism is incorporated for sequence decoding, enabling to fuse context into predicted next texts. To prepare appropriate train corpus, LDA (Latent Dirichlet Allocation) is applied for theme extraction. Generated lyrics is grammatically correct and semantically coherent with selected themes, which offers a valuable modelling method in other fields including multi-turn chatbots, long paragraph generation and etc. |
2301.03221 | Arnaud de Mesmay | Eun Jung Kim, Arnaud de Mesmay, Tillmann Miltzow | Representing Matroids over the Reals is $\exists \mathbb R$-complete | v2 and v3: Minor changes v4: Final version, to appear in DMTCS | null | null | null | cs.CC math.CO | http://creativecommons.org/licenses/by/4.0/ | A matroid $M$ is an ordered pair $(E,I)$, where $E$ is a finite set called
the ground set and a collection $I\subset 2^{E}$ called the independent sets
which satisfy the conditions: (i) $\emptyset \in I$, (ii) $I'\subset I \in I$
implies $I'\in I$, and (iii) $I_1,I_2 \in I$ and $|I_1| < |I_2|$ implies that
there is an $e\in I_2$ such that $I_1\cup \{e\} \in I$. The rank $rank(M)$ of a
matroid $M$ is the maximum size of an independent set. We say that a matroid
$M=(E,I)$ is representable over the reals if there is a map $\varphi \colon E
\rightarrow \mathbb{R}^{rank(M)}$ such that $I\in I$ if and only if
$\varphi(I)$ forms a linearly independent set.
We study the problem of matroid realizability over the reals. Given a matroid
$M$, we ask whether there is a set of points in the Euclidean space
representing $M$. We show that matroid realizability is $\exists \mathbb
R$-complete, already for matroids of rank 3. The complexity class $\exists
\mathbb R$ can be defined as the family of algorithmic problems that is
polynomial-time is equivalent to determining if a multivariate polynomial with
integers coefficients has a real root.
Our methods are similar to previous methods from the literature. Yet, the
result itself was never pointed out and there is no proof readily available in
the language of computer science.
| [
{
"created": "Mon, 9 Jan 2023 09:33:50 GMT",
"version": "v1"
},
{
"created": "Mon, 8 Jan 2024 16:49:45 GMT",
"version": "v2"
},
{
"created": "Tue, 9 Jan 2024 11:13:41 GMT",
"version": "v3"
},
{
"created": "Thu, 11 Jul 2024 14:35:55 GMT",
"version": "v4"
}
] | 2024-07-12 | [
[
"Kim",
"Eun Jung",
""
],
[
"de Mesmay",
"Arnaud",
""
],
[
"Miltzow",
"Tillmann",
""
]
] | A matroid $M$ is an ordered pair $(E,I)$, where $E$ is a finite set called the ground set and a collection $I\subset 2^{E}$ called the independent sets which satisfy the conditions: (i) $\emptyset \in I$, (ii) $I'\subset I \in I$ implies $I'\in I$, and (iii) $I_1,I_2 \in I$ and $|I_1| < |I_2|$ implies that there is an $e\in I_2$ such that $I_1\cup \{e\} \in I$. The rank $rank(M)$ of a matroid $M$ is the maximum size of an independent set. We say that a matroid $M=(E,I)$ is representable over the reals if there is a map $\varphi \colon E \rightarrow \mathbb{R}^{rank(M)}$ such that $I\in I$ if and only if $\varphi(I)$ forms a linearly independent set. We study the problem of matroid realizability over the reals. Given a matroid $M$, we ask whether there is a set of points in the Euclidean space representing $M$. We show that matroid realizability is $\exists \mathbb R$-complete, already for matroids of rank 3. The complexity class $\exists \mathbb R$ can be defined as the family of algorithmic problems that is polynomial-time is equivalent to determining if a multivariate polynomial with integers coefficients has a real root. Our methods are similar to previous methods from the literature. Yet, the result itself was never pointed out and there is no proof readily available in the language of computer science. |
1212.2036 | Jiwei Li | Jiwei Li and Sujian Li | Query-focused Multi-document Summarization: Combining a Novel Topic
Model with Graph-based Semi-supervised Learning | This paper has been withdrawn by the author due to a crucial sign
error in equation | null | null | null | cs.CL cs.IR | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Graph-based semi-supervised learning has proven to be an effective approach
for query-focused multi-document summarization. The problem of previous
semi-supervised learning is that sentences are ranked without considering the
higher level information beyond sentence level. Researches on general
summarization illustrated that the addition of topic level can effectively
improve the summary quality. Inspired by previous researches, we propose a
two-layer (i.e. sentence layer and topic layer) graph-based semi-supervised
learning approach. At the same time, we propose a novel topic model which makes
full use of the dependence between sentences and words. Experimental results on
DUC and TAC data sets demonstrate the effectiveness of our proposed approach.
| [
{
"created": "Mon, 10 Dec 2012 11:35:29 GMT",
"version": "v1"
},
{
"created": "Fri, 27 Dec 2013 17:24:00 GMT",
"version": "v2"
},
{
"created": "Tue, 31 Dec 2013 17:13:33 GMT",
"version": "v3"
}
] | 2014-01-03 | [
[
"Li",
"Jiwei",
""
],
[
"Li",
"Sujian",
""
]
] | Graph-based semi-supervised learning has proven to be an effective approach for query-focused multi-document summarization. The problem of previous semi-supervised learning is that sentences are ranked without considering the higher level information beyond sentence level. Researches on general summarization illustrated that the addition of topic level can effectively improve the summary quality. Inspired by previous researches, we propose a two-layer (i.e. sentence layer and topic layer) graph-based semi-supervised learning approach. At the same time, we propose a novel topic model which makes full use of the dependence between sentences and words. Experimental results on DUC and TAC data sets demonstrate the effectiveness of our proposed approach. |
2207.09774 | Edoardo Remelli | Edoardo Remelli, Timur Bagautdinov, Shunsuke Saito, Tomas Simon,
Chenglei Wu, Shih-En Wei, Kaiwen Guo, Zhe Cao, Fabian Prada, Jason Saragih,
Yaser Sheikh | Drivable Volumetric Avatars using Texel-Aligned Features | null | SIGGRAPH 2022 Conference Proceedings | null | null | cs.CV | http://creativecommons.org/licenses/by/4.0/ | Photorealistic telepresence requires both high-fidelity body modeling and
faithful driving to enable dynamically synthesized appearance that is
indistinguishable from reality. In this work, we propose an end-to-end
framework that addresses two core challenges in modeling and driving full-body
avatars of real people. One challenge is driving an avatar while staying
faithful to details and dynamics that cannot be captured by a global
low-dimensional parameterization such as body pose. Our approach supports
driving of clothed avatars with wrinkles and motion that a real driving
performer exhibits beyond the training corpus. Unlike existing global state
representations or non-parametric screen-space approaches, we introduce
texel-aligned features -- a localised representation which can leverage both
the structural prior of a skeleton-based parametric model and observed sparse
image signals at the same time. Another challenge is modeling a temporally
coherent clothed avatar, which typically requires precise surface tracking. To
circumvent this, we propose a novel volumetric avatar representation by
extending mixtures of volumetric primitives to articulated objects. By
explicitly incorporating articulation, our approach naturally generalizes to
unseen poses. We also introduce a localized viewpoint conditioning, which leads
to a large improvement in generalization of view-dependent appearance. The
proposed volumetric representation does not require high-quality mesh tracking
as a prerequisite and brings significant quality improvements compared to
mesh-based counterparts. In our experiments, we carefully examine our design
choices and demonstrate the efficacy of our approach, outperforming the
state-of-the-art methods on challenging driving scenarios.
| [
{
"created": "Wed, 20 Jul 2022 09:28:16 GMT",
"version": "v1"
}
] | 2022-07-21 | [
[
"Remelli",
"Edoardo",
""
],
[
"Bagautdinov",
"Timur",
""
],
[
"Saito",
"Shunsuke",
""
],
[
"Simon",
"Tomas",
""
],
[
"Wu",
"Chenglei",
""
],
[
"Wei",
"Shih-En",
""
],
[
"Guo",
"Kaiwen",
""
],
[
"Cao",
"Zhe",
""
],
[
"Prada",
"Fabian",
""
],
[
"Saragih",
"Jason",
""
],
[
"Sheikh",
"Yaser",
""
]
] | Photorealistic telepresence requires both high-fidelity body modeling and faithful driving to enable dynamically synthesized appearance that is indistinguishable from reality. In this work, we propose an end-to-end framework that addresses two core challenges in modeling and driving full-body avatars of real people. One challenge is driving an avatar while staying faithful to details and dynamics that cannot be captured by a global low-dimensional parameterization such as body pose. Our approach supports driving of clothed avatars with wrinkles and motion that a real driving performer exhibits beyond the training corpus. Unlike existing global state representations or non-parametric screen-space approaches, we introduce texel-aligned features -- a localised representation which can leverage both the structural prior of a skeleton-based parametric model and observed sparse image signals at the same time. Another challenge is modeling a temporally coherent clothed avatar, which typically requires precise surface tracking. To circumvent this, we propose a novel volumetric avatar representation by extending mixtures of volumetric primitives to articulated objects. By explicitly incorporating articulation, our approach naturally generalizes to unseen poses. We also introduce a localized viewpoint conditioning, which leads to a large improvement in generalization of view-dependent appearance. The proposed volumetric representation does not require high-quality mesh tracking as a prerequisite and brings significant quality improvements compared to mesh-based counterparts. In our experiments, we carefully examine our design choices and demonstrate the efficacy of our approach, outperforming the state-of-the-art methods on challenging driving scenarios. |
1809.06965 | Seung Bin Baik | Seung Bin Baik, Keum Gang Cha | A Study on Deep Learning Based Sauvegrain Method for Measurement of
Puberty Bone Age | 5 pages, 6 figures, 1 table | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | This study applies a technique to expand the number of images to a level that
allows deep learning. And the applicability of the Sauvegrain method through
deep learning with relatively few elbow X-rays is studied. The study was
composed of processes similar to the physicians' bone age assessment
procedures. The selected reference images were learned without being included
in the evaluation data, and at the same time, the data was extended to
accommodate the number of cases. In addition, we adjusted the X-ray images to
better images using U-Net and selected the ROI with RPN + so as to be able to
perform bone age estimation through CNN. The mean absolute error of the
Sauvegrain method based on deep learning is 2.8 months and the Mean Absolute
Percentage Error (MAPE) is 0.018. This result shows that X - ray analysis using
the Sauvegrain method shows higher accuracy than that of the age group of
puberty even in the deep learning base. This means that deep learning of the
Suvegrain method can be measured at a level similar to that of an expert, based
on the extended X-ray image with the image data extension technique. Finally,
we applied the Sauvegrain method to deep learning for accurate measurement of
bone age at puberty. As a result, the present study is based on deep learning,
and compared with the evaluation results of experts, it is possible to overcome
limitations of the method of measuring bone age based on machine learning which
was in TW3 or Greulich & Pyle due to lack of X- I confirmed the fact. And we
also presented the Sauvegrain method, which is applicable to adolescents as
well.
| [
{
"created": "Tue, 18 Sep 2018 23:47:08 GMT",
"version": "v1"
}
] | 2018-09-20 | [
[
"Baik",
"Seung Bin",
""
],
[
"Cha",
"Keum Gang",
""
]
] | This study applies a technique to expand the number of images to a level that allows deep learning. And the applicability of the Sauvegrain method through deep learning with relatively few elbow X-rays is studied. The study was composed of processes similar to the physicians' bone age assessment procedures. The selected reference images were learned without being included in the evaluation data, and at the same time, the data was extended to accommodate the number of cases. In addition, we adjusted the X-ray images to better images using U-Net and selected the ROI with RPN + so as to be able to perform bone age estimation through CNN. The mean absolute error of the Sauvegrain method based on deep learning is 2.8 months and the Mean Absolute Percentage Error (MAPE) is 0.018. This result shows that X - ray analysis using the Sauvegrain method shows higher accuracy than that of the age group of puberty even in the deep learning base. This means that deep learning of the Suvegrain method can be measured at a level similar to that of an expert, based on the extended X-ray image with the image data extension technique. Finally, we applied the Sauvegrain method to deep learning for accurate measurement of bone age at puberty. As a result, the present study is based on deep learning, and compared with the evaluation results of experts, it is possible to overcome limitations of the method of measuring bone age based on machine learning which was in TW3 or Greulich & Pyle due to lack of X- I confirmed the fact. And we also presented the Sauvegrain method, which is applicable to adolescents as well. |
2404.02608 | Jeferson Gonzalez-Gomez | Jeferson Gonzalez-Gomez, Hassan Nassar, Lars Bauer and Jorg Henkel | LightFAt: Mitigating Control-flow Explosion via Lightweight PMU-based
Control-flow Attestation | This official version of this paper will appear in the 2024 IEEE
International Symposium on Hardware Oriented Security and Trust (HOST) | null | null | null | cs.CR | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | With the continuous evolution of computational devices, more and more
applications are being executed remotely. The applications operate on a wide
spectrum of devices, ranging from IoT nodes with low computational capabilities
to large cloud providers with high capabilities. Remote execution often deals
with sensitive data or executes proprietary software. Hence, the challenge of
ensuring that the code execution will not be compromised rises. Remote
Attestation deals with this challenge. It ensures the code is executed in a
non-compromised environment by calculating a potentially large sequence of
cryptographic hash values. Each hash calculation is computationally intensive
and over a large sequence the overhead becomes extremely high. In this work, we
propose LightFAt: a Lightweight Control Flow Attestation scheme. Instead of
relying on the expensive cryptographic hash calculation, LightFAt leverages the
readings from the processor's Performance Monitor Unit (PMU) in conjunction
with a lightweight unsupervised machine learning (ML) classifier to detect
whether a target application's control flow is compromised, hence improving the
system's security. On the verifier's side, LightFAt reaches a detection
accuracy of over 95%, with low false-negative and false-positive rates.
| [
{
"created": "Wed, 3 Apr 2024 09:55:15 GMT",
"version": "v1"
},
{
"created": "Thu, 4 Apr 2024 09:20:33 GMT",
"version": "v2"
}
] | 2024-04-05 | [
[
"Gonzalez-Gomez",
"Jeferson",
""
],
[
"Nassar",
"Hassan",
""
],
[
"Bauer",
"Lars",
""
],
[
"Henkel",
"Jorg",
""
]
] | With the continuous evolution of computational devices, more and more applications are being executed remotely. The applications operate on a wide spectrum of devices, ranging from IoT nodes with low computational capabilities to large cloud providers with high capabilities. Remote execution often deals with sensitive data or executes proprietary software. Hence, the challenge of ensuring that the code execution will not be compromised rises. Remote Attestation deals with this challenge. It ensures the code is executed in a non-compromised environment by calculating a potentially large sequence of cryptographic hash values. Each hash calculation is computationally intensive and over a large sequence the overhead becomes extremely high. In this work, we propose LightFAt: a Lightweight Control Flow Attestation scheme. Instead of relying on the expensive cryptographic hash calculation, LightFAt leverages the readings from the processor's Performance Monitor Unit (PMU) in conjunction with a lightweight unsupervised machine learning (ML) classifier to detect whether a target application's control flow is compromised, hence improving the system's security. On the verifier's side, LightFAt reaches a detection accuracy of over 95%, with low false-negative and false-positive rates. |
2103.02654 | Yudi Dong | Yudi Dong and Huaxia Wang and Yu-Dong Yao | A Robust Adversarial Network-Based End-to-End Communications System With
Strong Generalization Ability Against Adversarial Attacks | 5 pages letter | ICC 2022 - IEEE International Conference on Communications | 10.1109/ICC45855.2022.9838452 | null | cs.LG cs.AI eess.SP | http://creativecommons.org/licenses/by/4.0/ | We propose a novel defensive mechanism based on a generative adversarial
network (GAN) framework to defend against adversarial attacks in end-to-end
communications systems. Specifically, we utilize a generative network to model
a powerful adversary and enable the end-to-end communications system to combat
the generative attack network via a minimax game. We show that the proposed
system not only works well against white-box and black-box adversarial attacks
but also possesses excellent generalization capabilities to maintain good
performance under no attacks. We also show that our GAN-based end-to-end system
outperforms the conventional communications system and the end-to-end
communications system with/without adversarial training.
| [
{
"created": "Wed, 3 Mar 2021 20:04:42 GMT",
"version": "v1"
}
] | 2022-08-16 | [
[
"Dong",
"Yudi",
""
],
[
"Wang",
"Huaxia",
""
],
[
"Yao",
"Yu-Dong",
""
]
] | We propose a novel defensive mechanism based on a generative adversarial network (GAN) framework to defend against adversarial attacks in end-to-end communications systems. Specifically, we utilize a generative network to model a powerful adversary and enable the end-to-end communications system to combat the generative attack network via a minimax game. We show that the proposed system not only works well against white-box and black-box adversarial attacks but also possesses excellent generalization capabilities to maintain good performance under no attacks. We also show that our GAN-based end-to-end system outperforms the conventional communications system and the end-to-end communications system with/without adversarial training. |
2311.14534 | Ali Ismail-Fawaz | Ali Ismail-Fawaz, Maxime Devanne, Stefano Berretti, Jonathan Weber,
Germain Forestier | Finding Foundation Models for Time Series Classification with a PreText
Task | null | null | null | null | cs.LG | http://creativecommons.org/licenses/by/4.0/ | Over the past decade, Time Series Classification (TSC) has gained an
increasing attention. While various methods were explored, deep learning -
particularly through Convolutional Neural Networks (CNNs)-stands out as an
effective approach. However, due to the limited availability of training data,
defining a foundation model for TSC that overcomes the overfitting problem is
still a challenging task. The UCR archive, encompassing a wide spectrum of
datasets ranging from motion recognition to ECG-based heart disease detection,
serves as a prime example for exploring this issue in diverse TSC scenarios. In
this paper, we address the overfitting challenge by introducing pre-trained
domain foundation models. A key aspect of our methodology is a novel pretext
task that spans multiple datasets. This task is designed to identify the
originating dataset of each time series sample, with the goal of creating
flexible convolution filters that can be applied across different datasets. The
research process consists of two phases: a pre-training phase where the model
acquires general features through the pretext task, and a subsequent
fine-tuning phase for specific dataset classifications. Our extensive
experiments on the UCR archive demonstrate that this pre-training strategy
significantly outperforms the conventional training approach without
pre-training. This strategy effectively reduces overfitting in small datasets
and provides an efficient route for adapting these models to new datasets, thus
advancing the capabilities of deep learning in TSC.
| [
{
"created": "Fri, 24 Nov 2023 15:03:55 GMT",
"version": "v1"
},
{
"created": "Wed, 28 Feb 2024 13:58:20 GMT",
"version": "v2"
}
] | 2024-02-29 | [
[
"Ismail-Fawaz",
"Ali",
""
],
[
"Devanne",
"Maxime",
""
],
[
"Berretti",
"Stefano",
""
],
[
"Weber",
"Jonathan",
""
],
[
"Forestier",
"Germain",
""
]
] | Over the past decade, Time Series Classification (TSC) has gained an increasing attention. While various methods were explored, deep learning - particularly through Convolutional Neural Networks (CNNs)-stands out as an effective approach. However, due to the limited availability of training data, defining a foundation model for TSC that overcomes the overfitting problem is still a challenging task. The UCR archive, encompassing a wide spectrum of datasets ranging from motion recognition to ECG-based heart disease detection, serves as a prime example for exploring this issue in diverse TSC scenarios. In this paper, we address the overfitting challenge by introducing pre-trained domain foundation models. A key aspect of our methodology is a novel pretext task that spans multiple datasets. This task is designed to identify the originating dataset of each time series sample, with the goal of creating flexible convolution filters that can be applied across different datasets. The research process consists of two phases: a pre-training phase where the model acquires general features through the pretext task, and a subsequent fine-tuning phase for specific dataset classifications. Our extensive experiments on the UCR archive demonstrate that this pre-training strategy significantly outperforms the conventional training approach without pre-training. This strategy effectively reduces overfitting in small datasets and provides an efficient route for adapting these models to new datasets, thus advancing the capabilities of deep learning in TSC. |
2306.13064 | Kate Boxer | Kate S. Boxer, Edward McFowland III, Daniel B. Neill | Auditing Predictive Models for Intersectional Biases | 29 pages, 7 figures | null | null | null | cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Predictive models that satisfy group fairness criteria in aggregate for
members of a protected class, but do not guarantee subgroup fairness, could
produce biased predictions for individuals at the intersection of two or more
protected classes. To address this risk, we propose Conditional Bias Scan
(CBS), a flexible auditing framework for detecting intersectional biases in
classification models. CBS identifies the subgroup for which there is the most
significant bias against the protected class, as compared to the equivalent
subgroup in the non-protected class, and can incorporate multiple commonly used
fairness definitions for both probabilistic and binarized predictions. We show
that this methodology can detect previously unidentified intersectional and
contextual biases in the COMPAS pre-trial risk assessment tool and has higher
bias detection power compared to similar methods that audit for subgroup
fairness.
| [
{
"created": "Thu, 22 Jun 2023 17:32:12 GMT",
"version": "v1"
}
] | 2023-06-23 | [
[
"Boxer",
"Kate S.",
""
],
[
"McFowland",
"Edward",
"III"
],
[
"Neill",
"Daniel B.",
""
]
] | Predictive models that satisfy group fairness criteria in aggregate for members of a protected class, but do not guarantee subgroup fairness, could produce biased predictions for individuals at the intersection of two or more protected classes. To address this risk, we propose Conditional Bias Scan (CBS), a flexible auditing framework for detecting intersectional biases in classification models. CBS identifies the subgroup for which there is the most significant bias against the protected class, as compared to the equivalent subgroup in the non-protected class, and can incorporate multiple commonly used fairness definitions for both probabilistic and binarized predictions. We show that this methodology can detect previously unidentified intersectional and contextual biases in the COMPAS pre-trial risk assessment tool and has higher bias detection power compared to similar methods that audit for subgroup fairness. |
2310.19182 | Junjiao Tian | Junjiao Tian, Yen-Cheng Liu, James Seale Smith, Zsolt Kira | Fast Trainable Projection for Robust Fine-Tuning | Accepted to NeurIPS 2023 | null | null | null | cs.CV | http://creativecommons.org/licenses/by/4.0/ | Robust fine-tuning aims to achieve competitive in-distribution (ID)
performance while maintaining the out-of-distribution (OOD) robustness of a
pre-trained model when transferring it to a downstream task. Recently,
projected gradient descent has been successfully used in robust fine-tuning by
constraining the deviation from the initialization of the fine-tuned model
explicitly through projection. However, algorithmically, two limitations
prevent this method from being adopted more widely, scalability and efficiency.
In this paper, we propose a new projection-based fine-tuning algorithm, Fast
Trainable Projection (FTP) for computationally efficient learning of per-layer
projection constraints, resulting in an average $35\%$ speedup on our
benchmarks compared to prior works. FTP can be combined with existing
optimizers such as AdamW, and be used in a plug-and-play fashion. Finally, we
show that FTP is a special instance of hyper-optimizers that tune the
hyper-parameters of optimizers in a learnable manner through nested
differentiation. Empirically, we show superior robustness on OOD datasets,
including domain shifts and natural corruptions, across four different vision
tasks with five different pre-trained models. Additionally, we demonstrate that
FTP is broadly applicable and beneficial to other learning scenarios such as
low-label and continual learning settings thanks to its easy adaptability. The
code will be available at https://github.com/GT-RIPL/FTP.git.
| [
{
"created": "Sun, 29 Oct 2023 22:52:43 GMT",
"version": "v1"
}
] | 2023-10-31 | [
[
"Tian",
"Junjiao",
""
],
[
"Liu",
"Yen-Cheng",
""
],
[
"Smith",
"James Seale",
""
],
[
"Kira",
"Zsolt",
""
]
] | Robust fine-tuning aims to achieve competitive in-distribution (ID) performance while maintaining the out-of-distribution (OOD) robustness of a pre-trained model when transferring it to a downstream task. Recently, projected gradient descent has been successfully used in robust fine-tuning by constraining the deviation from the initialization of the fine-tuned model explicitly through projection. However, algorithmically, two limitations prevent this method from being adopted more widely, scalability and efficiency. In this paper, we propose a new projection-based fine-tuning algorithm, Fast Trainable Projection (FTP) for computationally efficient learning of per-layer projection constraints, resulting in an average $35\%$ speedup on our benchmarks compared to prior works. FTP can be combined with existing optimizers such as AdamW, and be used in a plug-and-play fashion. Finally, we show that FTP is a special instance of hyper-optimizers that tune the hyper-parameters of optimizers in a learnable manner through nested differentiation. Empirically, we show superior robustness on OOD datasets, including domain shifts and natural corruptions, across four different vision tasks with five different pre-trained models. Additionally, we demonstrate that FTP is broadly applicable and beneficial to other learning scenarios such as low-label and continual learning settings thanks to its easy adaptability. The code will be available at https://github.com/GT-RIPL/FTP.git. |
2205.12428 | Ivan Kobyzev | Ivan Kobyzev, Aref Jafari, Mehdi Rezagholizadeh, Tianda Li, Alan
Do-Omri, Peng Lu, Pascal Poupart, Ali Ghodsi | Do we need Label Regularization to Fine-tune Pre-trained Language
Models? | Published at EACL 2023 | null | null | null | cs.LG cs.CL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Knowledge Distillation (KD) is a prominent neural model compression technique
that heavily relies on teacher network predictions to guide the training of a
student model. Considering the ever-growing size of pre-trained language models
(PLMs), KD is often adopted in many NLP tasks involving PLMs. However, it is
evident that in KD, deploying the teacher network during training adds to the
memory and computational requirements of training. In the computer vision
literature, the necessity of the teacher network is put under scrutiny by
showing that KD is a label regularization technique that can be replaced with
lighter teacher-free variants such as the label-smoothing technique. However,
to the best of our knowledge, this issue is not investigated in NLP. Therefore,
this work concerns studying different label regularization techniques and
whether we actually need them to improve the fine-tuning of smaller PLM
networks on downstream tasks. In this regard, we did a comprehensive set of
experiments on different PLMs such as BERT, RoBERTa, and GPT with more than 600
distinct trials and ran each configuration five times. This investigation led
to a surprising observation that KD and other label regularization techniques
do not play any meaningful role over regular fine-tuning when the student model
is pre-trained. We further explore this phenomenon in different settings of NLP
and computer vision tasks and demonstrate that pre-training itself acts as a
kind of regularization, and additional label regularization is unnecessary.
| [
{
"created": "Wed, 25 May 2022 01:26:31 GMT",
"version": "v1"
},
{
"created": "Wed, 12 Apr 2023 15:34:03 GMT",
"version": "v2"
}
] | 2023-04-13 | [
[
"Kobyzev",
"Ivan",
""
],
[
"Jafari",
"Aref",
""
],
[
"Rezagholizadeh",
"Mehdi",
""
],
[
"Li",
"Tianda",
""
],
[
"Do-Omri",
"Alan",
""
],
[
"Lu",
"Peng",
""
],
[
"Poupart",
"Pascal",
""
],
[
"Ghodsi",
"Ali",
""
]
] | Knowledge Distillation (KD) is a prominent neural model compression technique that heavily relies on teacher network predictions to guide the training of a student model. Considering the ever-growing size of pre-trained language models (PLMs), KD is often adopted in many NLP tasks involving PLMs. However, it is evident that in KD, deploying the teacher network during training adds to the memory and computational requirements of training. In the computer vision literature, the necessity of the teacher network is put under scrutiny by showing that KD is a label regularization technique that can be replaced with lighter teacher-free variants such as the label-smoothing technique. However, to the best of our knowledge, this issue is not investigated in NLP. Therefore, this work concerns studying different label regularization techniques and whether we actually need them to improve the fine-tuning of smaller PLM networks on downstream tasks. In this regard, we did a comprehensive set of experiments on different PLMs such as BERT, RoBERTa, and GPT with more than 600 distinct trials and ran each configuration five times. This investigation led to a surprising observation that KD and other label regularization techniques do not play any meaningful role over regular fine-tuning when the student model is pre-trained. We further explore this phenomenon in different settings of NLP and computer vision tasks and demonstrate that pre-training itself acts as a kind of regularization, and additional label regularization is unnecessary. |
2112.11117 | Konrad Kollnig | Konrad Kollnig, Reuben Binns, Max Van Kleek, Ulrik Lyngs, Jun Zhao,
Claudine Tinsman, Nigel Shadbolt | Before and after GDPR: tracking in mobile apps | null | Internet Policy Review, 2021, 10(4) | 10.14763/2021.4.1611 | null | cs.CY | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Third-party tracking, the collection and sharing of behavioural data about
individuals, is a significant and ubiquitous privacy threat in mobile apps. The
EU General Data Protection Regulation (GDPR) was introduced in 2018 to protect
personal data better, but there exists, thus far, limited empirical evidence
about its efficacy. This paper studies tracking in nearly two million Android
apps from before and after the introduction of the GDPR. Our analysis suggests
that there has been limited change in the presence of third-party tracking in
apps, and that the concentration of tracking capabilities among a few large
gatekeeper companies persists. However, change might be imminent.
| [
{
"created": "Tue, 21 Dec 2021 11:45:01 GMT",
"version": "v1"
}
] | 2021-12-22 | [
[
"Kollnig",
"Konrad",
""
],
[
"Binns",
"Reuben",
""
],
[
"Van Kleek",
"Max",
""
],
[
"Lyngs",
"Ulrik",
""
],
[
"Zhao",
"Jun",
""
],
[
"Tinsman",
"Claudine",
""
],
[
"Shadbolt",
"Nigel",
""
]
] | Third-party tracking, the collection and sharing of behavioural data about individuals, is a significant and ubiquitous privacy threat in mobile apps. The EU General Data Protection Regulation (GDPR) was introduced in 2018 to protect personal data better, but there exists, thus far, limited empirical evidence about its efficacy. This paper studies tracking in nearly two million Android apps from before and after the introduction of the GDPR. Our analysis suggests that there has been limited change in the presence of third-party tracking in apps, and that the concentration of tracking capabilities among a few large gatekeeper companies persists. However, change might be imminent. |
2202.05413 | Yun-Hsin Kuo | Yun-Hsin Kuo, Takanori Fujiwara, Charles C.-K. Chou, Chun-houh Chen,
Kwan-Liu Ma | A Machine-Learning-Aided Visual Analysis Workflow for Investigating Air
Pollution Data | To appear in the Proceedings of IEEE PacificVis 2022 | null | null | null | cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Analyzing air pollution data is challenging as there are various analysis
focuses from different aspects: feature (what), space (where), and time (when).
As in most geospatial analysis problems, besides high-dimensional features, the
temporal and spatial dependencies of air pollution induce the complexity of
performing analysis. Machine learning methods, such as dimensionality
reduction, can extract and summarize important information of the data to lift
the burden of understanding such a complicated environment. In this paper, we
present a methodology that utilizes multiple machine learning methods to
uniformly explore these aspects. With this methodology, we develop a visual
analytic system that supports a flexible analysis workflow, allowing domain
experts to freely explore different aspects based on their analysis needs. We
demonstrate the capability of our system and analysis workflow supporting a
variety of analysis tasks with multiple use cases.
| [
{
"created": "Fri, 11 Feb 2022 02:24:21 GMT",
"version": "v1"
}
] | 2022-02-14 | [
[
"Kuo",
"Yun-Hsin",
""
],
[
"Fujiwara",
"Takanori",
""
],
[
"Chou",
"Charles C. -K.",
""
],
[
"Chen",
"Chun-houh",
""
],
[
"Ma",
"Kwan-Liu",
""
]
] | Analyzing air pollution data is challenging as there are various analysis focuses from different aspects: feature (what), space (where), and time (when). As in most geospatial analysis problems, besides high-dimensional features, the temporal and spatial dependencies of air pollution induce the complexity of performing analysis. Machine learning methods, such as dimensionality reduction, can extract and summarize important information of the data to lift the burden of understanding such a complicated environment. In this paper, we present a methodology that utilizes multiple machine learning methods to uniformly explore these aspects. With this methodology, we develop a visual analytic system that supports a flexible analysis workflow, allowing domain experts to freely explore different aspects based on their analysis needs. We demonstrate the capability of our system and analysis workflow supporting a variety of analysis tasks with multiple use cases. |
2208.10781 | Jongha Kim | Jongha Kim, Jinheon Baek, Sung Ju Hwang | Object Detection in Aerial Images with Uncertainty-Aware Graph Network | ECCV Workshop 2022 | null | null | null | cs.CV | http://creativecommons.org/licenses/by/4.0/ | In this work, we propose a novel uncertainty-aware object detection framework
with a structured-graph, where nodes and edges are denoted by objects and their
spatial-semantic similarities, respectively. Specifically, we aim to consider
relationships among objects for effectively contextualizing them. To achieve
this, we first detect objects and then measure their semantic and spatial
distances to construct an object graph, which is then represented by a graph
neural network (GNN) for refining visual CNN features for objects. However,
refining CNN features and detection results of every object are inefficient and
may not be necessary, as that include correct predictions with low
uncertainties. Therefore, we propose to handle uncertain objects by not only
transferring the representation from certain objects (sources) to uncertain
objects (targets) over the directed graph, but also improving CNN features only
on objects regarded as uncertain with their representational outputs from the
GNN. Furthermore, we calculate a training loss by giving larger weights on
uncertain objects, to concentrate on improving uncertain object predictions
while maintaining high performances on certain objects. We refer to our model
as Uncertainty-Aware Graph network for object DETection (UAGDet). We then
experimentally validate ours on the challenging large-scale aerial image
dataset, namely DOTA, that consists of lots of objects with small to large
sizes in an image, on which ours improves the performance of the existing
object detection network.
| [
{
"created": "Tue, 23 Aug 2022 07:29:03 GMT",
"version": "v1"
},
{
"created": "Wed, 24 Aug 2022 05:45:37 GMT",
"version": "v2"
}
] | 2022-08-25 | [
[
"Kim",
"Jongha",
""
],
[
"Baek",
"Jinheon",
""
],
[
"Hwang",
"Sung Ju",
""
]
] | In this work, we propose a novel uncertainty-aware object detection framework with a structured-graph, where nodes and edges are denoted by objects and their spatial-semantic similarities, respectively. Specifically, we aim to consider relationships among objects for effectively contextualizing them. To achieve this, we first detect objects and then measure their semantic and spatial distances to construct an object graph, which is then represented by a graph neural network (GNN) for refining visual CNN features for objects. However, refining CNN features and detection results of every object are inefficient and may not be necessary, as that include correct predictions with low uncertainties. Therefore, we propose to handle uncertain objects by not only transferring the representation from certain objects (sources) to uncertain objects (targets) over the directed graph, but also improving CNN features only on objects regarded as uncertain with their representational outputs from the GNN. Furthermore, we calculate a training loss by giving larger weights on uncertain objects, to concentrate on improving uncertain object predictions while maintaining high performances on certain objects. We refer to our model as Uncertainty-Aware Graph network for object DETection (UAGDet). We then experimentally validate ours on the challenging large-scale aerial image dataset, namely DOTA, that consists of lots of objects with small to large sizes in an image, on which ours improves the performance of the existing object detection network. |
2308.13663 | Paula Mercurio | Paula Mercurio and Di Liu | Network Embedding Using Sparse Approximations of Random Walks | 20 pages, 4 figures | null | null | null | cs.LG | http://creativecommons.org/licenses/by/4.0/ | In this paper, we propose an efficient numerical implementation of Network
Embedding based on commute times, using sparse approximation of a diffusion
process on the network obtained by a modified version of the diffusion wavelet
algorithm. The node embeddings are computed by optimizing the cross entropy
loss via the stochastic gradient descent method with sampling of
low-dimensional representations of green functions. We demonstrate the efficacy
of this method for data clustering and multi-label classification through
several examples, and compare its performance over existing methods in terms of
efficiency and accuracy. Theoretical issues justifying the scheme are also
discussed.
| [
{
"created": "Fri, 25 Aug 2023 20:35:45 GMT",
"version": "v1"
}
] | 2023-08-29 | [
[
"Mercurio",
"Paula",
""
],
[
"Liu",
"Di",
""
]
] | In this paper, we propose an efficient numerical implementation of Network Embedding based on commute times, using sparse approximation of a diffusion process on the network obtained by a modified version of the diffusion wavelet algorithm. The node embeddings are computed by optimizing the cross entropy loss via the stochastic gradient descent method with sampling of low-dimensional representations of green functions. We demonstrate the efficacy of this method for data clustering and multi-label classification through several examples, and compare its performance over existing methods in terms of efficiency and accuracy. Theoretical issues justifying the scheme are also discussed. |
1812.10668 | \'Alvaro L\'opez Garc\'ia | \'Alvaro L\'opez Garc\'ia, Enol Fern\'andez-del-Castillo, Isabel
Campos Plasencia | An efficient cloud scheduler design supporting preemptible instances | null | Future Generation Computer Systems (2019) | 10.1016/j.future.2018.12.057 | null | cs.DC | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Maximizing resource utilization by performing an efficient resource
provisioning is a key factor for any cloud provider: commercial actors can
maximize their revenues, whereas scientific and non-commercial providers can
maximize their infrastructure utilization. Traditionally, batch systems have
allowed data centers to fill their resources as much as possible by using
backfilling and similar techniques. However, in an IaaS cloud, where virtual
machines are supposed to live indefinitely, or at least as long as the user is
able to pay for them, these policies are not easily implementable. In this work
we present a new scheduling algorithm for IaaS providers that is able to
support preemptible instances, that can be stopped by higher priority requests
without introducing large modifications in the current cloud schedulers. This
scheduler enables the implementation of new cloud usage and payment models that
allow more efficient usage of the resources and potential new revenue sources
for commercial providers. We also study the correctness and the performace
overhead of the proposed scheduler agains existing solutions.
| [
{
"created": "Thu, 27 Dec 2018 09:14:37 GMT",
"version": "v1"
},
{
"created": "Thu, 3 Jan 2019 10:10:37 GMT",
"version": "v2"
},
{
"created": "Tue, 28 Jan 2020 08:42:41 GMT",
"version": "v3"
}
] | 2020-01-29 | [
[
"García",
"Álvaro López",
""
],
[
"Fernández-del-Castillo",
"Enol",
""
],
[
"Plasencia",
"Isabel Campos",
""
]
] | Maximizing resource utilization by performing an efficient resource provisioning is a key factor for any cloud provider: commercial actors can maximize their revenues, whereas scientific and non-commercial providers can maximize their infrastructure utilization. Traditionally, batch systems have allowed data centers to fill their resources as much as possible by using backfilling and similar techniques. However, in an IaaS cloud, where virtual machines are supposed to live indefinitely, or at least as long as the user is able to pay for them, these policies are not easily implementable. In this work we present a new scheduling algorithm for IaaS providers that is able to support preemptible instances, that can be stopped by higher priority requests without introducing large modifications in the current cloud schedulers. This scheduler enables the implementation of new cloud usage and payment models that allow more efficient usage of the resources and potential new revenue sources for commercial providers. We also study the correctness and the performace overhead of the proposed scheduler agains existing solutions. |
2204.01807 | Scott Workman | Scott Workman, M. Usman Rafique, Hunter Blanton, Nathan Jacobs | Revisiting Near/Remote Sensing with Geospatial Attention | IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
2022 | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | This work addresses the task of overhead image segmentation when auxiliary
ground-level images are available. Recent work has shown that performing joint
inference over these two modalities, often called near/remote sensing, can
yield significant accuracy improvements. Extending this line of work, we
introduce the concept of geospatial attention, a geometry-aware attention
mechanism that explicitly considers the geospatial relationship between the
pixels in a ground-level image and a geographic location. We propose an
approach for computing geospatial attention that incorporates geometric
features and the appearance of the overhead and ground-level imagery. We
introduce a novel architecture for near/remote sensing that is based on
geospatial attention and demonstrate its use for five segmentation tasks. The
results demonstrate that our method significantly outperforms the previous
state-of-the-art methods.
| [
{
"created": "Mon, 4 Apr 2022 19:19:50 GMT",
"version": "v1"
}
] | 2022-04-06 | [
[
"Workman",
"Scott",
""
],
[
"Rafique",
"M. Usman",
""
],
[
"Blanton",
"Hunter",
""
],
[
"Jacobs",
"Nathan",
""
]
] | This work addresses the task of overhead image segmentation when auxiliary ground-level images are available. Recent work has shown that performing joint inference over these two modalities, often called near/remote sensing, can yield significant accuracy improvements. Extending this line of work, we introduce the concept of geospatial attention, a geometry-aware attention mechanism that explicitly considers the geospatial relationship between the pixels in a ground-level image and a geographic location. We propose an approach for computing geospatial attention that incorporates geometric features and the appearance of the overhead and ground-level imagery. We introduce a novel architecture for near/remote sensing that is based on geospatial attention and demonstrate its use for five segmentation tasks. The results demonstrate that our method significantly outperforms the previous state-of-the-art methods. |
1603.00816 | Kezhi Li | Kezhi Li, Daniel Holland | A Nonlinear Weighted Total Variation Image Reconstruction Algorithm for
Electrical Capacitance Tomography | null | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | A new iterative image reconstruction algorithm for electrical capacitance
tomography (ECT) is proposed that is based on iterative soft thresholding of a
total variation penalty and adaptive reweighted compressive sensing. This
algorithm encourages sharp changes in the ECT image and overcomes the
disadvantage of the $l_1$ minimization by equipping the total variation with an
adaptive weighting depending on the reconstructed image. Moreover, the
non-linear effect is also partially reduced due to the adoption of an updated
sensitivity matrix. Simulation results show that the proposed algorithm
recovers ECT images more precisely than existing state-of-the-art algorithms
and therefore is suitable for the imaging of multiphase systems in industrial
or medical applications.
| [
{
"created": "Wed, 2 Mar 2016 18:24:32 GMT",
"version": "v1"
},
{
"created": "Mon, 21 Nov 2016 15:01:41 GMT",
"version": "v2"
}
] | 2016-11-22 | [
[
"Li",
"Kezhi",
""
],
[
"Holland",
"Daniel",
""
]
] | A new iterative image reconstruction algorithm for electrical capacitance tomography (ECT) is proposed that is based on iterative soft thresholding of a total variation penalty and adaptive reweighted compressive sensing. This algorithm encourages sharp changes in the ECT image and overcomes the disadvantage of the $l_1$ minimization by equipping the total variation with an adaptive weighting depending on the reconstructed image. Moreover, the non-linear effect is also partially reduced due to the adoption of an updated sensitivity matrix. Simulation results show that the proposed algorithm recovers ECT images more precisely than existing state-of-the-art algorithms and therefore is suitable for the imaging of multiphase systems in industrial or medical applications. |
2010.01478 | Shruthi Chari | Shruthi Chari, Oshani Seneviratne, Daniel M. Gruen, Morgan A. Foreman,
Amar K. Das, Deborah L. McGuinness | Explanation Ontology in Action: A Clinical Use-Case | 5 pages, 2 figures, 1 protocol | International Semantic Web Conference, Poster and Demo Track, 2020 | null | null | cs.AI cs.HC cs.LG | http://creativecommons.org/licenses/by/4.0/ | We addressed the problem of a lack of semantic representation for
user-centric explanations and different explanation types in our Explanation
Ontology (https://purl.org/heals/eo). Such a representation is increasingly
necessary as explainability has become an important problem in Artificial
Intelligence with the emergence of complex methods and an uptake in
high-precision and user-facing settings. In this submission, we provide
step-by-step guidance for system designers to utilize our ontology, introduced
in our resource track paper, to plan and model for explanations during the
design of their Artificial Intelligence systems. We also provide a detailed
example with our utilization of this guidance in a clinical setting.
| [
{
"created": "Sun, 4 Oct 2020 03:52:39 GMT",
"version": "v1"
}
] | 2020-10-06 | [
[
"Chari",
"Shruthi",
""
],
[
"Seneviratne",
"Oshani",
""
],
[
"Gruen",
"Daniel M.",
""
],
[
"Foreman",
"Morgan A.",
""
],
[
"Das",
"Amar K.",
""
],
[
"McGuinness",
"Deborah L.",
""
]
] | We addressed the problem of a lack of semantic representation for user-centric explanations and different explanation types in our Explanation Ontology (https://purl.org/heals/eo). Such a representation is increasingly necessary as explainability has become an important problem in Artificial Intelligence with the emergence of complex methods and an uptake in high-precision and user-facing settings. In this submission, we provide step-by-step guidance for system designers to utilize our ontology, introduced in our resource track paper, to plan and model for explanations during the design of their Artificial Intelligence systems. We also provide a detailed example with our utilization of this guidance in a clinical setting. |
2306.09561 | Renan Leandro Fernandes | Renan Fernandes (1), Fred Freitas (1), Ivan Varzinczak (2, 3 and 4)
and Pedro PM Farias (1 and 5) ((1) Centro de Inform\'atica - Universidade
Federal de Pernambuco, (2) LIASD - Universit\'e Paris 8, (3) CAIR -
University of Cape Town, (4) ISTI - CNR and (5) ARCE, Public Services
Regulation Agency-CE) | A connection method for a defeasible extension of $\mathcal{ALCH}$ | null | null | null | null | cs.LO | http://creativecommons.org/licenses/by-nc-nd/4.0/ | This paper proposes a connection method \`a la Bibel for an
exception-tolerant family of description logics (DLs). As for the language, we
assume the DL $\mathcal{ALCH}$ extended with two typicality operators: one on
(complex) concepts and one on role names. The language is a variant of
defeasible DLs, as broadly studied in the literature over the past decade, in
which most of these can be embedded. We revisit the definition of the matrix
representation of a knowledge base and establish the conditions for a given
axiom to be provable. We show that the calculus terminates and is sound and
complete w.r.t. a DL version of the preferential semantics widely adopted in
non-monotonic reasoning.
| [
{
"created": "Fri, 16 Jun 2023 00:32:14 GMT",
"version": "v1"
},
{
"created": "Thu, 22 Jun 2023 13:38:48 GMT",
"version": "v2"
}
] | 2023-06-23 | [
[
"Fernandes",
"Renan",
"",
"2, 3 and 4"
],
[
"Freitas",
"Fred",
"",
"2, 3 and 4"
],
[
"Varzinczak",
"Ivan",
"",
"2, 3 and 4"
],
[
"Farias",
"Pedro PM",
"",
"1 and 5"
]
] | This paper proposes a connection method \`a la Bibel for an exception-tolerant family of description logics (DLs). As for the language, we assume the DL $\mathcal{ALCH}$ extended with two typicality operators: one on (complex) concepts and one on role names. The language is a variant of defeasible DLs, as broadly studied in the literature over the past decade, in which most of these can be embedded. We revisit the definition of the matrix representation of a knowledge base and establish the conditions for a given axiom to be provable. We show that the calculus terminates and is sound and complete w.r.t. a DL version of the preferential semantics widely adopted in non-monotonic reasoning. |
2305.07710 | Anubhav Jain | Anubhav Jain, Nasir Memon, Julian Togelius | Zero-shot racially balanced dataset generation using an existing biased
StyleGAN2 | null | null | null | null | cs.CV | http://creativecommons.org/licenses/by/4.0/ | Facial recognition systems have made significant strides thanks to data-heavy
deep learning models, but these models rely on large privacy-sensitive
datasets. Further, many of these datasets lack diversity in terms of ethnicity
and demographics, which can lead to biased models that can have serious
societal and security implications. To address these issues, we propose a
methodology that leverages the biased generative model StyleGAN2 to create
demographically diverse images of synthetic individuals. The synthetic dataset
is created using a novel evolutionary search algorithm that targets specific
demographic groups. By training face recognition models with the resulting
balanced dataset containing 50,000 identities per race (13.5 million images in
total), we can improve their performance and minimize biases that might have
been present in a model trained on a real dataset.
| [
{
"created": "Fri, 12 May 2023 18:07:10 GMT",
"version": "v1"
},
{
"created": "Mon, 18 Sep 2023 17:48:17 GMT",
"version": "v2"
}
] | 2023-09-20 | [
[
"Jain",
"Anubhav",
""
],
[
"Memon",
"Nasir",
""
],
[
"Togelius",
"Julian",
""
]
] | Facial recognition systems have made significant strides thanks to data-heavy deep learning models, but these models rely on large privacy-sensitive datasets. Further, many of these datasets lack diversity in terms of ethnicity and demographics, which can lead to biased models that can have serious societal and security implications. To address these issues, we propose a methodology that leverages the biased generative model StyleGAN2 to create demographically diverse images of synthetic individuals. The synthetic dataset is created using a novel evolutionary search algorithm that targets specific demographic groups. By training face recognition models with the resulting balanced dataset containing 50,000 identities per race (13.5 million images in total), we can improve their performance and minimize biases that might have been present in a model trained on a real dataset. |
0904.1729 | Sugumar Murugesan | Sugumar Murugesan, Philip Schniter | Joint Opportunistic Scheduling in Multi-Cellular Systems | null | null | null | null | cs.NI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We address the problem of multiuser scheduling with partial channel
information in a multi-cell environment. The scheduling problem is formulated
jointly with the ARQ based channel learning process and the intercell
interference mitigating cell breathing protocol. The optimal joint scheduling
policy under various system constraints is established. The general problem is
posed as a generalized Restless Multiarmed Bandit process and the notion of
indexability is studied. We conjecture, with numerical support, that the
multicell multiuser scheduling problem is indexable and obtain a partial
structure of the index policy.
| [
{
"created": "Fri, 10 Apr 2009 18:55:12 GMT",
"version": "v1"
}
] | 2009-04-13 | [
[
"Murugesan",
"Sugumar",
""
],
[
"Schniter",
"Philip",
""
]
] | We address the problem of multiuser scheduling with partial channel information in a multi-cell environment. The scheduling problem is formulated jointly with the ARQ based channel learning process and the intercell interference mitigating cell breathing protocol. The optimal joint scheduling policy under various system constraints is established. The general problem is posed as a generalized Restless Multiarmed Bandit process and the notion of indexability is studied. We conjecture, with numerical support, that the multicell multiuser scheduling problem is indexable and obtain a partial structure of the index policy. |
2111.05505 | Zhikun Chen | Zhikun Chen, Daofeng Li, Jinkang Zhu and Sihai Zhang | DACFL: Dynamic Average Consensus Based Federated Learning in
Decentralized Topology | null | Sensors 2022, 22(9), 3317 | 10.3390/s22093317 | null | cs.LG | http://creativecommons.org/licenses/by/4.0/ | Federated learning (FL) is a burgeoning distributed machine learning
framework where a central parameter server (PS) coordinates many local users to
train a globally consistent model. Conventional federated learning inevitably
relies on a centralized topology with a PS. As a result, it will paralyze once
the PS fails. To alleviate such a single point failure, especially on the PS,
some existing work has provided decentralized FL (DFL) implementations like
CDSGD and D-PSGD to facilitate FL in a decentralized topology. However, there
are still some problems with these methods, e.g., significant divergence
between users' final models in CDSGD and a network-wide model average necessity
in D-PSGD. In order to solve these deficiency, this paper devises a new DFL
implementation coined as DACFL, where each user trains its model using its own
training data and exchanges the intermediate models with its neighbors through
a symmetric and doubly stochastic matrix. The DACFL treats the progress of each
user's local training as a discrete-time process and employs a first order
dynamic average consensus (FODAC) method to track the \textit{average model} in
the absence of the PS. In this paper, we also provide a theoretical convergence
analysis of DACFL on the premise of i.i.d data to strengthen its rationality.
The experimental results on MNIST, Fashion-MNIST and CIFAR-10 validate the
feasibility of our solution in both time-invariant and time-varying network
topologies, and declare that DACFL outperforms D-PSGD and CDSGD in most cases.
| [
{
"created": "Wed, 10 Nov 2021 03:00:40 GMT",
"version": "v1"
}
] | 2023-12-13 | [
[
"Chen",
"Zhikun",
""
],
[
"Li",
"Daofeng",
""
],
[
"Zhu",
"Jinkang",
""
],
[
"Zhang",
"Sihai",
""
]
] | Federated learning (FL) is a burgeoning distributed machine learning framework where a central parameter server (PS) coordinates many local users to train a globally consistent model. Conventional federated learning inevitably relies on a centralized topology with a PS. As a result, it will paralyze once the PS fails. To alleviate such a single point failure, especially on the PS, some existing work has provided decentralized FL (DFL) implementations like CDSGD and D-PSGD to facilitate FL in a decentralized topology. However, there are still some problems with these methods, e.g., significant divergence between users' final models in CDSGD and a network-wide model average necessity in D-PSGD. In order to solve these deficiency, this paper devises a new DFL implementation coined as DACFL, where each user trains its model using its own training data and exchanges the intermediate models with its neighbors through a symmetric and doubly stochastic matrix. The DACFL treats the progress of each user's local training as a discrete-time process and employs a first order dynamic average consensus (FODAC) method to track the \textit{average model} in the absence of the PS. In this paper, we also provide a theoretical convergence analysis of DACFL on the premise of i.i.d data to strengthen its rationality. The experimental results on MNIST, Fashion-MNIST and CIFAR-10 validate the feasibility of our solution in both time-invariant and time-varying network topologies, and declare that DACFL outperforms D-PSGD and CDSGD in most cases. |
2101.02429 | Burak Bartan | Burak Bartan, Mert Pilanci | Neural Spectrahedra and Semidefinite Lifts: Global Convex Optimization
of Polynomial Activation Neural Networks in Fully Polynomial-Time | null | null | null | null | cs.LG cs.CC math.OC stat.ML | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The training of two-layer neural networks with nonlinear activation functions
is an important non-convex optimization problem with numerous applications and
promising performance in layerwise deep learning. In this paper, we develop
exact convex optimization formulations for two-layer neural networks with
second degree polynomial activations based on semidefinite programming.
Remarkably, we show that semidefinite lifting is always exact and therefore
computational complexity for global optimization is polynomial in the input
dimension and sample size for all input data. The developed convex formulations
are proven to achieve the same global optimal solution set as their non-convex
counterparts. More specifically, the globally optimal two-layer neural network
with polynomial activations can be found by solving a semidefinite program
(SDP) and decomposing the solution using a procedure we call Neural
Decomposition. Moreover, the choice of regularizers plays a crucial role in the
computational tractability of neural network training. We show that the
standard weight decay regularization formulation is NP-hard, whereas other
simple convex penalties render the problem tractable in polynomial time via
convex programming. We extend the results beyond the fully connected
architecture to different neural network architectures including networks with
vector outputs and convolutional architectures with pooling. We provide
extensive numerical simulations showing that the standard backpropagation
approach often fails to achieve the global optimum of the training loss. The
proposed approach is significantly faster to obtain better test accuracy
compared to the standard backpropagation procedure.
| [
{
"created": "Thu, 7 Jan 2021 08:43:01 GMT",
"version": "v1"
}
] | 2021-01-11 | [
[
"Bartan",
"Burak",
""
],
[
"Pilanci",
"Mert",
""
]
] | The training of two-layer neural networks with nonlinear activation functions is an important non-convex optimization problem with numerous applications and promising performance in layerwise deep learning. In this paper, we develop exact convex optimization formulations for two-layer neural networks with second degree polynomial activations based on semidefinite programming. Remarkably, we show that semidefinite lifting is always exact and therefore computational complexity for global optimization is polynomial in the input dimension and sample size for all input data. The developed convex formulations are proven to achieve the same global optimal solution set as their non-convex counterparts. More specifically, the globally optimal two-layer neural network with polynomial activations can be found by solving a semidefinite program (SDP) and decomposing the solution using a procedure we call Neural Decomposition. Moreover, the choice of regularizers plays a crucial role in the computational tractability of neural network training. We show that the standard weight decay regularization formulation is NP-hard, whereas other simple convex penalties render the problem tractable in polynomial time via convex programming. We extend the results beyond the fully connected architecture to different neural network architectures including networks with vector outputs and convolutional architectures with pooling. We provide extensive numerical simulations showing that the standard backpropagation approach often fails to achieve the global optimum of the training loss. The proposed approach is significantly faster to obtain better test accuracy compared to the standard backpropagation procedure. |
1803.01504 | Dan Xu | Dan Xu, Xavier Alameda-Pineda, Jingkuan Song, Elisa Ricci, Nicu Sebe | Cross-Paced Representation Learning with Partial Curricula for
Sketch-based Image Retrieval | null | null | 10.1109/TIP.2018.2837381 | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In this paper we address the problem of learning robust cross-domain
representations for sketch-based image retrieval (SBIR). While most SBIR
approaches focus on extracting low- and mid-level descriptors for direct
feature matching, recent works have shown the benefit of learning coupled
feature representations to describe data from two related sources. However,
cross-domain representation learning methods are typically cast into non-convex
minimization problems that are difficult to optimize, leading to unsatisfactory
performance. Inspired by self-paced learning, a learning methodology designed
to overcome convergence issues related to local optima by exploiting the
samples in a meaningful order (i.e. easy to hard), we introduce the cross-paced
partial curriculum learning (CPPCL) framework. Compared with existing
self-paced learning methods which only consider a single modality and cannot
deal with prior knowledge, CPPCL is specifically designed to assess the
learning pace by jointly handling data from dual sources and modality-specific
prior information provided in the form of partial curricula. Additionally,
thanks to the learned dictionaries, we demonstrate that the proposed CPPCL
embeds robust coupled representations for SBIR. Our approach is extensively
evaluated on four publicly available datasets (i.e. CUFS, Flickr15K, QueenMary
SBIR and TU-Berlin Extension datasets), showing superior performance over
competing SBIR methods.
| [
{
"created": "Mon, 5 Mar 2018 05:30:08 GMT",
"version": "v1"
}
] | 2018-08-01 | [
[
"Xu",
"Dan",
""
],
[
"Alameda-Pineda",
"Xavier",
""
],
[
"Song",
"Jingkuan",
""
],
[
"Ricci",
"Elisa",
""
],
[
"Sebe",
"Nicu",
""
]
] | In this paper we address the problem of learning robust cross-domain representations for sketch-based image retrieval (SBIR). While most SBIR approaches focus on extracting low- and mid-level descriptors for direct feature matching, recent works have shown the benefit of learning coupled feature representations to describe data from two related sources. However, cross-domain representation learning methods are typically cast into non-convex minimization problems that are difficult to optimize, leading to unsatisfactory performance. Inspired by self-paced learning, a learning methodology designed to overcome convergence issues related to local optima by exploiting the samples in a meaningful order (i.e. easy to hard), we introduce the cross-paced partial curriculum learning (CPPCL) framework. Compared with existing self-paced learning methods which only consider a single modality and cannot deal with prior knowledge, CPPCL is specifically designed to assess the learning pace by jointly handling data from dual sources and modality-specific prior information provided in the form of partial curricula. Additionally, thanks to the learned dictionaries, we demonstrate that the proposed CPPCL embeds robust coupled representations for SBIR. Our approach is extensively evaluated on four publicly available datasets (i.e. CUFS, Flickr15K, QueenMary SBIR and TU-Berlin Extension datasets), showing superior performance over competing SBIR methods. |
1507.04885 | Akbar Rafiey | Akbar Rafiey, Jeff Kinne, J\'an Manuch, Arash Rafiey | Ordering with precedence constraints and budget minimization | null | null | null | null | cs.DM | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We introduce a variation of the scheduling with precedence constraints
problem that has applications to molecular folding and production management.
We are given a bipartite graph $H=(B,S)$. Vertices in $B$ are thought of as
goods or services that must be \emph{bought} to produce items in $S$ that are
to be \emph{sold}. An edge from $j\in S$ to $i\in B$ indicates that the
production of $j$ requires the purchase of $i$. Each vertex in $B$ has a cost,
and each vertex in $S$ results in some gain. The goal is to obtain an ordering
of $B\cup S$ that respects the precedence constraints and maximizes the minimal
net profit encountered as the vertices are processed. We call this optimal
value the \emph{budget} or \emph{capital} investment required for the bipartite
graph, and refer to our problem as \emph{the bipartite graph ordering problem}.
The problem is equivalent to a version of an NP-complete molecular folding
problem that has been studied recently [12]. Work on the molecular folding
problem has focused on heuristic algorithms and exponential-time exact
algorithms for the un-weighted problem where costs are $\pm 1$ and when
restricted to graphs arising from RNA folding.
The bipartite graph present work seeks exact algorithms for solving the
bipartite ordering problem. We demonstrate an algorithm that computes the
optimal ordering in time $O^*(2^n)$ when $n$ is the number of vertices in the
input bipartite graph. Our main result is a general strategy that can be used
to find an optimal ordering in polynomial time for bipartite graphs that
satisfy certain properties. We apply the technique to a variety of graph
classes, obtaining polynomial-time solutions to the bipartite graph ordering
problem for bipartite permutation graphs, trivially perfect, co-bipartite
graphs, and trees.
| [
{
"created": "Fri, 17 Jul 2015 09:11:56 GMT",
"version": "v1"
},
{
"created": "Wed, 22 Jun 2016 01:21:00 GMT",
"version": "v2"
},
{
"created": "Mon, 3 Oct 2016 05:23:37 GMT",
"version": "v3"
}
] | 2016-10-04 | [
[
"Rafiey",
"Akbar",
""
],
[
"Kinne",
"Jeff",
""
],
[
"Manuch",
"Ján",
""
],
[
"Rafiey",
"Arash",
""
]
] | We introduce a variation of the scheduling with precedence constraints problem that has applications to molecular folding and production management. We are given a bipartite graph $H=(B,S)$. Vertices in $B$ are thought of as goods or services that must be \emph{bought} to produce items in $S$ that are to be \emph{sold}. An edge from $j\in S$ to $i\in B$ indicates that the production of $j$ requires the purchase of $i$. Each vertex in $B$ has a cost, and each vertex in $S$ results in some gain. The goal is to obtain an ordering of $B\cup S$ that respects the precedence constraints and maximizes the minimal net profit encountered as the vertices are processed. We call this optimal value the \emph{budget} or \emph{capital} investment required for the bipartite graph, and refer to our problem as \emph{the bipartite graph ordering problem}. The problem is equivalent to a version of an NP-complete molecular folding problem that has been studied recently [12]. Work on the molecular folding problem has focused on heuristic algorithms and exponential-time exact algorithms for the un-weighted problem where costs are $\pm 1$ and when restricted to graphs arising from RNA folding. The bipartite graph present work seeks exact algorithms for solving the bipartite ordering problem. We demonstrate an algorithm that computes the optimal ordering in time $O^*(2^n)$ when $n$ is the number of vertices in the input bipartite graph. Our main result is a general strategy that can be used to find an optimal ordering in polynomial time for bipartite graphs that satisfy certain properties. We apply the technique to a variety of graph classes, obtaining polynomial-time solutions to the bipartite graph ordering problem for bipartite permutation graphs, trivially perfect, co-bipartite graphs, and trees. |
2201.08042 | Ervin Dervishaj | Ervin Dervishaj and Paolo Cremonesi | GAN-based Matrix Factorization for Recommender Systems | Accepted at the 37th ACM/SIGAPP Symposium on Applied Computing (SAC
'22) | null | 10.1145/3477314.3507099 | null | cs.IR cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Proposed in 2014, Generative Adversarial Networks (GAN) initiated a fresh
interest in generative modelling. They immediately achieved state-of-the-art in
image synthesis, image-to-image translation, text-to-image generation, image
inpainting and have been used in sciences ranging from medicine to high-energy
particle physics. Despite their popularity and ability to learn arbitrary
distributions, GAN have not been widely applied in recommender systems (RS).
Moreover, only few of the techniques that have introduced GAN in RS have
employed them directly as a collaborative filtering (CF) model. In this work we
propose a new GAN-based approach that learns user and item latent factors in a
matrix factorization setting for the generic top-N recommendation problem.
Following the vector-wise GAN training approach for RS introduced by CFGAN, we
identify 2 unique issues when utilizing GAN for CF. We propose solutions for
both of them by using an autoencoder as discriminator and incorporating an
additional loss function for the generator. We evaluate our model, GANMF,
through well-known datasets in the RS community and show improvements over
traditional CF approaches and GAN-based models. Through an ablation study on
the components of GANMF we aim to understand the effects of our architectural
choices. Finally, we provide a qualitative evaluation of the matrix
factorization performance of GANMF.
| [
{
"created": "Thu, 20 Jan 2022 08:14:29 GMT",
"version": "v1"
}
] | 2022-01-21 | [
[
"Dervishaj",
"Ervin",
""
],
[
"Cremonesi",
"Paolo",
""
]
] | Proposed in 2014, Generative Adversarial Networks (GAN) initiated a fresh interest in generative modelling. They immediately achieved state-of-the-art in image synthesis, image-to-image translation, text-to-image generation, image inpainting and have been used in sciences ranging from medicine to high-energy particle physics. Despite their popularity and ability to learn arbitrary distributions, GAN have not been widely applied in recommender systems (RS). Moreover, only few of the techniques that have introduced GAN in RS have employed them directly as a collaborative filtering (CF) model. In this work we propose a new GAN-based approach that learns user and item latent factors in a matrix factorization setting for the generic top-N recommendation problem. Following the vector-wise GAN training approach for RS introduced by CFGAN, we identify 2 unique issues when utilizing GAN for CF. We propose solutions for both of them by using an autoencoder as discriminator and incorporating an additional loss function for the generator. We evaluate our model, GANMF, through well-known datasets in the RS community and show improvements over traditional CF approaches and GAN-based models. Through an ablation study on the components of GANMF we aim to understand the effects of our architectural choices. Finally, we provide a qualitative evaluation of the matrix factorization performance of GANMF. |
2302.12170 | Joel Lehman | Elliot Meyerson and Mark J. Nelson and Herbie Bradley and Adam Gaier
and Arash Moradi and Amy K. Hoover and Joel Lehman | Language Model Crossover: Variation through Few-Shot Prompting | null | null | null | null | cs.NE | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | This paper pursues the insight that language models naturally enable an
intelligent variation operator similar in spirit to evolutionary crossover. In
particular, language models of sufficient scale demonstrate in-context
learning, i.e. they can learn from associations between a small number of input
patterns to generate outputs incorporating such associations (also called
few-shot prompting). This ability can be leveraged to form a simple but
powerful variation operator, i.e. to prompt a language model with a few
text-based genotypes (such as code, plain-text sentences, or equations), and to
parse its corresponding output as those genotypes' offspring. The promise of
such language model crossover (which is simple to implement and can leverage
many different open-source language models) is that it enables a simple
mechanism to evolve semantically-rich text representations (with few
domain-specific tweaks), and naturally benefits from current progress in
language models. Experiments in this paper highlight the versatility of
language-model crossover, through evolving binary bit-strings, sentences,
equations, text-to-image prompts, and Python code. The conclusion is that
language model crossover is a promising method for evolving genomes
representable as text.
| [
{
"created": "Thu, 23 Feb 2023 17:12:34 GMT",
"version": "v1"
},
{
"created": "Sat, 7 Oct 2023 17:07:48 GMT",
"version": "v2"
},
{
"created": "Mon, 13 May 2024 23:57:11 GMT",
"version": "v3"
}
] | 2024-05-15 | [
[
"Meyerson",
"Elliot",
""
],
[
"Nelson",
"Mark J.",
""
],
[
"Bradley",
"Herbie",
""
],
[
"Gaier",
"Adam",
""
],
[
"Moradi",
"Arash",
""
],
[
"Hoover",
"Amy K.",
""
],
[
"Lehman",
"Joel",
""
]
] | This paper pursues the insight that language models naturally enable an intelligent variation operator similar in spirit to evolutionary crossover. In particular, language models of sufficient scale demonstrate in-context learning, i.e. they can learn from associations between a small number of input patterns to generate outputs incorporating such associations (also called few-shot prompting). This ability can be leveraged to form a simple but powerful variation operator, i.e. to prompt a language model with a few text-based genotypes (such as code, plain-text sentences, or equations), and to parse its corresponding output as those genotypes' offspring. The promise of such language model crossover (which is simple to implement and can leverage many different open-source language models) is that it enables a simple mechanism to evolve semantically-rich text representations (with few domain-specific tweaks), and naturally benefits from current progress in language models. Experiments in this paper highlight the versatility of language-model crossover, through evolving binary bit-strings, sentences, equations, text-to-image prompts, and Python code. The conclusion is that language model crossover is a promising method for evolving genomes representable as text. |
2011.04696 | Fernando M. Espinoza-Cuadros | Fernando M. Espinoza-Cuadros, Juan M. Perero-Codosero, Javier
Ant\'on-Mart\'in, Luis A. Hern\'andez-G\'omez | Speaker De-identification System using Autoencoders and Adversarial
Training | null | null | null | null | cs.SD cs.CL eess.AS | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The fast increase of web services and mobile apps, which collect personal
data from users, increases the risk that their privacy may be severely
compromised. In particular, the increasing variety of spoken language
interfaces and voice assistants empowered by the vertiginous breakthroughs in
Deep Learning are prompting important concerns in the European Union to
preserve speech data privacy. For instance, an attacker can record speech from
users and impersonate them to get access to systems requiring voice
identification. Hacking speaker profiles from users is also possible by means
of existing technology to extract speaker, linguistic (e.g., dialect) and
paralinguistic features (e.g., age) from the speech signal. In order to
mitigate these weaknesses, in this paper, we propose a speaker
de-identification system based on adversarial training and autoencoders in
order to suppress speaker, gender, and accent information from speech.
Experimental results show that combining adversarial learning and autoencoders
increase the equal error rate of a speaker verification system while preserving
the intelligibility of the anonymized spoken content.
| [
{
"created": "Mon, 9 Nov 2020 19:22:05 GMT",
"version": "v1"
}
] | 2021-02-01 | [
[
"Espinoza-Cuadros",
"Fernando M.",
""
],
[
"Perero-Codosero",
"Juan M.",
""
],
[
"Antón-Martín",
"Javier",
""
],
[
"Hernández-Gómez",
"Luis A.",
""
]
] | The fast increase of web services and mobile apps, which collect personal data from users, increases the risk that their privacy may be severely compromised. In particular, the increasing variety of spoken language interfaces and voice assistants empowered by the vertiginous breakthroughs in Deep Learning are prompting important concerns in the European Union to preserve speech data privacy. For instance, an attacker can record speech from users and impersonate them to get access to systems requiring voice identification. Hacking speaker profiles from users is also possible by means of existing technology to extract speaker, linguistic (e.g., dialect) and paralinguistic features (e.g., age) from the speech signal. In order to mitigate these weaknesses, in this paper, we propose a speaker de-identification system based on adversarial training and autoencoders in order to suppress speaker, gender, and accent information from speech. Experimental results show that combining adversarial learning and autoencoders increase the equal error rate of a speaker verification system while preserving the intelligibility of the anonymized spoken content. |
2308.15235 | Nicos Isaak | Nicos Isaak | PronounFlow: A Hybrid Approach for Calibrating Pronouns in Sentences | 13 pages, 4 figures, 3 tables | null | null | null | cs.CL cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Flip through any book or listen to any song lyrics, and you will come across
pronouns that, in certain cases, can hinder meaning comprehension, especially
for machines. As the role of having cognitive machines becomes pervasive in our
lives, numerous systems have been developed to resolve pronouns under various
challenges. Commensurate with this, it is believed that having systems able to
disambiguate pronouns in sentences will help towards the endowment of machines
with commonsense and reasoning abilities like those found in humans. However,
one problem these systems face with modern English is the lack of gender
pronouns, where people try to alternate by using masculine, feminine, or plural
to avoid the whole issue. Since humanity aims to the building of systems in the
full-bodied sense we usually reserve for people, what happens when pronouns in
written text, like plural or epicene ones, refer to unspecified entities whose
gender is not necessarily known? Wouldn't that put extra barriers to existing
coreference resolution systems? Towards answering those questions, through the
implementation of a neural-symbolic system that utilizes the best of both
worlds, we are employing PronounFlow, a system that reads any English sentence
with pronouns and entities, identifies which of them are not tied to each
other, and makes suggestions on which to use to avoid biases. Undertaken
experiments show that PronounFlow not only alternates pronouns in sentences
based on the collective human knowledge around us but also considerably helps
coreference resolution systems with the pronoun disambiguation process.
| [
{
"created": "Tue, 29 Aug 2023 11:46:27 GMT",
"version": "v1"
}
] | 2023-08-30 | [
[
"Isaak",
"Nicos",
""
]
] | Flip through any book or listen to any song lyrics, and you will come across pronouns that, in certain cases, can hinder meaning comprehension, especially for machines. As the role of having cognitive machines becomes pervasive in our lives, numerous systems have been developed to resolve pronouns under various challenges. Commensurate with this, it is believed that having systems able to disambiguate pronouns in sentences will help towards the endowment of machines with commonsense and reasoning abilities like those found in humans. However, one problem these systems face with modern English is the lack of gender pronouns, where people try to alternate by using masculine, feminine, or plural to avoid the whole issue. Since humanity aims to the building of systems in the full-bodied sense we usually reserve for people, what happens when pronouns in written text, like plural or epicene ones, refer to unspecified entities whose gender is not necessarily known? Wouldn't that put extra barriers to existing coreference resolution systems? Towards answering those questions, through the implementation of a neural-symbolic system that utilizes the best of both worlds, we are employing PronounFlow, a system that reads any English sentence with pronouns and entities, identifies which of them are not tied to each other, and makes suggestions on which to use to avoid biases. Undertaken experiments show that PronounFlow not only alternates pronouns in sentences based on the collective human knowledge around us but also considerably helps coreference resolution systems with the pronoun disambiguation process. |
0906.1086 | Jean-Marie Vanherpe | Jean-Luc Fouquet (LIFO), Jean-Marie Vanherpe (LIFO) | On Fulkerson conjecture | Accepted for publication in Discussiones Mathematicae Graph Theory;
Discussiones Mathematicae Graph Theory (2010) xxx-yyy | Discussiones Mathematicae Graph Theory 31, 2 (2011) 253-272 | null | null | cs.DM | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | If $G$ is a bridgeless cubic graph, Fulkerson conjectured that we can find 6
perfect matchings (a{\em Fulkerson covering}) with the property that every edge
of $G$ is contained in exactly two of them. A consequence of the Fulkerson
conjecture would be that every bridgeless cubic graph has 3 perfect matchings
with empty intersection (this problem is known as the Fan Raspaud Conjecture).
A {\em FR-triple} is a set of 3 such perfect matchings. We show here how to
derive a Fulkerson covering from two FR-triples. Moreover, we give a simple
proof that the Fulkerson conjecture holds true for some classes of well known
snarks.
| [
{
"created": "Fri, 5 Jun 2009 10:54:55 GMT",
"version": "v1"
},
{
"created": "Mon, 31 May 2010 09:12:31 GMT",
"version": "v2"
}
] | 2011-04-01 | [
[
"Fouquet",
"Jean-Luc",
"",
"LIFO"
],
[
"Vanherpe",
"Jean-Marie",
"",
"LIFO"
]
] | If $G$ is a bridgeless cubic graph, Fulkerson conjectured that we can find 6 perfect matchings (a{\em Fulkerson covering}) with the property that every edge of $G$ is contained in exactly two of them. A consequence of the Fulkerson conjecture would be that every bridgeless cubic graph has 3 perfect matchings with empty intersection (this problem is known as the Fan Raspaud Conjecture). A {\em FR-triple} is a set of 3 such perfect matchings. We show here how to derive a Fulkerson covering from two FR-triples. Moreover, we give a simple proof that the Fulkerson conjecture holds true for some classes of well known snarks. |
2406.05290 | Sergiy Shelyag | Abhiram Anand Thiruthummal, Sergiy Shelyag, Eun-jin Kim | Extremization to Fine Tune Physics Informed Neural Networks for Solving
Boundary Value Problems | Accepted for publication in CNSNS | null | null | null | cs.LG cs.CE cs.NA math.NA physics.comp-ph | http://creativecommons.org/licenses/by-nc-nd/4.0/ | We propose a novel method for fast and accurate training of physics-informed
neural networks (PINNs) to find solutions to boundary value problems (BVPs) and
initial boundary value problems (IBVPs). By combining the methods of training
deep neural networks (DNNs) and Extreme Learning Machines (ELMs), we develop a
model which has the expressivity of DNNs with the fine-tuning ability of ELMs.
We showcase the superiority of our proposed method by solving several BVPs and
IBVPs which include linear and non-linear ordinary differential equations
(ODEs), partial differential equations (PDEs) and coupled PDEs. The examples we
consider include a stiff coupled ODE system where traditional numerical methods
fail, a 3+1D non-linear PDE, Kovasznay flow and Taylor-Green vortex solutions
to incompressible Navier-Stokes equations and pure advection solution of 1+1 D
compressible Euler equation.
The Theory of Functional Connections (TFC) is used to exactly impose initial
and boundary conditions (IBCs) of (I)BVPs on PINNs. We propose a modification
to the TFC framework named Reduced TFC and show a significant improvement in
the training and inference time of PINNs compared to IBCs imposed using TFC.
Furthermore, Reduced TFC is shown to be able to generalize to more complex
boundary geometries which is not possible with TFC. We also introduce a method
of applying boundary conditions at infinity for BVPs and numerically solve the
pure advection in 1+1 D Euler equations using these boundary conditions.
| [
{
"created": "Fri, 7 Jun 2024 23:25:13 GMT",
"version": "v1"
}
] | 2024-06-11 | [
[
"Thiruthummal",
"Abhiram Anand",
""
],
[
"Shelyag",
"Sergiy",
""
],
[
"Kim",
"Eun-jin",
""
]
] | We propose a novel method for fast and accurate training of physics-informed neural networks (PINNs) to find solutions to boundary value problems (BVPs) and initial boundary value problems (IBVPs). By combining the methods of training deep neural networks (DNNs) and Extreme Learning Machines (ELMs), we develop a model which has the expressivity of DNNs with the fine-tuning ability of ELMs. We showcase the superiority of our proposed method by solving several BVPs and IBVPs which include linear and non-linear ordinary differential equations (ODEs), partial differential equations (PDEs) and coupled PDEs. The examples we consider include a stiff coupled ODE system where traditional numerical methods fail, a 3+1D non-linear PDE, Kovasznay flow and Taylor-Green vortex solutions to incompressible Navier-Stokes equations and pure advection solution of 1+1 D compressible Euler equation. The Theory of Functional Connections (TFC) is used to exactly impose initial and boundary conditions (IBCs) of (I)BVPs on PINNs. We propose a modification to the TFC framework named Reduced TFC and show a significant improvement in the training and inference time of PINNs compared to IBCs imposed using TFC. Furthermore, Reduced TFC is shown to be able to generalize to more complex boundary geometries which is not possible with TFC. We also introduce a method of applying boundary conditions at infinity for BVPs and numerically solve the pure advection in 1+1 D Euler equations using these boundary conditions. |
1901.08274 | Zaiqiang Wu | Zaiqiang Wu, Wei Jiang, Hao Luo, Lin Cheng | A Novel Self-Intersection Penalty Term for Statistical Body Shape Models
and Its Applications in 3D Pose Estimation | null | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Statistical body shape models are widely used in 3D pose estimation due to
their low-dimensional parameters representation. However, it is difficult to
avoid self-intersection between body parts accurately. Motivated by this fact,
we proposed a novel self-intersection penalty term for statistical body shape
models applied in 3D pose estimation. To avoid the trouble of computing
self-intersection for complex surfaces like the body meshes, the gradient of
our proposed self-intersection penalty term is manually derived from the
perspective of geometry. First, the self-intersection penalty term is defined
as the volume of the self-intersection region. To calculate the partial
derivatives with respect to the coordinates of the vertices, we employed
detection rays to divide vertices of statistical body shape models into
different groups depending on whether the vertex is in the region of
self-intersection. Second, the partial derivatives could be easily derived by
the normal vectors of neighboring triangles of the vertices. Finally, this
penalty term could be applied in gradient-based optimization algorithms to
remove the self-intersection of triangular meshes without using any
approximation. Qualitative and quantitative evaluations were conducted to
demonstrate the effectiveness and generality of our proposed method compared
with previous approaches. The experimental results show that our proposed
penalty term can avoid self-intersection to exclude unreasonable predictions
and improves the accuracy of 3D pose estimation indirectly. Further more, the
proposed method could be employed universally in triangular mesh based 3D
reconstruction.
| [
{
"created": "Thu, 24 Jan 2019 08:19:37 GMT",
"version": "v1"
}
] | 2019-01-25 | [
[
"Wu",
"Zaiqiang",
""
],
[
"Jiang",
"Wei",
""
],
[
"Luo",
"Hao",
""
],
[
"Cheng",
"Lin",
""
]
] | Statistical body shape models are widely used in 3D pose estimation due to their low-dimensional parameters representation. However, it is difficult to avoid self-intersection between body parts accurately. Motivated by this fact, we proposed a novel self-intersection penalty term for statistical body shape models applied in 3D pose estimation. To avoid the trouble of computing self-intersection for complex surfaces like the body meshes, the gradient of our proposed self-intersection penalty term is manually derived from the perspective of geometry. First, the self-intersection penalty term is defined as the volume of the self-intersection region. To calculate the partial derivatives with respect to the coordinates of the vertices, we employed detection rays to divide vertices of statistical body shape models into different groups depending on whether the vertex is in the region of self-intersection. Second, the partial derivatives could be easily derived by the normal vectors of neighboring triangles of the vertices. Finally, this penalty term could be applied in gradient-based optimization algorithms to remove the self-intersection of triangular meshes without using any approximation. Qualitative and quantitative evaluations were conducted to demonstrate the effectiveness and generality of our proposed method compared with previous approaches. The experimental results show that our proposed penalty term can avoid self-intersection to exclude unreasonable predictions and improves the accuracy of 3D pose estimation indirectly. Further more, the proposed method could be employed universally in triangular mesh based 3D reconstruction. |
2304.09406 | Minh Hua | Minh Hua, Rita Raley | How to Do Things with Deep Learning Code | Accepted for publication in Digital Humanities Quarterly | null | null | null | cs.CL | http://creativecommons.org/licenses/by/4.0/ | The premise of this article is that a basic understanding of the composition
and functioning of large language models is critically urgent. To that end, we
extract a representational map of OpenAI's GPT-2 with what we articulate as two
classes of deep learning code, that which pertains to the model and that which
underwrites applications built around the model. We then verify this map
through case studies of two popular GPT-2 applications: the text adventure
game, AI Dungeon, and the language art project, This Word Does Not Exist. Such
an exercise allows us to test the potential of Critical Code Studies when the
object of study is deep learning code and to demonstrate the validity of code
as an analytical focus for researchers in the subfields of Critical Artificial
Intelligence and Critical Machine Learning Studies. More broadly, however, our
work draws attention to the means by which ordinary users might interact with,
and even direct, the behavior of deep learning systems, and by extension works
toward demystifying some of the auratic mystery of "AI." What is at stake is
the possibility of achieving an informed sociotechnical consensus about the
responsible applications of large language models, as well as a more expansive
sense of their creative capabilities-indeed, understanding how and where
engagement occurs allows all of us to become more active participants in the
development of machine learning systems.
| [
{
"created": "Wed, 19 Apr 2023 03:46:12 GMT",
"version": "v1"
}
] | 2023-04-20 | [
[
"Hua",
"Minh",
""
],
[
"Raley",
"Rita",
""
]
] | The premise of this article is that a basic understanding of the composition and functioning of large language models is critically urgent. To that end, we extract a representational map of OpenAI's GPT-2 with what we articulate as two classes of deep learning code, that which pertains to the model and that which underwrites applications built around the model. We then verify this map through case studies of two popular GPT-2 applications: the text adventure game, AI Dungeon, and the language art project, This Word Does Not Exist. Such an exercise allows us to test the potential of Critical Code Studies when the object of study is deep learning code and to demonstrate the validity of code as an analytical focus for researchers in the subfields of Critical Artificial Intelligence and Critical Machine Learning Studies. More broadly, however, our work draws attention to the means by which ordinary users might interact with, and even direct, the behavior of deep learning systems, and by extension works toward demystifying some of the auratic mystery of "AI." What is at stake is the possibility of achieving an informed sociotechnical consensus about the responsible applications of large language models, as well as a more expansive sense of their creative capabilities-indeed, understanding how and where engagement occurs allows all of us to become more active participants in the development of machine learning systems. |
2204.06127 | Mingshuo Nie | Mingshuo Nie, Dongming Chen, Dongqi Wang | Reinforcement learning on graphs: A survey | Accepted by IEEE Transactions on Emerging Topics in Computational
Intelligence | null | null | null | cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Graph mining tasks arise from many different application domains, ranging
from social networks, transportation to E-commerce, etc., which have been
receiving great attention from the theoretical and algorithmic design
communities in recent years, and there has been some pioneering work employing
the research-rich Reinforcement Learning (RL) techniques to address graph data
mining tasks. However, these graph mining methods and RL models are dispersed
in different research areas, which makes it hard to compare them. In this
survey, we provide a comprehensive overview of RL and graph mining methods and
generalize these methods to Graph Reinforcement Learning (GRL) as a unified
formulation. We further discuss the applications of GRL methods across various
domains and summarize the method descriptions, open-source codes, and benchmark
datasets of GRL methods. Furthermore, we propose important directions and
challenges to be solved in the future. As far as we know, this is the latest
work on a comprehensive survey of GRL, this work provides a global view and a
learning resource for scholars. In addition, we create an online open-source
for both interested scholars who want to enter this rapidly developing domain
and experts who would like to compare GRL methods.
| [
{
"created": "Wed, 13 Apr 2022 01:25:58 GMT",
"version": "v1"
},
{
"created": "Wed, 27 Apr 2022 12:50:20 GMT",
"version": "v2"
},
{
"created": "Fri, 11 Nov 2022 11:19:28 GMT",
"version": "v3"
},
{
"created": "Sun, 15 Jan 2023 01:44:48 GMT",
"version": "v4"
}
] | 2023-01-18 | [
[
"Nie",
"Mingshuo",
""
],
[
"Chen",
"Dongming",
""
],
[
"Wang",
"Dongqi",
""
]
] | Graph mining tasks arise from many different application domains, ranging from social networks, transportation to E-commerce, etc., which have been receiving great attention from the theoretical and algorithmic design communities in recent years, and there has been some pioneering work employing the research-rich Reinforcement Learning (RL) techniques to address graph data mining tasks. However, these graph mining methods and RL models are dispersed in different research areas, which makes it hard to compare them. In this survey, we provide a comprehensive overview of RL and graph mining methods and generalize these methods to Graph Reinforcement Learning (GRL) as a unified formulation. We further discuss the applications of GRL methods across various domains and summarize the method descriptions, open-source codes, and benchmark datasets of GRL methods. Furthermore, we propose important directions and challenges to be solved in the future. As far as we know, this is the latest work on a comprehensive survey of GRL, this work provides a global view and a learning resource for scholars. In addition, we create an online open-source for both interested scholars who want to enter this rapidly developing domain and experts who would like to compare GRL methods. |
2301.01808 | Adar Kahana | Adar Kahana and Oren Elisha | MessageNet: Message Classification using Natural Language Processing and
Meta-data | null | null | null | null | cs.LG cs.CL | http://creativecommons.org/licenses/by/4.0/ | In this paper we propose a new Deep Learning (DL) approach for message
classification. Our method is based on the state-of-the-art Natural Language
Processing (NLP) building blocks, combined with a novel technique for infusing
the meta-data input that is typically available in messages such as the sender
information, timestamps, attached image, audio, affiliations, and more. As we
demonstrate throughout the paper, going beyond the mere text by leveraging all
available channels in the message, could yield an improved representation and
higher classification accuracy. To achieve message representation, each type of
input is processed in a dedicated block in the neural network architecture that
is suitable for the data type. Such an implementation enables training all
blocks together simultaneously, and forming cross channels features in the
network. We show in the Experiments Section that in some cases, message's
meta-data holds an additional information that cannot be extracted just from
the text, and when using this information we achieve better performance.
Furthermore, we demonstrate that our multi-modality block approach outperforms
other approaches for injecting the meta data to the the text classifier.
| [
{
"created": "Wed, 4 Jan 2023 20:11:00 GMT",
"version": "v1"
}
] | 2023-01-06 | [
[
"Kahana",
"Adar",
""
],
[
"Elisha",
"Oren",
""
]
] | In this paper we propose a new Deep Learning (DL) approach for message classification. Our method is based on the state-of-the-art Natural Language Processing (NLP) building blocks, combined with a novel technique for infusing the meta-data input that is typically available in messages such as the sender information, timestamps, attached image, audio, affiliations, and more. As we demonstrate throughout the paper, going beyond the mere text by leveraging all available channels in the message, could yield an improved representation and higher classification accuracy. To achieve message representation, each type of input is processed in a dedicated block in the neural network architecture that is suitable for the data type. Such an implementation enables training all blocks together simultaneously, and forming cross channels features in the network. We show in the Experiments Section that in some cases, message's meta-data holds an additional information that cannot be extracted just from the text, and when using this information we achieve better performance. Furthermore, we demonstrate that our multi-modality block approach outperforms other approaches for injecting the meta data to the the text classifier. |
1611.04021 | Tseng-Hung Chen | Kuo-Hao Zeng, Tseng-Hung Chen, Ching-Yao Chuang, Yuan-Hong Liao, Juan
Carlos Niebles, Min Sun | Leveraging Video Descriptions to Learn Video Question Answering | 7 pages, 5 figures. Accepted to AAAI 2017. Camera-ready version | null | null | null | cs.CV cs.AI cs.MM | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We propose a scalable approach to learn video-based question answering (QA):
answer a "free-form natural language question" about a video content. Our
approach automatically harvests a large number of videos and descriptions
freely available online. Then, a large number of candidate QA pairs are
automatically generated from descriptions rather than manually annotated. Next,
we use these candidate QA pairs to train a number of video-based QA methods
extended fromMN (Sukhbaatar et al. 2015), VQA (Antol et al. 2015), SA (Yao et
al. 2015), SS (Venugopalan et al. 2015). In order to handle non-perfect
candidate QA pairs, we propose a self-paced learning procedure to iteratively
identify them and mitigate their effects in training. Finally, we evaluate
performance on manually generated video-based QA pairs. The results show that
our self-paced learning procedure is effective, and the extended SS model
outperforms various baselines.
| [
{
"created": "Sat, 12 Nov 2016 17:15:57 GMT",
"version": "v1"
},
{
"created": "Mon, 19 Dec 2016 16:07:33 GMT",
"version": "v2"
}
] | 2016-12-20 | [
[
"Zeng",
"Kuo-Hao",
""
],
[
"Chen",
"Tseng-Hung",
""
],
[
"Chuang",
"Ching-Yao",
""
],
[
"Liao",
"Yuan-Hong",
""
],
[
"Niebles",
"Juan Carlos",
""
],
[
"Sun",
"Min",
""
]
] | We propose a scalable approach to learn video-based question answering (QA): answer a "free-form natural language question" about a video content. Our approach automatically harvests a large number of videos and descriptions freely available online. Then, a large number of candidate QA pairs are automatically generated from descriptions rather than manually annotated. Next, we use these candidate QA pairs to train a number of video-based QA methods extended fromMN (Sukhbaatar et al. 2015), VQA (Antol et al. 2015), SA (Yao et al. 2015), SS (Venugopalan et al. 2015). In order to handle non-perfect candidate QA pairs, we propose a self-paced learning procedure to iteratively identify them and mitigate their effects in training. Finally, we evaluate performance on manually generated video-based QA pairs. The results show that our self-paced learning procedure is effective, and the extended SS model outperforms various baselines. |
2009.06184 | Yifan Wang | Yifan Wang, Guoli Yan, Haikuan Zhu, Sagar Buch, Ying Wang, Ewart Mark
Haacke, Jing Hua, and Zichun Zhong | VC-Net: Deep Volume-Composition Networks for Segmentation and
Visualization of Highly Sparse and Noisy Image Data | 15 pages, 10 figures, proceeding to IEEE Transactions on
Visualization and Computer Graphics (TVCG) (IEEE SciVis 2020), October, 2020 | null | null | null | cs.GR cs.CV eess.IV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The motivation of our work is to present a new visualization-guided computing
paradigm to combine direct 3D volume processing and volume rendered clues for
effective 3D exploration such as extracting and visualizing microstructures
in-vivo. However, it is still challenging to extract and visualize high
fidelity 3D vessel structure due to its high sparseness, noisiness, and complex
topology variations. In this paper, we present an end-to-end deep learning
method, VC-Net, for robust extraction of 3D microvasculature through embedding
the image composition, generated by maximum intensity projection (MIP), into 3D
volume image learning to enhance the performance. The core novelty is to
automatically leverage the volume visualization technique (MIP) to enhance the
3D data exploration at deep learning level. The MIP embedding features can
enhance the local vessel signal and are adaptive to the geometric variability
and scalability of vessels, which is crucial in microvascular tracking. A
multi-stream convolutional neural network is proposed to learn the 3D volume
and 2D MIP features respectively and then explore their inter-dependencies in a
joint volume-composition embedding space by unprojecting the MIP features into
3D volume embedding space. The proposed framework can better capture small /
micro vessels and improve vessel connectivity. To our knowledge, this is the
first deep learning framework to construct a joint convolutional embedding
space, where the computed vessel probabilities from volume rendering based 2D
projection and 3D volume can be explored and integrated synergistically.
Experimental results are compared with the traditional 3D vessel segmentation
methods and the deep learning state-of-the-art on public and real patient
(micro-)cerebrovascular image datasets. Our method demonstrates the potential
in a powerful MR arteriogram and venogram diagnosis of vascular diseases.
| [
{
"created": "Mon, 14 Sep 2020 04:15:02 GMT",
"version": "v1"
}
] | 2020-09-15 | [
[
"Wang",
"Yifan",
""
],
[
"Yan",
"Guoli",
""
],
[
"Zhu",
"Haikuan",
""
],
[
"Buch",
"Sagar",
""
],
[
"Wang",
"Ying",
""
],
[
"Haacke",
"Ewart Mark",
""
],
[
"Hua",
"Jing",
""
],
[
"Zhong",
"Zichun",
""
]
] | The motivation of our work is to present a new visualization-guided computing paradigm to combine direct 3D volume processing and volume rendered clues for effective 3D exploration such as extracting and visualizing microstructures in-vivo. However, it is still challenging to extract and visualize high fidelity 3D vessel structure due to its high sparseness, noisiness, and complex topology variations. In this paper, we present an end-to-end deep learning method, VC-Net, for robust extraction of 3D microvasculature through embedding the image composition, generated by maximum intensity projection (MIP), into 3D volume image learning to enhance the performance. The core novelty is to automatically leverage the volume visualization technique (MIP) to enhance the 3D data exploration at deep learning level. The MIP embedding features can enhance the local vessel signal and are adaptive to the geometric variability and scalability of vessels, which is crucial in microvascular tracking. A multi-stream convolutional neural network is proposed to learn the 3D volume and 2D MIP features respectively and then explore their inter-dependencies in a joint volume-composition embedding space by unprojecting the MIP features into 3D volume embedding space. The proposed framework can better capture small / micro vessels and improve vessel connectivity. To our knowledge, this is the first deep learning framework to construct a joint convolutional embedding space, where the computed vessel probabilities from volume rendering based 2D projection and 3D volume can be explored and integrated synergistically. Experimental results are compared with the traditional 3D vessel segmentation methods and the deep learning state-of-the-art on public and real patient (micro-)cerebrovascular image datasets. Our method demonstrates the potential in a powerful MR arteriogram and venogram diagnosis of vascular diseases. |
1808.00435 | Sheng Chen | Sheng Chen, Jia Guo, Yang Liu, Xiang Gao, Zhen Han | Global Norm-Aware Pooling for Pose-Robust Face Recognition at Low False
Positive Rate | null | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In this paper, we propose a novel Global Norm-Aware Pooling (GNAP) block,
which reweights local features in a convolutional neural network (CNN)
adaptively according to their L2 norms and outputs a global feature vector with
a global average pooling layer. Our GNAP block is designed to give dynamic
weights to local features in different spatial positions without losing spatial
symmetry. We use a GNAP block in a face feature embedding CNN to produce
discriminative face feature vectors for pose-robust face recognition. The GNAP
block is of very cheap computational cost, but it is very powerful for
frontal-profile face recognition. Under the CFP frontal-profile protocol, the
GNAP block can not only reduce EER dramatically but also boost TPR@FPR=0.1%
(TPR i.e. True Positive Rate, FPR i.e. False Positive Rate) substantially. Our
experiments show that the GNAP block greatly promotes pose-robust face
recognition over the base model especially at low false positive rate.
| [
{
"created": "Wed, 1 Aug 2018 17:32:31 GMT",
"version": "v1"
}
] | 2018-08-02 | [
[
"Chen",
"Sheng",
""
],
[
"Guo",
"Jia",
""
],
[
"Liu",
"Yang",
""
],
[
"Gao",
"Xiang",
""
],
[
"Han",
"Zhen",
""
]
] | In this paper, we propose a novel Global Norm-Aware Pooling (GNAP) block, which reweights local features in a convolutional neural network (CNN) adaptively according to their L2 norms and outputs a global feature vector with a global average pooling layer. Our GNAP block is designed to give dynamic weights to local features in different spatial positions without losing spatial symmetry. We use a GNAP block in a face feature embedding CNN to produce discriminative face feature vectors for pose-robust face recognition. The GNAP block is of very cheap computational cost, but it is very powerful for frontal-profile face recognition. Under the CFP frontal-profile protocol, the GNAP block can not only reduce EER dramatically but also boost TPR@FPR=0.1% (TPR i.e. True Positive Rate, FPR i.e. False Positive Rate) substantially. Our experiments show that the GNAP block greatly promotes pose-robust face recognition over the base model especially at low false positive rate. |
1902.09968 | Runsheng Zhang | Runsheng Zhang, Yaping Huang, Mengyang Pu, Jian Zhang, Qingji Guan, Qi
Zou, Haibin Ling | Object Discovery From a Single Unlabeled Image by Mining Frequent
Itemset With Multi-scale Features | null | null | 10.1109/TIP.2020.3015543 | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | TThe goal of our work is to discover dominant objects in a very general
setting where only a single unlabeled image is given. This is far more
challenge than typical co-localization or weakly-supervised localization tasks.
To tackle this problem, we propose a simple but effective pattern mining-based
method, called Object Location Mining (OLM), which exploits the advantages of
data mining and feature representation of pre-trained convolutional neural
networks (CNNs). Specifically, we first convert the feature maps from a
pre-trained CNN model into a set of transactions, and then discovers frequent
patterns from transaction database through pattern mining techniques. We
observe that those discovered patterns, i.e., co-occurrence highlighted
regions, typically hold appearance and spatial consistency. Motivated by this
observation, we can easily discover and localize possible objects by merging
relevant meaningful patterns. Extensive experiments on a variety of benchmarks
demonstrate that OLM achieves competitive localization performance compared
with the state-of-the-art methods. We also evaluate our approach compared with
unsupervised saliency detection methods and achieves competitive results on
seven benchmark datasets. Moreover, we conduct experiments on fine-grained
classification to show that our proposed method can locate the entire object
and parts accurately, which can benefit to improving the classification results
significantly.
| [
{
"created": "Tue, 26 Feb 2019 14:37:01 GMT",
"version": "v1"
},
{
"created": "Tue, 24 Sep 2019 11:19:00 GMT",
"version": "v2"
},
{
"created": "Sat, 8 Aug 2020 05:05:12 GMT",
"version": "v3"
}
] | 2023-07-19 | [
[
"Zhang",
"Runsheng",
""
],
[
"Huang",
"Yaping",
""
],
[
"Pu",
"Mengyang",
""
],
[
"Zhang",
"Jian",
""
],
[
"Guan",
"Qingji",
""
],
[
"Zou",
"Qi",
""
],
[
"Ling",
"Haibin",
""
]
] | TThe goal of our work is to discover dominant objects in a very general setting where only a single unlabeled image is given. This is far more challenge than typical co-localization or weakly-supervised localization tasks. To tackle this problem, we propose a simple but effective pattern mining-based method, called Object Location Mining (OLM), which exploits the advantages of data mining and feature representation of pre-trained convolutional neural networks (CNNs). Specifically, we first convert the feature maps from a pre-trained CNN model into a set of transactions, and then discovers frequent patterns from transaction database through pattern mining techniques. We observe that those discovered patterns, i.e., co-occurrence highlighted regions, typically hold appearance and spatial consistency. Motivated by this observation, we can easily discover and localize possible objects by merging relevant meaningful patterns. Extensive experiments on a variety of benchmarks demonstrate that OLM achieves competitive localization performance compared with the state-of-the-art methods. We also evaluate our approach compared with unsupervised saliency detection methods and achieves competitive results on seven benchmark datasets. Moreover, we conduct experiments on fine-grained classification to show that our proposed method can locate the entire object and parts accurately, which can benefit to improving the classification results significantly. |
2007.00413 | Yamin Li | Yamin Li, Kai Tang, Dong He, Xiangyu Wang | Multi-Axis Support-Free Printing of Freeform Parts with Lattice Infill
Structures | arXiv admin note: text overlap with arXiv:2003.05938 | null | null | null | cs.GR cs.CG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In additive manufacturing, infill structures are commonly used to reduce the
weight and cost of a solid part. Currently, most infill structure generation
methods are based on the conventional 2.5-axis printing configuration, which,
although able to satisfy the self-supporting condition on the infills, suffer
from the well-known stair-case effect on the finished surface and the need of
extensive support for overhang features. In this paper, based on the emerging
continuous multi-axis printing configuration, we present a new lattice infill
structure generation algorithm, which is able to achieve both the
self-supporting condition for the infills and the support-free requirement at
the boundary surface of the part. The algorithm critically relies on the use of
three mutually orthogonal geodesic distance fields that are embedded in the
tetrahedral mesh of the solid model. The intersection between the iso-geodesic
distance surfaces of these three geodesic distance fields naturally forms the
desired lattice of infill structure, while the density of the infills can be
conveniently controlled by adjusting the iso-values. The lattice infill pattern
in each curved slicing layer is trimmed to conform to an Eulerian graph so to
generate a continuous printing path, which can effectively reduce the nozzle
retractions during the printing process. In addition, to cater to the
collision-free requirement and to improve the printing efficiency, we also
propose a printing sequence optimization algorithm for determining a
collision-free order of printing of the connected lattice infills, which seeks
to reduce the air-move length of the nozzle. Ample experiments in both computer
simulation and physical printing are performed, and the results give a
preliminary confirmation of the advantages of our methodology.
| [
{
"created": "Tue, 30 Jun 2020 06:08:00 GMT",
"version": "v1"
}
] | 2020-07-02 | [
[
"Li",
"Yamin",
""
],
[
"Tang",
"Kai",
""
],
[
"He",
"Dong",
""
],
[
"Wang",
"Xiangyu",
""
]
] | In additive manufacturing, infill structures are commonly used to reduce the weight and cost of a solid part. Currently, most infill structure generation methods are based on the conventional 2.5-axis printing configuration, which, although able to satisfy the self-supporting condition on the infills, suffer from the well-known stair-case effect on the finished surface and the need of extensive support for overhang features. In this paper, based on the emerging continuous multi-axis printing configuration, we present a new lattice infill structure generation algorithm, which is able to achieve both the self-supporting condition for the infills and the support-free requirement at the boundary surface of the part. The algorithm critically relies on the use of three mutually orthogonal geodesic distance fields that are embedded in the tetrahedral mesh of the solid model. The intersection between the iso-geodesic distance surfaces of these three geodesic distance fields naturally forms the desired lattice of infill structure, while the density of the infills can be conveniently controlled by adjusting the iso-values. The lattice infill pattern in each curved slicing layer is trimmed to conform to an Eulerian graph so to generate a continuous printing path, which can effectively reduce the nozzle retractions during the printing process. In addition, to cater to the collision-free requirement and to improve the printing efficiency, we also propose a printing sequence optimization algorithm for determining a collision-free order of printing of the connected lattice infills, which seeks to reduce the air-move length of the nozzle. Ample experiments in both computer simulation and physical printing are performed, and the results give a preliminary confirmation of the advantages of our methodology. |
2005.10039 | Hinrikus Wolf | Tobias Schumacher, Hinrikus Wolf, Martin Ritzert, Florian Lemmerich,
Jan Bachmann, Florian Frantzen, Max Klabunde, Martin Grohe, Markus Strohmaier | The Effects of Randomness on the Stability of Node Embeddings | null | null | null | null | cs.LG cs.SI stat.ML | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We systematically evaluate the (in-)stability of state-of-the-art node
embedding algorithms due to randomness, i.e., the random variation of their
outcomes given identical algorithms and graphs. We apply five node embeddings
algorithms---HOPE, LINE, node2vec, SDNE, and GraphSAGE---to synthetic and
empirical graphs and assess their stability under randomness with respect to
(i) the geometry of embedding spaces as well as (ii) their performance in
downstream tasks. We find significant instabilities in the geometry of
embedding spaces independent of the centrality of a node. In the evaluation of
downstream tasks, we find that the accuracy of node classification seems to be
unaffected by random seeding while the actual classification of nodes can vary
significantly. This suggests that instability effects need to be taken into
account when working with node embeddings. Our work is relevant for researchers
and engineers interested in the effectiveness, reliability, and reproducibility
of node embedding approaches.
| [
{
"created": "Wed, 20 May 2020 13:36:09 GMT",
"version": "v1"
}
] | 2020-05-21 | [
[
"Schumacher",
"Tobias",
""
],
[
"Wolf",
"Hinrikus",
""
],
[
"Ritzert",
"Martin",
""
],
[
"Lemmerich",
"Florian",
""
],
[
"Bachmann",
"Jan",
""
],
[
"Frantzen",
"Florian",
""
],
[
"Klabunde",
"Max",
""
],
[
"Grohe",
"Martin",
""
],
[
"Strohmaier",
"Markus",
""
]
] | We systematically evaluate the (in-)stability of state-of-the-art node embedding algorithms due to randomness, i.e., the random variation of their outcomes given identical algorithms and graphs. We apply five node embeddings algorithms---HOPE, LINE, node2vec, SDNE, and GraphSAGE---to synthetic and empirical graphs and assess their stability under randomness with respect to (i) the geometry of embedding spaces as well as (ii) their performance in downstream tasks. We find significant instabilities in the geometry of embedding spaces independent of the centrality of a node. In the evaluation of downstream tasks, we find that the accuracy of node classification seems to be unaffected by random seeding while the actual classification of nodes can vary significantly. This suggests that instability effects need to be taken into account when working with node embeddings. Our work is relevant for researchers and engineers interested in the effectiveness, reliability, and reproducibility of node embedding approaches. |
2105.11879 | Marcin Namysl | Marcin Namysl, Alexander M. Esser, Sven Behnke, Joachim K\"ohler | Flexible Table Recognition and Semantic Interpretation System | Accepted for publication in Proceedings of the 17th International
Conference on Computer Vision Theory and Applications (VISAPP 2022) | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Table extraction is an important but still unsolved problem. In this paper,
we introduce a flexible and modular table extraction system. We develop two
rule-based algorithms that perform the complete table recognition process,
including table detection and segmentation, and support the most frequent table
formats. Moreover, to incorporate the extraction of semantic information, we
develop a graph-based table interpretation method. We conduct extensive
experiments on the challenging table recognition benchmarks ICDAR 2013 and
ICDAR 2019, achieving results competitive with state-of-the-art approaches. Our
complete information extraction system exhibited a high F1 score of 0.7380. To
support future research on information extraction from documents, we make the
resources (ground-truth annotations, evaluation scripts, algorithm parameters)
from our table interpretation experiment publicly available.
| [
{
"created": "Tue, 25 May 2021 12:31:02 GMT",
"version": "v1"
},
{
"created": "Thu, 2 Dec 2021 17:33:35 GMT",
"version": "v2"
}
] | 2021-12-03 | [
[
"Namysl",
"Marcin",
""
],
[
"Esser",
"Alexander M.",
""
],
[
"Behnke",
"Sven",
""
],
[
"Köhler",
"Joachim",
""
]
] | Table extraction is an important but still unsolved problem. In this paper, we introduce a flexible and modular table extraction system. We develop two rule-based algorithms that perform the complete table recognition process, including table detection and segmentation, and support the most frequent table formats. Moreover, to incorporate the extraction of semantic information, we develop a graph-based table interpretation method. We conduct extensive experiments on the challenging table recognition benchmarks ICDAR 2013 and ICDAR 2019, achieving results competitive with state-of-the-art approaches. Our complete information extraction system exhibited a high F1 score of 0.7380. To support future research on information extraction from documents, we make the resources (ground-truth annotations, evaluation scripts, algorithm parameters) from our table interpretation experiment publicly available. |
2008.08791 | Monica Perusquia-Hernandez | Monica Perusquia-Hernandez, Felix Dollack, Chun Kwang Tan, Shushi
Namba, Saho Ayabe-Kanamura, Kenji Suzuki | Facial movement synergies and Action Unit detection from distal wearable
Electromyography and Computer Vision | 11 pages, 11 figures, 2 tables | null | null | null | cs.HC cs.CV eess.IV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Distal facial Electromyography (EMG) can be used to detect smiles and frowns
with reasonable accuracy. It capitalizes on volume conduction to detect
relevant muscle activity, even when the electrodes are not placed directly on
the source muscle. The main advantage of this method is to prevent occlusion
and obstruction of the facial expression production, whilst allowing EMG
measurements. However, measuring EMG distally entails that the exact source of
the facial movement is unknown. We propose a novel method to estimate specific
Facial Action Units (AUs) from distal facial EMG and Computer Vision (CV). This
method is based on Independent Component Analysis (ICA), Non-Negative Matrix
Factorization (NNMF), and sorting of the resulting components to determine
which is the most likely to correspond to each CV-labeled action unit (AU).
Performance on the detection of AU06 (Orbicularis Oculi) and AU12 (Zygomaticus
Major) was estimated by calculating the agreement with Human Coders. The
results of our proposed algorithm showed an accuracy of 81% and a Cohen's Kappa
of 0.49 for AU6; and accuracy of 82% and a Cohen's Kappa of 0.53 for AU12. This
demonstrates the potential of distal EMG to detect individual facial movements.
Using this multimodal method, several AU synergies were identified. We
quantified the co-occurrence and timing of AU6 and AU12 in posed and
spontaneous smiles using the human-coded labels, and for comparison, using the
continuous CV-labels. The co-occurrence analysis was also performed on the
EMG-based labels to uncover the relationship between muscle synergies and the
kinematics of visible facial movement.
| [
{
"created": "Thu, 20 Aug 2020 06:09:03 GMT",
"version": "v1"
}
] | 2020-08-21 | [
[
"Perusquia-Hernandez",
"Monica",
""
],
[
"Dollack",
"Felix",
""
],
[
"Tan",
"Chun Kwang",
""
],
[
"Namba",
"Shushi",
""
],
[
"Ayabe-Kanamura",
"Saho",
""
],
[
"Suzuki",
"Kenji",
""
]
] | Distal facial Electromyography (EMG) can be used to detect smiles and frowns with reasonable accuracy. It capitalizes on volume conduction to detect relevant muscle activity, even when the electrodes are not placed directly on the source muscle. The main advantage of this method is to prevent occlusion and obstruction of the facial expression production, whilst allowing EMG measurements. However, measuring EMG distally entails that the exact source of the facial movement is unknown. We propose a novel method to estimate specific Facial Action Units (AUs) from distal facial EMG and Computer Vision (CV). This method is based on Independent Component Analysis (ICA), Non-Negative Matrix Factorization (NNMF), and sorting of the resulting components to determine which is the most likely to correspond to each CV-labeled action unit (AU). Performance on the detection of AU06 (Orbicularis Oculi) and AU12 (Zygomaticus Major) was estimated by calculating the agreement with Human Coders. The results of our proposed algorithm showed an accuracy of 81% and a Cohen's Kappa of 0.49 for AU6; and accuracy of 82% and a Cohen's Kappa of 0.53 for AU12. This demonstrates the potential of distal EMG to detect individual facial movements. Using this multimodal method, several AU synergies were identified. We quantified the co-occurrence and timing of AU6 and AU12 in posed and spontaneous smiles using the human-coded labels, and for comparison, using the continuous CV-labels. The co-occurrence analysis was also performed on the EMG-based labels to uncover the relationship between muscle synergies and the kinematics of visible facial movement. |
2208.05343 | Pino Caballero-Gil | Pino Caballero-Gil, Francisco Mart\'in-Fern\'andez, C\'andido
Caballero-Gil | Using query frequencies in tree-based revocation for certificateless
authentication in VANETs | null | The 9th International Conference for Internet Technology and
Secured Transactions (ICITST-2014), pp. 268-273, 2014 | 10.1109/ICITST.2014.7038819 | null | cs.CR | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Revocation of dishonest users is not an easy problem. This paper proposes a
new way to manage revocation of pseudonyms in vehicular ad-hoc networks when
using identity-based authentication to increase efficiency and security through
certificateless authentication. In order to improve the performance of
revocation lists, this paper proposes the use of a data structure based on
authenticated dynamic hash k-ary trees and the frequency with which revoked
pseudonyms are consulted. The use of the knowledge about the frequency of
consultation of revoked pseudonyms allows an easier access to the most popular
revoked pseudonyms to the detriment of revoked pseudonyms that are the least
consulted. Accordingly, the proposal is especially useful in urban environments
where there are vehicles that spend more time on road than others, such as
public service vehicles.
| [
{
"created": "Sat, 6 Aug 2022 18:45:34 GMT",
"version": "v1"
}
] | 2022-08-11 | [
[
"Caballero-Gil",
"Pino",
""
],
[
"Martín-Fernández",
"Francisco",
""
],
[
"Caballero-Gil",
"Cándido",
""
]
] | Revocation of dishonest users is not an easy problem. This paper proposes a new way to manage revocation of pseudonyms in vehicular ad-hoc networks when using identity-based authentication to increase efficiency and security through certificateless authentication. In order to improve the performance of revocation lists, this paper proposes the use of a data structure based on authenticated dynamic hash k-ary trees and the frequency with which revoked pseudonyms are consulted. The use of the knowledge about the frequency of consultation of revoked pseudonyms allows an easier access to the most popular revoked pseudonyms to the detriment of revoked pseudonyms that are the least consulted. Accordingly, the proposal is especially useful in urban environments where there are vehicles that spend more time on road than others, such as public service vehicles. |
1902.06422 | Hirofumi Tsuda | Hirofumi Tsuda | Optimal Sequence and Performance for Desired User in Asynchronous CDMA
System | null | null | null | null | cs.IT math.IT | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We consider asynchronous CDMA systems in no-fading environments with a
particular focus on a certain user. This certain user is called a desired user
in this paper. In such a situation, an optimal sequence, maximum
Signal-to-Interference plus Noise Ratio (SINR) and the maximum capacity for a
desired user are derived with other spreading sequences being given and fixed.
In addition, the maximum SINR and the optimal sequence for a desired user are
written in terms of the minimum eigenvalue and the corresponding eigenvector of
a matrix, respectively. Since it is not straightforward to obtain an explicit
form of the maximum SINR, we evaluate SINR and obtain the lower and upper
bounds of the maximum SINR. From these bounds, the maximum SINR may get larger
as the quantities written in terms of quadratic forms of other spreading
sequences decrease. Further, we propose a method to obtain spreading sequences
for all the users which achieve large SINRs. The performance of our proposed
method is numerically verified.
| [
{
"created": "Mon, 18 Feb 2019 06:48:50 GMT",
"version": "v1"
},
{
"created": "Sun, 24 Mar 2019 06:01:47 GMT",
"version": "v2"
},
{
"created": "Wed, 24 Apr 2019 13:10:41 GMT",
"version": "v3"
},
{
"created": "Thu, 6 Jun 2019 13:57:17 GMT",
"version": "v4"
},
{
"created": "Sun, 9 Jun 2019 09:19:26 GMT",
"version": "v5"
}
] | 2019-06-11 | [
[
"Tsuda",
"Hirofumi",
""
]
] | We consider asynchronous CDMA systems in no-fading environments with a particular focus on a certain user. This certain user is called a desired user in this paper. In such a situation, an optimal sequence, maximum Signal-to-Interference plus Noise Ratio (SINR) and the maximum capacity for a desired user are derived with other spreading sequences being given and fixed. In addition, the maximum SINR and the optimal sequence for a desired user are written in terms of the minimum eigenvalue and the corresponding eigenvector of a matrix, respectively. Since it is not straightforward to obtain an explicit form of the maximum SINR, we evaluate SINR and obtain the lower and upper bounds of the maximum SINR. From these bounds, the maximum SINR may get larger as the quantities written in terms of quadratic forms of other spreading sequences decrease. Further, we propose a method to obtain spreading sequences for all the users which achieve large SINRs. The performance of our proposed method is numerically verified. |
1604.02694 | Hao Fu | Hao Fu, Xing Xie, Yong Rui, Defu Lian, Guangzhong Sun, Enhong Chen | Predicting Social Status via Social Networks: A Case Study on
University, Occupation, and Region | null | null | null | null | cs.SI physics.soc-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Social status refers to the relative position within the society. It is an
important notion in sociology and related research. The problem of measuring
social status has been studied for many years. Various indicators are proposed
to assess social status of individuals, including educational attainment,
occupation, and income/wealth. However, these indicators are sometimes
difficult to collect or measure.
We investigate social networks for alternative measures of social status.
Online activities expose certain traits of users in the real world. We are
interested in how these activities are related to social status, and how social
status can be predicted with social network data. To the best of our knowledge,
this is the first study on connecting online activities with social status in
reality.
In particular, we focus on the network structure of microblogs in this study.
A user following another implies some kind of status. We cast the predicted
social status of users to the "status" of real-world entities, e.g.,
universities, occupations, and regions, so that we can compare and validate
predicted results with facts in the real world. We propose an efficient
algorithm for this task and evaluate it on a dataset consisting of 3.4 million
users from Sina Weibo. The result shows that it is possible to predict social
status with reasonable accuracy using social network data. We also point out
challenges and limitations of this approach, e.g., inconsistence between online
popularity and real-world status for certain users. Our findings provide
insights on analyzing online social status and future designs of ranking
schemes for social networks.
| [
{
"created": "Sun, 10 Apr 2016 14:21:29 GMT",
"version": "v1"
}
] | 2016-04-12 | [
[
"Fu",
"Hao",
""
],
[
"Xie",
"Xing",
""
],
[
"Rui",
"Yong",
""
],
[
"Lian",
"Defu",
""
],
[
"Sun",
"Guangzhong",
""
],
[
"Chen",
"Enhong",
""
]
] | Social status refers to the relative position within the society. It is an important notion in sociology and related research. The problem of measuring social status has been studied for many years. Various indicators are proposed to assess social status of individuals, including educational attainment, occupation, and income/wealth. However, these indicators are sometimes difficult to collect or measure. We investigate social networks for alternative measures of social status. Online activities expose certain traits of users in the real world. We are interested in how these activities are related to social status, and how social status can be predicted with social network data. To the best of our knowledge, this is the first study on connecting online activities with social status in reality. In particular, we focus on the network structure of microblogs in this study. A user following another implies some kind of status. We cast the predicted social status of users to the "status" of real-world entities, e.g., universities, occupations, and regions, so that we can compare and validate predicted results with facts in the real world. We propose an efficient algorithm for this task and evaluate it on a dataset consisting of 3.4 million users from Sina Weibo. The result shows that it is possible to predict social status with reasonable accuracy using social network data. We also point out challenges and limitations of this approach, e.g., inconsistence between online popularity and real-world status for certain users. Our findings provide insights on analyzing online social status and future designs of ranking schemes for social networks. |
2010.03738 | Yang Deng | Yang Deng, Wenxuan Zhang, Wai Lam | Multi-hop Inference for Question-driven Summarization | Accepted by EMNLP 2020 (main conference, long paper) | null | null | null | cs.CL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Question-driven summarization has been recently studied as an effective
approach to summarizing the source document to produce concise but informative
answers for non-factoid questions. In this work, we propose a novel
question-driven abstractive summarization method, Multi-hop Selective Generator
(MSG), to incorporate multi-hop reasoning into question-driven summarization
and, meanwhile, provide justifications for the generated summaries.
Specifically, we jointly model the relevance to the question and the
interrelation among different sentences via a human-like multi-hop inference
module, which captures important sentences for justifying the summarized
answer. A gated selective pointer generator network with a multi-view coverage
mechanism is designed to integrate diverse information from different
perspectives. Experimental results show that the proposed method consistently
outperforms state-of-the-art methods on two non-factoid QA datasets, namely
WikiHow and PubMedQA.
| [
{
"created": "Thu, 8 Oct 2020 02:36:39 GMT",
"version": "v1"
}
] | 2020-10-09 | [
[
"Deng",
"Yang",
""
],
[
"Zhang",
"Wenxuan",
""
],
[
"Lam",
"Wai",
""
]
] | Question-driven summarization has been recently studied as an effective approach to summarizing the source document to produce concise but informative answers for non-factoid questions. In this work, we propose a novel question-driven abstractive summarization method, Multi-hop Selective Generator (MSG), to incorporate multi-hop reasoning into question-driven summarization and, meanwhile, provide justifications for the generated summaries. Specifically, we jointly model the relevance to the question and the interrelation among different sentences via a human-like multi-hop inference module, which captures important sentences for justifying the summarized answer. A gated selective pointer generator network with a multi-view coverage mechanism is designed to integrate diverse information from different perspectives. Experimental results show that the proposed method consistently outperforms state-of-the-art methods on two non-factoid QA datasets, namely WikiHow and PubMedQA. |
1105.1014 | Maurizio Martina | Maurizio Martina and Guido Masera | Improving Network-on-Chip-based turbo decoder architectures | null | null | null | null | cs.AR | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In this work novel results concerning Network-on-Chip-based turbo decoder
architectures are presented. Stemming from previous publications, this work
concentrates first on improving the throughput by exploiting adaptive-bandwidth
reduction techniques. This technique shows in the best case an improvement of
more than 60 Mb/s. Moreover, it is known that double-binary turbo decoders
require higher area than binary ones. This characteristic has the negative
effect of increasing the data width of the network nodes. Thus, the second
contribution of this work is to reduce the network complexity to support
doublebinary codes, by exploiting bit-level and pseudo-floating-point
representation of the extrinsic information. These two techniques allow for an
area reduction of up to more than the 40% with a performance degradation of
about 0.2 dB.
| [
{
"created": "Thu, 5 May 2011 08:41:43 GMT",
"version": "v1"
}
] | 2011-05-06 | [
[
"Martina",
"Maurizio",
""
],
[
"Masera",
"Guido",
""
]
] | In this work novel results concerning Network-on-Chip-based turbo decoder architectures are presented. Stemming from previous publications, this work concentrates first on improving the throughput by exploiting adaptive-bandwidth reduction techniques. This technique shows in the best case an improvement of more than 60 Mb/s. Moreover, it is known that double-binary turbo decoders require higher area than binary ones. This characteristic has the negative effect of increasing the data width of the network nodes. Thus, the second contribution of this work is to reduce the network complexity to support doublebinary codes, by exploiting bit-level and pseudo-floating-point representation of the extrinsic information. These two techniques allow for an area reduction of up to more than the 40% with a performance degradation of about 0.2 dB. |
2011.03894 | Christoforos Mavrogiannis | Junha Roh, Christoforos Mavrogiannis, Rishabh Madan, Dieter Fox,
Siddhartha S. Srinivasa | Multimodal Trajectory Prediction via Topological Invariance for
Navigation at Uncontrolled Intersections | Preprint of a paper with the same title, accepted to the Conference
on Robot Learning 2020 | null | null | null | cs.RO cs.AI cs.LG cs.MA | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We focus on decentralized navigation among multiple non-communicating
rational agents at \emph{uncontrolled} intersections, i.e., street
intersections without traffic signs or signals. Avoiding collisions in such
domains relies on the ability of agents to predict each others' intentions
reliably, and react quickly. Multiagent trajectory prediction is NP-hard
whereas the sample complexity of existing data-driven approaches limits their
applicability. Our key insight is that the geometric structure of the
intersection and the incentive of agents to move efficiently and avoid
collisions (rationality) reduces the space of likely behaviors, effectively
relaxing the problem of trajectory prediction. In this paper, we collapse the
space of multiagent trajectories at an intersection into a set of modes
representing different classes of multiagent behavior, formalized using a
notion of topological invariance. Based on this formalism, we design Multiple
Topologies Prediction (MTP), a data-driven trajectory-prediction mechanism that
reconstructs trajectory representations of high-likelihood modes in multiagent
intersection scenes. We show that MTP outperforms a state-of-the-art multimodal
trajectory prediction baseline (MFP) in terms of prediction accuracy by 78.24%
on a challenging simulated dataset. Finally, we show that MTP enables our
optimization-based planner, MTPnav, to achieve collision-free and
time-efficient navigation across a variety of challenging intersection
scenarios on the CARLA simulator.
| [
{
"created": "Sun, 8 Nov 2020 02:56:42 GMT",
"version": "v1"
}
] | 2020-11-10 | [
[
"Roh",
"Junha",
""
],
[
"Mavrogiannis",
"Christoforos",
""
],
[
"Madan",
"Rishabh",
""
],
[
"Fox",
"Dieter",
""
],
[
"Srinivasa",
"Siddhartha S.",
""
]
] | We focus on decentralized navigation among multiple non-communicating rational agents at \emph{uncontrolled} intersections, i.e., street intersections without traffic signs or signals. Avoiding collisions in such domains relies on the ability of agents to predict each others' intentions reliably, and react quickly. Multiagent trajectory prediction is NP-hard whereas the sample complexity of existing data-driven approaches limits their applicability. Our key insight is that the geometric structure of the intersection and the incentive of agents to move efficiently and avoid collisions (rationality) reduces the space of likely behaviors, effectively relaxing the problem of trajectory prediction. In this paper, we collapse the space of multiagent trajectories at an intersection into a set of modes representing different classes of multiagent behavior, formalized using a notion of topological invariance. Based on this formalism, we design Multiple Topologies Prediction (MTP), a data-driven trajectory-prediction mechanism that reconstructs trajectory representations of high-likelihood modes in multiagent intersection scenes. We show that MTP outperforms a state-of-the-art multimodal trajectory prediction baseline (MFP) in terms of prediction accuracy by 78.24% on a challenging simulated dataset. Finally, we show that MTP enables our optimization-based planner, MTPnav, to achieve collision-free and time-efficient navigation across a variety of challenging intersection scenarios on the CARLA simulator. |
2002.00556 | Jeong-Hyun Cho | Jeong-Hyun Cho, Ji-Hoon Jeong, Dong-Joo Kim, and Seong-Whan Lee | A novel approach to classify natural grasp actions by estimating muscle
activity patterns from EEG signals | 4 pages, 4 figures, conference | null | null | null | cs.HC eess.SP | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Developing electroencephalogram (EEG) based brain-computer interface (BCI)
systems is challenging. In this study, we analyzed natural grasp actions from
EEG. Ten healthy subjects participated in this experiment. They executed and
imagined three sustained grasp actions. We proposed a novel approach which
estimates muscle activity patterns from EEG signals to improve the overall
classification accuracy. For implementation, we have recorded EEG and
electromyogram (EMG) simultaneously. Using the similarity of the estimated
pattern from EEG signals compare to the activity pattern from EMG signals
showed higher classification accuracy than competitive methods. As a result, we
obtained the average classification accuracy of 63.89($\pm$7.54)% for actual
movement and 46.96($\pm$15.30)% for motor imagery. These are 21.59% and 5.66%
higher than the result of the competitive model, respectively. This result is
encouraging, and the proposed method could potentially be used in future
applications, such as a BCI-driven robot control for handling various daily use
objects.
| [
{
"created": "Mon, 3 Feb 2020 04:40:17 GMT",
"version": "v1"
}
] | 2020-02-06 | [
[
"Cho",
"Jeong-Hyun",
""
],
[
"Jeong",
"Ji-Hoon",
""
],
[
"Kim",
"Dong-Joo",
""
],
[
"Lee",
"Seong-Whan",
""
]
] | Developing electroencephalogram (EEG) based brain-computer interface (BCI) systems is challenging. In this study, we analyzed natural grasp actions from EEG. Ten healthy subjects participated in this experiment. They executed and imagined three sustained grasp actions. We proposed a novel approach which estimates muscle activity patterns from EEG signals to improve the overall classification accuracy. For implementation, we have recorded EEG and electromyogram (EMG) simultaneously. Using the similarity of the estimated pattern from EEG signals compare to the activity pattern from EMG signals showed higher classification accuracy than competitive methods. As a result, we obtained the average classification accuracy of 63.89($\pm$7.54)% for actual movement and 46.96($\pm$15.30)% for motor imagery. These are 21.59% and 5.66% higher than the result of the competitive model, respectively. This result is encouraging, and the proposed method could potentially be used in future applications, such as a BCI-driven robot control for handling various daily use objects. |
2407.10302 | Oshani Seneviratne | Oshani Seneviratne | The Feasibility of a Smart Contract "Kill Switch" | null | null | null | null | cs.CR cs.ET | http://creativecommons.org/licenses/by/4.0/ | The advent of blockchain technology and its adoption across various sectors
have raised critical discussions about the need for regulatory mechanisms to
ensure consumer protection, maintain financial stability, and address privacy
concerns without compromising the foundational principles of decentralization
and immutability inherent in blockchain platforms. We examine the existing
mechanisms for smart contract termination across several major blockchain
platforms, including Ethereum, BNB Smart Chain, Cardano, Solana, Hyperledger
Fabric, Corda, IOTA, Apotos, and Sui. We assess the compatibility of these
mechanisms with the requirements of the EU Data Act, focusing on aspects such
as consumer protection, error correction, and regulatory compliance. Our
analysis reveals a diverse landscape of approaches, from immutable smart
contracts with built-in termination conditions to upgradable smart contracts
that allow for post-deployment modifications. We discuss the challenges
associated with implementing the so-called smart contract "kill switches," such
as the balance between enabling regulatory compliance and preserving the
decentralized ethos, the technical feasibility of such mechanisms, and the
implications for security and trust in the ecosystem.
| [
{
"created": "Sun, 14 Jul 2024 19:31:15 GMT",
"version": "v1"
}
] | 2024-07-16 | [
[
"Seneviratne",
"Oshani",
""
]
] | The advent of blockchain technology and its adoption across various sectors have raised critical discussions about the need for regulatory mechanisms to ensure consumer protection, maintain financial stability, and address privacy concerns without compromising the foundational principles of decentralization and immutability inherent in blockchain platforms. We examine the existing mechanisms for smart contract termination across several major blockchain platforms, including Ethereum, BNB Smart Chain, Cardano, Solana, Hyperledger Fabric, Corda, IOTA, Apotos, and Sui. We assess the compatibility of these mechanisms with the requirements of the EU Data Act, focusing on aspects such as consumer protection, error correction, and regulatory compliance. Our analysis reveals a diverse landscape of approaches, from immutable smart contracts with built-in termination conditions to upgradable smart contracts that allow for post-deployment modifications. We discuss the challenges associated with implementing the so-called smart contract "kill switches," such as the balance between enabling regulatory compliance and preserving the decentralized ethos, the technical feasibility of such mechanisms, and the implications for security and trust in the ecosystem. |
1504.03754 | Milad Mahdian | Milad Mahdian, Edmund Yeh | Throughput and Delay Scaling of Content-Centric Ad Hoc and Heterogeneous
Wireless Networks | null | null | null | null | cs.NI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We study the throughput and delay characteristics of wireless caching
networks, where users are mainly interested in retrieving content stored in the
network, rather than in maintaining source-destination communication. Nodes are
assumed to be uniformly distributed in the network area. Each node has a
limited-capacity content store, which it uses to cache contents. We propose an
achievable caching and transmission scheme whereby requesters retrieve content
from the caching point which is closest in Euclidean distance. We establish the
throughput and delay scaling of the achievable scheme, and show that the
throughput and delay performance are order-optimal within a class of schemes.
We then solve the caching optimization problem, and evaluate the network
performance for a Zipf content popularity distribution, letting the number of
content types and the network size both go to infinity. Finally, we extend our
analysis to heterogeneous wireless networks where, in addition to wireless
nodes, there are a number of base stations uniformly distributed at random in
the network area. We show that in order to achieve a better performance in a
heterogeneous network in the order sense, the number of base stations needs to
be greater than the ratio of the number of nodes to the number of content
types. Furthermore, we show that the heterogeneous network does not yield
performance advantages in the order sense if the Zipf content popularity
distribution exponent exceeds 3/2.
| [
{
"created": "Wed, 15 Apr 2015 01:09:11 GMT",
"version": "v1"
},
{
"created": "Wed, 24 Feb 2016 00:37:33 GMT",
"version": "v2"
},
{
"created": "Mon, 25 Apr 2016 14:30:06 GMT",
"version": "v3"
},
{
"created": "Sat, 22 Apr 2017 23:59:49 GMT",
"version": "v4"
}
] | 2017-04-25 | [
[
"Mahdian",
"Milad",
""
],
[
"Yeh",
"Edmund",
""
]
] | We study the throughput and delay characteristics of wireless caching networks, where users are mainly interested in retrieving content stored in the network, rather than in maintaining source-destination communication. Nodes are assumed to be uniformly distributed in the network area. Each node has a limited-capacity content store, which it uses to cache contents. We propose an achievable caching and transmission scheme whereby requesters retrieve content from the caching point which is closest in Euclidean distance. We establish the throughput and delay scaling of the achievable scheme, and show that the throughput and delay performance are order-optimal within a class of schemes. We then solve the caching optimization problem, and evaluate the network performance for a Zipf content popularity distribution, letting the number of content types and the network size both go to infinity. Finally, we extend our analysis to heterogeneous wireless networks where, in addition to wireless nodes, there are a number of base stations uniformly distributed at random in the network area. We show that in order to achieve a better performance in a heterogeneous network in the order sense, the number of base stations needs to be greater than the ratio of the number of nodes to the number of content types. Furthermore, we show that the heterogeneous network does not yield performance advantages in the order sense if the Zipf content popularity distribution exponent exceeds 3/2. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.