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
2210.01055
Tianyu Huang
Tianyu Huang, Bowen Dong, Yunhan Yang, Xiaoshui Huang, Rynson W.H. Lau, Wanli Ouyang, Wangmeng Zuo
CLIP2Point: Transfer CLIP to Point Cloud Classification with Image-Depth Pre-training
Accepted by ICCV2023
null
null
null
cs.CV
http://creativecommons.org/licenses/by-nc-nd/4.0/
Pre-training across 3D vision and language remains under development because of limited training data. Recent works attempt to transfer vision-language pre-training models to 3D vision. PointCLIP converts point cloud data to multi-view depth maps, adopting CLIP for shape classification. However, its performance is restricted by the domain gap between rendered depth maps and images, as well as the diversity of depth distributions. To address this issue, we propose CLIP2Point, an image-depth pre-training method by contrastive learning to transfer CLIP to the 3D domain, and adapt it to point cloud classification. We introduce a new depth rendering setting that forms a better visual effect, and then render 52,460 pairs of images and depth maps from ShapeNet for pre-training. The pre-training scheme of CLIP2Point combines cross-modality learning to enforce the depth features for capturing expressive visual and textual features and intra-modality learning to enhance the invariance of depth aggregation. Additionally, we propose a novel Dual-Path Adapter (DPA) module, i.e., a dual-path structure with simplified adapters for few-shot learning. The dual-path structure allows the joint use of CLIP and CLIP2Point, and the simplified adapter can well fit few-shot tasks without post-search. Experimental results show that CLIP2Point is effective in transferring CLIP knowledge to 3D vision. Our CLIP2Point outperforms PointCLIP and other self-supervised 3D networks, achieving state-of-the-art results on zero-shot and few-shot classification.
[ { "created": "Mon, 3 Oct 2022 16:13:14 GMT", "version": "v1" }, { "created": "Sun, 20 Nov 2022 12:08:19 GMT", "version": "v2" }, { "created": "Wed, 23 Aug 2023 03:24:13 GMT", "version": "v3" } ]
2023-08-24
[ [ "Huang", "Tianyu", "" ], [ "Dong", "Bowen", "" ], [ "Yang", "Yunhan", "" ], [ "Huang", "Xiaoshui", "" ], [ "Lau", "Rynson W. H.", "" ], [ "Ouyang", "Wanli", "" ], [ "Zuo", "Wangmeng", "" ] ]
Pre-training across 3D vision and language remains under development because of limited training data. Recent works attempt to transfer vision-language pre-training models to 3D vision. PointCLIP converts point cloud data to multi-view depth maps, adopting CLIP for shape classification. However, its performance is restricted by the domain gap between rendered depth maps and images, as well as the diversity of depth distributions. To address this issue, we propose CLIP2Point, an image-depth pre-training method by contrastive learning to transfer CLIP to the 3D domain, and adapt it to point cloud classification. We introduce a new depth rendering setting that forms a better visual effect, and then render 52,460 pairs of images and depth maps from ShapeNet for pre-training. The pre-training scheme of CLIP2Point combines cross-modality learning to enforce the depth features for capturing expressive visual and textual features and intra-modality learning to enhance the invariance of depth aggregation. Additionally, we propose a novel Dual-Path Adapter (DPA) module, i.e., a dual-path structure with simplified adapters for few-shot learning. The dual-path structure allows the joint use of CLIP and CLIP2Point, and the simplified adapter can well fit few-shot tasks without post-search. Experimental results show that CLIP2Point is effective in transferring CLIP knowledge to 3D vision. Our CLIP2Point outperforms PointCLIP and other self-supervised 3D networks, achieving state-of-the-art results on zero-shot and few-shot classification.
2304.07162
Thomas Neele
Thomas Neele, Jaco van de Pol
Operations on Fixpoint Equation Systems
null
Logical Methods in Computer Science, Volume 20, Issue 3 (July 10, 2024) lmcs:11199
10.46298/lmcs-20(3:5)2024
null
cs.LO
http://creativecommons.org/licenses/by/4.0/
We study operations on fixpoint equation systems (FES) over arbitrary complete lattices. We investigate under which conditions these operations, such as substituting variables by their definition, and swapping the ordering of equations, preserve the solution of a FES. We provide rigorous, computer-checked proofs. Along the way, we list a number of known and new identities and inequalities on extremal fixpoints in complete lattices.
[ { "created": "Fri, 14 Apr 2023 14:35:18 GMT", "version": "v1" }, { "created": "Fri, 9 Feb 2024 14:23:48 GMT", "version": "v2" }, { "created": "Tue, 14 May 2024 08:49:22 GMT", "version": "v3" }, { "created": "Tue, 9 Jul 2024 11:24:33 GMT", "version": "v4" } ]
2024-08-07
[ [ "Neele", "Thomas", "" ], [ "van de Pol", "Jaco", "" ] ]
We study operations on fixpoint equation systems (FES) over arbitrary complete lattices. We investigate under which conditions these operations, such as substituting variables by their definition, and swapping the ordering of equations, preserve the solution of a FES. We provide rigorous, computer-checked proofs. Along the way, we list a number of known and new identities and inequalities on extremal fixpoints in complete lattices.
2201.04678
Weidong Luo
Weidong Luo
Polynomial Turing Compressions for Some Graph Problems Parameterized by Modular-Width
18 pages
null
null
null
cs.DS
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
A polynomial Turing compression (PTC) for a parameterized problem $L$ is a polynomial time Turing machine that has access to an oracle for a problem $L'$ such that a polynomial in the input parameter bounds each query. Meanwhile, a polynomial (many-one) compression (PC) can be regarded as a restricted variant of PTC where the machine can query the oracle exactly once and must output the same answer as the oracle. Bodlaender et al. (ICALP 2008) and Fortnow and Santhanam (STOC 2008) initiated an impressive hardness theory for PC under the assumption coNP $\not\subseteq$ NP/poly. Since PTC is a generalization of PC, we define $\mathcal{C}$ as the set of all problems that have PTCs but have no PCs under the assumption coNP $\not\subseteq$ NP/poly. Based on the hardness theory for PC, Fernau et al. (STACS 2009) found the first problem Leaf Out-tree($k$) in $\mathcal{C}$. However, very little is known about $\mathcal{C}$, as only a dozen problems were shown to belong to the complexity class in the last ten years. Several problems are open, for example, whether CNF-SAT($n$) and $k$-path are in $\mathcal{C}$, and novel ideas are required to better understand the fundamental differences between PTCs and PCs. In this paper, we enrich our knowledge about $\mathcal{C}$ by showing that several problems parameterized by modular-width ($mw$) belong to $\mathcal{C}$. More specifically, exploiting the properties of the well-studied structural graph parameter $mw$, we demonstrate 17 problems parameterized by $mw$ are in $\mathcal{C}$, such as Chromatic Number($mw$) and Hamiltonian Cycle($mw$). In addition, we develop a general recipe to prove the existence of PTCs for a large class of problems, including our 17 problems.
[ { "created": "Wed, 12 Jan 2022 20:12:41 GMT", "version": "v1" }, { "created": "Thu, 21 Apr 2022 16:59:42 GMT", "version": "v2" }, { "created": "Thu, 14 Dec 2023 04:17:26 GMT", "version": "v3" } ]
2023-12-15
[ [ "Luo", "Weidong", "" ] ]
A polynomial Turing compression (PTC) for a parameterized problem $L$ is a polynomial time Turing machine that has access to an oracle for a problem $L'$ such that a polynomial in the input parameter bounds each query. Meanwhile, a polynomial (many-one) compression (PC) can be regarded as a restricted variant of PTC where the machine can query the oracle exactly once and must output the same answer as the oracle. Bodlaender et al. (ICALP 2008) and Fortnow and Santhanam (STOC 2008) initiated an impressive hardness theory for PC under the assumption coNP $\not\subseteq$ NP/poly. Since PTC is a generalization of PC, we define $\mathcal{C}$ as the set of all problems that have PTCs but have no PCs under the assumption coNP $\not\subseteq$ NP/poly. Based on the hardness theory for PC, Fernau et al. (STACS 2009) found the first problem Leaf Out-tree($k$) in $\mathcal{C}$. However, very little is known about $\mathcal{C}$, as only a dozen problems were shown to belong to the complexity class in the last ten years. Several problems are open, for example, whether CNF-SAT($n$) and $k$-path are in $\mathcal{C}$, and novel ideas are required to better understand the fundamental differences between PTCs and PCs. In this paper, we enrich our knowledge about $\mathcal{C}$ by showing that several problems parameterized by modular-width ($mw$) belong to $\mathcal{C}$. More specifically, exploiting the properties of the well-studied structural graph parameter $mw$, we demonstrate 17 problems parameterized by $mw$ are in $\mathcal{C}$, such as Chromatic Number($mw$) and Hamiltonian Cycle($mw$). In addition, we develop a general recipe to prove the existence of PTCs for a large class of problems, including our 17 problems.
0908.2083
Janusz Brzozowski
J. Brzozowski, G. Jir\'askov\'a, B. Li
Quotient complexity of ideal languages
24 pages, 9 .eepic figures, 2 tables, use llncs.cls
null
null
null
cs.FL
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We study the state complexity of regular operations in the class of ideal languages. A language L over an alphabet Sigma is a right (left) ideal if it satisfies L = L Sigma* (L = Sigma* L). It is a two-sided ideal if L = Sigma* L Sigma *, and an all-sided ideal if it is the shuffle of Sigma* with L. We prefer the term "quotient complexity" instead of "state complexity", and we use derivatives to calculate upper bounds on quotient complexity, whenever it is convenient. We find tight upper bounds on the quotient complexity of each type of ideal language in terms of the complexity of an arbitrary generator and of its minimal generator, the complexity of the minimal generator, and also on the operations union, intersection, set difference, symmetric difference, concatenation, star and reversal of ideal languages.
[ { "created": "Fri, 14 Aug 2009 15:24:15 GMT", "version": "v1" } ]
2009-08-17
[ [ "Brzozowski", "J.", "" ], [ "Jirásková", "G.", "" ], [ "Li", "B.", "" ] ]
We study the state complexity of regular operations in the class of ideal languages. A language L over an alphabet Sigma is a right (left) ideal if it satisfies L = L Sigma* (L = Sigma* L). It is a two-sided ideal if L = Sigma* L Sigma *, and an all-sided ideal if it is the shuffle of Sigma* with L. We prefer the term "quotient complexity" instead of "state complexity", and we use derivatives to calculate upper bounds on quotient complexity, whenever it is convenient. We find tight upper bounds on the quotient complexity of each type of ideal language in terms of the complexity of an arbitrary generator and of its minimal generator, the complexity of the minimal generator, and also on the operations union, intersection, set difference, symmetric difference, concatenation, star and reversal of ideal languages.
2108.11299
Tobias Lorenz
Tobias Lorenz, Marta Kwiatkowska, Mario Fritz
Certifiers Make Neural Networks Vulnerable to Availability Attacks
Published at 16th ACM Workshop on Artificial Intelligence and Security (AISec '23)
null
10.1145/3605764.3623917
null
cs.LG cs.CR cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
To achieve reliable, robust, and safe AI systems, it is vital to implement fallback strategies when AI predictions cannot be trusted. Certifiers for neural networks are a reliable way to check the robustness of these predictions. They guarantee for some predictions that a certain class of manipulations or attacks could not have changed the outcome. For the remaining predictions without guarantees, the method abstains from making a prediction, and a fallback strategy needs to be invoked, which typically incurs additional costs, can require a human operator, or even fail to provide any prediction. While this is a key concept towards safe and secure AI, we show for the first time that this approach comes with its own security risks, as such fallback strategies can be deliberately triggered by an adversary. In addition to naturally occurring abstains for some inputs and perturbations, the adversary can use training-time attacks to deliberately trigger the fallback with high probability. This transfers the main system load onto the fallback, reducing the overall system's integrity and/or availability. We design two novel availability attacks, which show the practical relevance of these threats. For example, adding 1% poisoned data during training is sufficient to trigger the fallback and hence make the model unavailable for up to 100% of all inputs by inserting the trigger. Our extensive experiments across multiple datasets, model architectures, and certifiers demonstrate the broad applicability of these attacks. An initial investigation into potential defenses shows that current approaches are insufficient to mitigate the issue, highlighting the need for new, specific solutions.
[ { "created": "Wed, 25 Aug 2021 15:49:10 GMT", "version": "v1" }, { "created": "Mon, 7 Mar 2022 09:42:15 GMT", "version": "v2" }, { "created": "Fri, 13 May 2022 12:11:56 GMT", "version": "v3" }, { "created": "Sun, 2 Oct 2022 16:58:47 GMT", "version": "v4" }, { "created": "Tue, 3 Oct 2023 13:08:50 GMT", "version": "v5" } ]
2023-10-04
[ [ "Lorenz", "Tobias", "" ], [ "Kwiatkowska", "Marta", "" ], [ "Fritz", "Mario", "" ] ]
To achieve reliable, robust, and safe AI systems, it is vital to implement fallback strategies when AI predictions cannot be trusted. Certifiers for neural networks are a reliable way to check the robustness of these predictions. They guarantee for some predictions that a certain class of manipulations or attacks could not have changed the outcome. For the remaining predictions without guarantees, the method abstains from making a prediction, and a fallback strategy needs to be invoked, which typically incurs additional costs, can require a human operator, or even fail to provide any prediction. While this is a key concept towards safe and secure AI, we show for the first time that this approach comes with its own security risks, as such fallback strategies can be deliberately triggered by an adversary. In addition to naturally occurring abstains for some inputs and perturbations, the adversary can use training-time attacks to deliberately trigger the fallback with high probability. This transfers the main system load onto the fallback, reducing the overall system's integrity and/or availability. We design two novel availability attacks, which show the practical relevance of these threats. For example, adding 1% poisoned data during training is sufficient to trigger the fallback and hence make the model unavailable for up to 100% of all inputs by inserting the trigger. Our extensive experiments across multiple datasets, model architectures, and certifiers demonstrate the broad applicability of these attacks. An initial investigation into potential defenses shows that current approaches are insufficient to mitigate the issue, highlighting the need for new, specific solutions.
2004.12091
Onur G\"unl\"u Dr.-Ing.
Onur G\"unl\"u, Peter Trifonov, Muah Kim, Rafael F. Schaefer, and Vladimir Sidorenko
Randomized Nested Polar Subcode Constructions for Privacy, Secrecy, and Storage
Shorter version to appear in 2020 IEEE International Symposium on Information Theory and Applications. Decoding complexity results are added
null
null
null
cs.IT cs.CR cs.MM eess.SP math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We consider polar subcodes (PSCs), which are polar codes (PCs) with dynamically-frozen symbols, to increase the minimum distance as compared to corresponding PCs. A randomized nested PSC construction with a low-rate PSC and a high-rate PC, is proposed for list and sequential successive cancellation decoders. This code construction aims to perform lossy compression with side information. Nested PSCs are used in the key agreement problem with physical identifiers. Gains in terms of the secret-key vs. storage rate ratio as compared to nested PCs with the same list size are illustrated to show that nested PSCs significantly improve on nested PCs. The performance of the nested PSCs is shown to improve with larger list sizes, which is not the case for nested PCs considered.
[ { "created": "Sat, 25 Apr 2020 08:57:17 GMT", "version": "v1" }, { "created": "Tue, 30 Jun 2020 12:57:02 GMT", "version": "v2" }, { "created": "Wed, 29 Jul 2020 10:26:19 GMT", "version": "v3" } ]
2020-07-30
[ [ "Günlü", "Onur", "" ], [ "Trifonov", "Peter", "" ], [ "Kim", "Muah", "" ], [ "Schaefer", "Rafael F.", "" ], [ "Sidorenko", "Vladimir", "" ] ]
We consider polar subcodes (PSCs), which are polar codes (PCs) with dynamically-frozen symbols, to increase the minimum distance as compared to corresponding PCs. A randomized nested PSC construction with a low-rate PSC and a high-rate PC, is proposed for list and sequential successive cancellation decoders. This code construction aims to perform lossy compression with side information. Nested PSCs are used in the key agreement problem with physical identifiers. Gains in terms of the secret-key vs. storage rate ratio as compared to nested PCs with the same list size are illustrated to show that nested PSCs significantly improve on nested PCs. The performance of the nested PSCs is shown to improve with larger list sizes, which is not the case for nested PCs considered.
2012.07959
Li-Yi Wei
Peihan Tu, Li-Yi Wei, Koji Yatani, Takeo Igarashi, Matthias Zwicker
Continuous Curve Textures
null
null
10.1145/3414685.3417780
null
cs.GR
http://creativecommons.org/licenses/by/4.0/
Repetitive patterns are ubiquitous in natural and human-made objects, and can be created with a variety of tools and methods. Manual authoring provides unmatched degree of freedom and control, but can require significant artistic expertise and manual labor. Computational methods can automate parts of the manual creation process, but are mainly tailored for discrete pixels or elements instead of more general continuous structures. We propose an example-based method to synthesize continuous curve patterns from exemplars. Our main idea is to extend prior sample-based discrete element synthesis methods to consider not only sample positions (geometry) but also their connections (topology). Since continuous structures can exhibit higher complexity than discrete elements, we also propose robust, hierarchical synthesis to enhance output quality. Our algorithm can generate a variety of continuous curve patterns fully automatically. For further quality improvement and customization, we also present an autocomplete user interface to facilitate interactive creation and iterative editing. We evaluate our methods and interface via different patterns, ablation studies, and comparisons with alternative methods.
[ { "created": "Mon, 14 Dec 2020 21:51:17 GMT", "version": "v1" } ]
2020-12-16
[ [ "Tu", "Peihan", "" ], [ "Wei", "Li-Yi", "" ], [ "Yatani", "Koji", "" ], [ "Igarashi", "Takeo", "" ], [ "Zwicker", "Matthias", "" ] ]
Repetitive patterns are ubiquitous in natural and human-made objects, and can be created with a variety of tools and methods. Manual authoring provides unmatched degree of freedom and control, but can require significant artistic expertise and manual labor. Computational methods can automate parts of the manual creation process, but are mainly tailored for discrete pixels or elements instead of more general continuous structures. We propose an example-based method to synthesize continuous curve patterns from exemplars. Our main idea is to extend prior sample-based discrete element synthesis methods to consider not only sample positions (geometry) but also their connections (topology). Since continuous structures can exhibit higher complexity than discrete elements, we also propose robust, hierarchical synthesis to enhance output quality. Our algorithm can generate a variety of continuous curve patterns fully automatically. For further quality improvement and customization, we also present an autocomplete user interface to facilitate interactive creation and iterative editing. We evaluate our methods and interface via different patterns, ablation studies, and comparisons with alternative methods.
2212.14527
Sikun Yang
Sikun Yang, Hongyuan Zha
Estimating Latent Population Flows from Aggregated Data via Inversing Multi-Marginal Optimal Transport
null
null
null
null
cs.LG
http://creativecommons.org/publicdomain/zero/1.0/
We study the problem of estimating latent population flows from aggregated count data. This problem arises when individual trajectories are not available due to privacy issues or measurement fidelity. Instead, the aggregated observations are measured over discrete-time points, for estimating the population flows among states. Most related studies tackle the problems by learning the transition parameters of a time-homogeneous Markov process. Nonetheless, most real-world population flows can be influenced by various uncertainties such as traffic jam and weather conditions. Thus, in many cases, a time-homogeneous Markov model is a poor approximation of the much more complex population flows. To circumvent this difficulty, we resort to a multi-marginal optimal transport (MOT) formulation that can naturally represent aggregated observations with constrained marginals, and encode time-dependent transition matrices by the cost functions. In particular, we propose to estimate the transition flows from aggregated data by learning the cost functions of the MOT framework, which enables us to capture time-varying dynamic patterns. The experiments demonstrate the improved accuracy of the proposed algorithms than the related methods in estimating several real-world transition flows.
[ { "created": "Fri, 30 Dec 2022 03:03:23 GMT", "version": "v1" } ]
2023-01-02
[ [ "Yang", "Sikun", "" ], [ "Zha", "Hongyuan", "" ] ]
We study the problem of estimating latent population flows from aggregated count data. This problem arises when individual trajectories are not available due to privacy issues or measurement fidelity. Instead, the aggregated observations are measured over discrete-time points, for estimating the population flows among states. Most related studies tackle the problems by learning the transition parameters of a time-homogeneous Markov process. Nonetheless, most real-world population flows can be influenced by various uncertainties such as traffic jam and weather conditions. Thus, in many cases, a time-homogeneous Markov model is a poor approximation of the much more complex population flows. To circumvent this difficulty, we resort to a multi-marginal optimal transport (MOT) formulation that can naturally represent aggregated observations with constrained marginals, and encode time-dependent transition matrices by the cost functions. In particular, we propose to estimate the transition flows from aggregated data by learning the cost functions of the MOT framework, which enables us to capture time-varying dynamic patterns. The experiments demonstrate the improved accuracy of the proposed algorithms than the related methods in estimating several real-world transition flows.
1612.04164
Stefan Wagner
Sebastian V\"ost and Stefan Wagner
Keeping Continuous Deliveries Safe
4 pages, 3 figures
ICSE-C '17 Proceedings of the 39th International Conference on Software Engineering Companion, pages 259-261. IEEE, 2017
10.1109/ICSE-C.2017.135
null
cs.SE
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Allowing swift release cycles, Continuous Delivery has become popular in application software development and is starting to be applied in safety-critical domains such as the automotive industry. These domains require thorough analysis regarding safety constraints, which can be achieved by formal verification and the execution of safety tests resulting from a safety analysis on the product. With continuous delivery in place, such tests need to be executed with every build to ensure the latest software still fulfills all safety requirements. Even more though, the safety analysis has to be updated with every change to ensure the safety test suite is still up-to-date. We thus propose that a safety analysis should be treated no differently from other deliverables such as source-code and dependencies, formulate guidelines on how to achieve this and advert areas where future research is needed.
[ { "created": "Tue, 13 Dec 2016 13:38:24 GMT", "version": "v1" } ]
2017-11-15
[ [ "Vöst", "Sebastian", "" ], [ "Wagner", "Stefan", "" ] ]
Allowing swift release cycles, Continuous Delivery has become popular in application software development and is starting to be applied in safety-critical domains such as the automotive industry. These domains require thorough analysis regarding safety constraints, which can be achieved by formal verification and the execution of safety tests resulting from a safety analysis on the product. With continuous delivery in place, such tests need to be executed with every build to ensure the latest software still fulfills all safety requirements. Even more though, the safety analysis has to be updated with every change to ensure the safety test suite is still up-to-date. We thus propose that a safety analysis should be treated no differently from other deliverables such as source-code and dependencies, formulate guidelines on how to achieve this and advert areas where future research is needed.
2404.03099
Leonardo Ferreira Guilhoto
Leonardo Ferreira Guilhoto, Paris Perdikaris
Composite Bayesian Optimization In Function Spaces Using NEON -- Neural Epistemic Operator Networks
null
null
null
null
cs.LG cs.AI cs.CE cs.IT math.IT stat.ML
http://creativecommons.org/licenses/by-nc-sa/4.0/
Operator learning is a rising field of scientific computing where inputs or outputs of a machine learning model are functions defined in infinite-dimensional spaces. In this paper, we introduce NEON (Neural Epistemic Operator Networks), an architecture for generating predictions with uncertainty using a single operator network backbone, which presents orders of magnitude less trainable parameters than deep ensembles of comparable performance. We showcase the utility of this method for sequential decision-making by examining the problem of composite Bayesian Optimization (BO), where we aim to optimize a function $f=g\circ h$, where $h:X\to C(\mathcal{Y},\mathbb{R}^{d_s})$ is an unknown map which outputs elements of a function space, and $g: C(\mathcal{Y},\mathbb{R}^{d_s})\to \mathbb{R}$ is a known and cheap-to-compute functional. By comparing our approach to other state-of-the-art methods on toy and real world scenarios, we demonstrate that NEON achieves state-of-the-art performance while requiring orders of magnitude less trainable parameters.
[ { "created": "Wed, 3 Apr 2024 22:42:37 GMT", "version": "v1" } ]
2024-04-05
[ [ "Guilhoto", "Leonardo Ferreira", "" ], [ "Perdikaris", "Paris", "" ] ]
Operator learning is a rising field of scientific computing where inputs or outputs of a machine learning model are functions defined in infinite-dimensional spaces. In this paper, we introduce NEON (Neural Epistemic Operator Networks), an architecture for generating predictions with uncertainty using a single operator network backbone, which presents orders of magnitude less trainable parameters than deep ensembles of comparable performance. We showcase the utility of this method for sequential decision-making by examining the problem of composite Bayesian Optimization (BO), where we aim to optimize a function $f=g\circ h$, where $h:X\to C(\mathcal{Y},\mathbb{R}^{d_s})$ is an unknown map which outputs elements of a function space, and $g: C(\mathcal{Y},\mathbb{R}^{d_s})\to \mathbb{R}$ is a known and cheap-to-compute functional. By comparing our approach to other state-of-the-art methods on toy and real world scenarios, we demonstrate that NEON achieves state-of-the-art performance while requiring orders of magnitude less trainable parameters.
1802.10233
Daniel Lemire
Edmon Begoli, Jes\'us Camacho Rodr\'iguez, Julian Hyde, Michael J. Mior, Daniel Lemire
Apache Calcite: A Foundational Framework for Optimized Query Processing Over Heterogeneous Data Sources
SIGMOD'18
null
10.1145/3183713.3190662
null
cs.DB
http://creativecommons.org/licenses/by/4.0/
Apache Calcite is a foundational software framework that provides query processing, optimization, and query language support to many popular open-source data processing systems such as Apache Hive, Apache Storm, Apache Flink, Druid, and MapD. Calcite's architecture consists of a modular and extensible query optimizer with hundreds of built-in optimization rules, a query processor capable of processing a variety of query languages, an adapter architecture designed for extensibility, and support for heterogeneous data models and stores (relational, semi-structured, streaming, and geospatial). This flexible, embeddable, and extensible architecture is what makes Calcite an attractive choice for adoption in big-data frameworks. It is an active project that continues to introduce support for the new types of data sources, query languages, and approaches to query processing and optimization.
[ { "created": "Wed, 28 Feb 2018 02:10:36 GMT", "version": "v1" } ]
2020-10-09
[ [ "Begoli", "Edmon", "" ], [ "Rodríguez", "Jesús Camacho", "" ], [ "Hyde", "Julian", "" ], [ "Mior", "Michael J.", "" ], [ "Lemire", "Daniel", "" ] ]
Apache Calcite is a foundational software framework that provides query processing, optimization, and query language support to many popular open-source data processing systems such as Apache Hive, Apache Storm, Apache Flink, Druid, and MapD. Calcite's architecture consists of a modular and extensible query optimizer with hundreds of built-in optimization rules, a query processor capable of processing a variety of query languages, an adapter architecture designed for extensibility, and support for heterogeneous data models and stores (relational, semi-structured, streaming, and geospatial). This flexible, embeddable, and extensible architecture is what makes Calcite an attractive choice for adoption in big-data frameworks. It is an active project that continues to introduce support for the new types of data sources, query languages, and approaches to query processing and optimization.
2006.05814
Paul Mireault
Paul Mireault
Implementation Strategies for Multidimensional Spreadsheets
12 Pages, 18 Colour Figures. arXiv admin note: text overlap with arXiv:1801.09777
Proceedings of the EuSpRIG 2019 Conference "Spreadsheet Risk Management", Browns, Covent Garden, London, pp103-114, ISBN: 978-1-905404-56-8
null
null
cs.SE
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Seasoned Excel developers were invited to participate in a challenge to implement a spreadsheet with multi-dimensional variables. We analyzed their spreadsheet to see the different implement strategies employed. We identified two strategies: most participants used a projection of three or four-dimensional variables on the two-dimensional plane used by Excel. A few participants used a database approach where the multi-dimensional variables are presented in the form of a dataset table with the appropriate primary key. This approach leads to simpler formulas.
[ { "created": "Thu, 4 Jun 2020 21:12:06 GMT", "version": "v1" } ]
2020-06-11
[ [ "Mireault", "Paul", "" ] ]
Seasoned Excel developers were invited to participate in a challenge to implement a spreadsheet with multi-dimensional variables. We analyzed their spreadsheet to see the different implement strategies employed. We identified two strategies: most participants used a projection of three or four-dimensional variables on the two-dimensional plane used by Excel. A few participants used a database approach where the multi-dimensional variables are presented in the form of a dataset table with the appropriate primary key. This approach leads to simpler formulas.
2004.00307
Filipe Assun\c{c}\~ao
Filipe Assun\c{c}\~ao, Nuno Louren\c{c}o, Bernardete Ribeiro, and Penousal Machado
Evolution of Scikit-Learn Pipelines with Dynamic Structured Grammatical Evolution
EvoApps 2020
null
null
null
cs.NE cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The deployment of Machine Learning (ML) models is a difficult and time-consuming job that comprises a series of sequential and correlated tasks that go from the data pre-processing, and the design and extraction of features, to the choice of the ML algorithm and its parameterisation. The task is even more challenging considering that the design of features is in many cases problem specific, and thus requires domain-expertise. To overcome these limitations Automated Machine Learning (AutoML) methods seek to automate, with few or no human-intervention, the design of pipelines, i.e., automate the selection of the sequence of methods that have to be applied to the raw data. These methods have the potential to enable non-expert users to use ML, and provide expert users with solutions that they would unlikely consider. In particular, this paper describes AutoML-DSGE - a novel grammar-based framework that adapts Dynamic Structured Grammatical Evolution (DSGE) to the evolution of Scikit-Learn classification pipelines. The experimental results include comparing AutoML-DSGE to another grammar-based AutoML framework, Resilient ClassificationPipeline Evolution (RECIPE), and show that the average performance of the classification pipelines generated by AutoML-DSGE is always superior to the average performance of RECIPE; the differences are statistically significant in 3 out of the 10 used datasets.
[ { "created": "Wed, 1 Apr 2020 09:31:34 GMT", "version": "v1" } ]
2020-04-02
[ [ "Assunção", "Filipe", "" ], [ "Lourenço", "Nuno", "" ], [ "Ribeiro", "Bernardete", "" ], [ "Machado", "Penousal", "" ] ]
The deployment of Machine Learning (ML) models is a difficult and time-consuming job that comprises a series of sequential and correlated tasks that go from the data pre-processing, and the design and extraction of features, to the choice of the ML algorithm and its parameterisation. The task is even more challenging considering that the design of features is in many cases problem specific, and thus requires domain-expertise. To overcome these limitations Automated Machine Learning (AutoML) methods seek to automate, with few or no human-intervention, the design of pipelines, i.e., automate the selection of the sequence of methods that have to be applied to the raw data. These methods have the potential to enable non-expert users to use ML, and provide expert users with solutions that they would unlikely consider. In particular, this paper describes AutoML-DSGE - a novel grammar-based framework that adapts Dynamic Structured Grammatical Evolution (DSGE) to the evolution of Scikit-Learn classification pipelines. The experimental results include comparing AutoML-DSGE to another grammar-based AutoML framework, Resilient ClassificationPipeline Evolution (RECIPE), and show that the average performance of the classification pipelines generated by AutoML-DSGE is always superior to the average performance of RECIPE; the differences are statistically significant in 3 out of the 10 used datasets.
2407.03059
Mariia Vladimirova
Mariia Vladimirova, Federico Pavone, Eustache Diemert
FairJob: A Real-World Dataset for Fairness in Online Systems
24 pages, 15 figures
null
null
null
cs.LG cs.AI cs.CY stat.ML
http://creativecommons.org/licenses/by/4.0/
We introduce a fairness-aware dataset for job recommendation in advertising, designed to foster research in algorithmic fairness within real-world scenarios. It was collected and prepared to comply with privacy standards and business confidentiality. An additional challenge is the lack of access to protected user attributes such as gender, for which we propose a solution to obtain a proxy estimate. Despite being anonymized and including a proxy for a sensitive attribute, our dataset preserves predictive power and maintains a realistic and challenging benchmark. This dataset addresses a significant gap in the availability of fairness-focused resources for high-impact domains like advertising -- the actual impact being having access or not to precious employment opportunities, where balancing fairness and utility is a common industrial challenge. We also explore various stages in the advertising process where unfairness can occur and introduce a method to compute a fair utility metric for the job recommendations in online systems case from a biased dataset. Experimental evaluations of bias mitigation techniques on the released dataset demonstrate potential improvements in fairness and the associated trade-offs with utility.
[ { "created": "Wed, 3 Jul 2024 12:30:39 GMT", "version": "v1" } ]
2024-07-04
[ [ "Vladimirova", "Mariia", "" ], [ "Pavone", "Federico", "" ], [ "Diemert", "Eustache", "" ] ]
We introduce a fairness-aware dataset for job recommendation in advertising, designed to foster research in algorithmic fairness within real-world scenarios. It was collected and prepared to comply with privacy standards and business confidentiality. An additional challenge is the lack of access to protected user attributes such as gender, for which we propose a solution to obtain a proxy estimate. Despite being anonymized and including a proxy for a sensitive attribute, our dataset preserves predictive power and maintains a realistic and challenging benchmark. This dataset addresses a significant gap in the availability of fairness-focused resources for high-impact domains like advertising -- the actual impact being having access or not to precious employment opportunities, where balancing fairness and utility is a common industrial challenge. We also explore various stages in the advertising process where unfairness can occur and introduce a method to compute a fair utility metric for the job recommendations in online systems case from a biased dataset. Experimental evaluations of bias mitigation techniques on the released dataset demonstrate potential improvements in fairness and the associated trade-offs with utility.
2001.10721
Shunchuan Yang
Yu Cheng, Guangzhi Chen, Xiang-Hua Wang, and Shunchuan Yang
Investigation of Numerical Dispersion with Time Step of The FDTD Methods: Avoiding Erroneous Conclusions
null
null
null
null
cs.CE
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
It is widely thought that small time steps lead to small numerical errors in the finite-difference time-domain (FDTD) simulations. In this paper, we investigated how time steps impact on numerical dispersion of two FDTD methods including the FDTD(2,2) method and the FDTD(2,4) method. Through rigorously analytical and numerical analysis, it is found that small time steps of the FDTD methods do not always have small numerical errors. Our findings reveal that these two FDTD methods present different behaviors with respect to time steps: (1) for the FDTD(2,2) method, smaller time steps limited by the Courant-Friedrichs-Lewy (CFL) condition increase numerical dispersion and lead to larger simulation errors; (2) for the FDTD(2,4) method, as time step increases, numerical dispersion errors first decrease and then increase. Our findings are also comprehensively validated from one- to three-dimensional cases through several numerical examples including wave propagation, resonant frequencies of cavities and a practical electromagnetic compatibility (EMC) problem.
[ { "created": "Wed, 29 Jan 2020 08:28:46 GMT", "version": "v1" }, { "created": "Thu, 20 Feb 2020 16:46:30 GMT", "version": "v2" } ]
2020-02-21
[ [ "Cheng", "Yu", "" ], [ "Chen", "Guangzhi", "" ], [ "Wang", "Xiang-Hua", "" ], [ "Yang", "Shunchuan", "" ] ]
It is widely thought that small time steps lead to small numerical errors in the finite-difference time-domain (FDTD) simulations. In this paper, we investigated how time steps impact on numerical dispersion of two FDTD methods including the FDTD(2,2) method and the FDTD(2,4) method. Through rigorously analytical and numerical analysis, it is found that small time steps of the FDTD methods do not always have small numerical errors. Our findings reveal that these two FDTD methods present different behaviors with respect to time steps: (1) for the FDTD(2,2) method, smaller time steps limited by the Courant-Friedrichs-Lewy (CFL) condition increase numerical dispersion and lead to larger simulation errors; (2) for the FDTD(2,4) method, as time step increases, numerical dispersion errors first decrease and then increase. Our findings are also comprehensively validated from one- to three-dimensional cases through several numerical examples including wave propagation, resonant frequencies of cavities and a practical electromagnetic compatibility (EMC) problem.
2205.09453
Roxana Danger
Roxana Danger
Differential Privacy: What is all the noise about?
27 pages, 7 figures
null
null
null
cs.CR cs.AI cs.LG
http://creativecommons.org/licenses/by/4.0/
Differential Privacy (DP) is a formal definition of privacy that provides rigorous guarantees against risks of privacy breaches during data processing. It makes no assumptions about the knowledge or computational power of adversaries, and provides an interpretable, quantifiable and composable formalism. DP has been actively researched during the last 15 years, but it is still hard to master for many Machine Learning (ML)) practitioners. This paper aims to provide an overview of the most important ideas, concepts and uses of DP in ML, with special focus on its intersection with Federated Learning (FL).
[ { "created": "Thu, 19 May 2022 10:12:29 GMT", "version": "v1" } ]
2022-05-20
[ [ "Danger", "Roxana", "" ] ]
Differential Privacy (DP) is a formal definition of privacy that provides rigorous guarantees against risks of privacy breaches during data processing. It makes no assumptions about the knowledge or computational power of adversaries, and provides an interpretable, quantifiable and composable formalism. DP has been actively researched during the last 15 years, but it is still hard to master for many Machine Learning (ML)) practitioners. This paper aims to provide an overview of the most important ideas, concepts and uses of DP in ML, with special focus on its intersection with Federated Learning (FL).
1807.04118
Takumi Ichimura
Takumi Ichimura, Takuya Uemoto, Akira Hara
Emergence of Altruism Behavior for Multi Feeding Areas in Army Ant Social Evolutionary System
6 pages, 11 figures
Proc. of 2014 IEEE International Conference on Systems, Man, and Cybernetics (IEEE SMC 2014)
10.1109/SMC.2014.6973902
null
cs.MA cs.ET cs.NE
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Army ants perform the altruism that an ant sacrifices its own well-being for the benefit of another ants. Army ants build bridges using their own bodies along the path from a food to the nest. We developed the army ant inspired social evolutionary system which can perform the altruism. The system has 2 kinds of ant agents, `Major ant' and `Minor ant' and the ants communicate with each other via pheromones. One ants can recognize them as the signals from the other ants. The pheromones evaporate with the certain ratio and diffused into the space of neighbors stochastically. If the optimal bridge is found, the path through the bridge is the shortest route from foods to the nest. We define the probability for an ant to leave a bridge at a low occupancy condition of ants and propose the constructing method of the optimal route. In this paper, the behaviors of ant under the environment with two or more feeding spots are observed. Some experimental results show the behaviors of great interest with respect to altruism of ants. The description in some computer simulation is reported in this paper.
[ { "created": "Tue, 10 Jul 2018 04:40:38 GMT", "version": "v1" } ]
2018-07-12
[ [ "Ichimura", "Takumi", "" ], [ "Uemoto", "Takuya", "" ], [ "Hara", "Akira", "" ] ]
Army ants perform the altruism that an ant sacrifices its own well-being for the benefit of another ants. Army ants build bridges using their own bodies along the path from a food to the nest. We developed the army ant inspired social evolutionary system which can perform the altruism. The system has 2 kinds of ant agents, `Major ant' and `Minor ant' and the ants communicate with each other via pheromones. One ants can recognize them as the signals from the other ants. The pheromones evaporate with the certain ratio and diffused into the space of neighbors stochastically. If the optimal bridge is found, the path through the bridge is the shortest route from foods to the nest. We define the probability for an ant to leave a bridge at a low occupancy condition of ants and propose the constructing method of the optimal route. In this paper, the behaviors of ant under the environment with two or more feeding spots are observed. Some experimental results show the behaviors of great interest with respect to altruism of ants. The description in some computer simulation is reported in this paper.
2106.11097
Pengfei Xiong
Han Fang, Pengfei Xiong, Luhui Xu, Yu Chen
CLIP2Video: Mastering Video-Text Retrieval via Image CLIP
null
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We present CLIP2Video network to transfer the image-language pre-training model to video-text retrieval in an end-to-end manner. Leading approaches in the domain of video-and-language learning try to distill the spatio-temporal video features and multi-modal interaction between videos and languages from a large-scale video-text dataset. Different from them, we leverage pretrained image-language model, simplify it as a two-stage framework with co-learning of image-text and enhancing temporal relations between video frames and video-text respectively, make it able to train on comparatively small datasets. Specifically, based on the spatial semantics captured by Contrastive Language-Image Pretraining (CLIP) model, our model involves a Temporal Difference Block to capture motions at fine temporal video frames, and a Temporal Alignment Block to re-align the tokens of video clips and phrases and enhance the multi-modal correlation. We conduct thorough ablation studies, and achieve state-of-the-art performance on major text-to-video and video-to-text retrieval benchmarks, including new records of retrieval accuracy on MSR-VTT, MSVD and VATEX.
[ { "created": "Mon, 21 Jun 2021 13:30:33 GMT", "version": "v1" } ]
2021-06-22
[ [ "Fang", "Han", "" ], [ "Xiong", "Pengfei", "" ], [ "Xu", "Luhui", "" ], [ "Chen", "Yu", "" ] ]
We present CLIP2Video network to transfer the image-language pre-training model to video-text retrieval in an end-to-end manner. Leading approaches in the domain of video-and-language learning try to distill the spatio-temporal video features and multi-modal interaction between videos and languages from a large-scale video-text dataset. Different from them, we leverage pretrained image-language model, simplify it as a two-stage framework with co-learning of image-text and enhancing temporal relations between video frames and video-text respectively, make it able to train on comparatively small datasets. Specifically, based on the spatial semantics captured by Contrastive Language-Image Pretraining (CLIP) model, our model involves a Temporal Difference Block to capture motions at fine temporal video frames, and a Temporal Alignment Block to re-align the tokens of video clips and phrases and enhance the multi-modal correlation. We conduct thorough ablation studies, and achieve state-of-the-art performance on major text-to-video and video-to-text retrieval benchmarks, including new records of retrieval accuracy on MSR-VTT, MSVD and VATEX.
0906.2135
Michael Nelson
Herbert Van de Sompel, Carl Lagoze, Michael L. Nelson, Simeon Warner, Robert Sanderson, Pete Johnston
Adding eScience Assets to the Data Web
10 pages, 7 figures. Proceedings of Linked Data on the Web (LDOW2009) Workshop
null
null
null
cs.DL
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Aggregations of Web resources are increasingly important in scholarship as it adopts new methods that are data-centric, collaborative, and networked-based. The same notion of aggregations of resources is common to the mashed-up, socially networked information environment of Web 2.0. We present a mechanism to identify and describe aggregations of Web resources that has resulted from the Open Archives Initiative - Object Reuse and Exchange (OAI-ORE) project. The OAI-ORE specifications are based on the principles of the Architecture of the World Wide Web, the Semantic Web, and the Linked Data effort. Therefore, their incorporation into the cyberinfrastructure that supports eScholarship will ensure the integration of the products of scholarly research into the Data Web.
[ { "created": "Thu, 11 Jun 2009 15:33:37 GMT", "version": "v1" } ]
2009-06-12
[ [ "Van de Sompel", "Herbert", "" ], [ "Lagoze", "Carl", "" ], [ "Nelson", "Michael L.", "" ], [ "Warner", "Simeon", "" ], [ "Sanderson", "Robert", "" ], [ "Johnston", "Pete", "" ] ]
Aggregations of Web resources are increasingly important in scholarship as it adopts new methods that are data-centric, collaborative, and networked-based. The same notion of aggregations of resources is common to the mashed-up, socially networked information environment of Web 2.0. We present a mechanism to identify and describe aggregations of Web resources that has resulted from the Open Archives Initiative - Object Reuse and Exchange (OAI-ORE) project. The OAI-ORE specifications are based on the principles of the Architecture of the World Wide Web, the Semantic Web, and the Linked Data effort. Therefore, their incorporation into the cyberinfrastructure that supports eScholarship will ensure the integration of the products of scholarly research into the Data Web.
1312.6461
Sho Sonoda
Sho Sonoda, Noboru Murata
Nonparametric Weight Initialization of Neural Networks via Integral Representation
For ICLR2014, revised into 9 pages; revised into 12 pages (with supplements)
null
null
null
cs.LG cs.NE
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
A new initialization method for hidden parameters in a neural network is proposed. Derived from the integral representation of the neural network, a nonparametric probability distribution of hidden parameters is introduced. In this proposal, hidden parameters are initialized by samples drawn from this distribution, and output parameters are fitted by ordinary linear regression. Numerical experiments show that backpropagation with proposed initialization converges faster than uniformly random initialization. Also it is shown that the proposed method achieves enough accuracy by itself without backpropagation in some cases.
[ { "created": "Mon, 23 Dec 2013 03:23:04 GMT", "version": "v1" }, { "created": "Tue, 24 Dec 2013 02:54:29 GMT", "version": "v2" }, { "created": "Wed, 19 Feb 2014 20:02:05 GMT", "version": "v3" } ]
2014-02-20
[ [ "Sonoda", "Sho", "" ], [ "Murata", "Noboru", "" ] ]
A new initialization method for hidden parameters in a neural network is proposed. Derived from the integral representation of the neural network, a nonparametric probability distribution of hidden parameters is introduced. In this proposal, hidden parameters are initialized by samples drawn from this distribution, and output parameters are fitted by ordinary linear regression. Numerical experiments show that backpropagation with proposed initialization converges faster than uniformly random initialization. Also it is shown that the proposed method achieves enough accuracy by itself without backpropagation in some cases.
1811.06295
Chen Du
Chen Du, Chunheng Wang, Yanna Wang, Cunzhao Shi, Baihua Xiao
Selective Feature Connection Mechanism: Concatenating Multi-layer CNN Features with a Feature Selector
The paper is under consideration at Pattern Recognition Letters
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Different layers of deep convolutional neural networks(CNNs) can encode different-level information. High-layer features always contain more semantic information, and low-layer features contain more detail information. However, low-layer features suffer from the background clutter and semantic ambiguity. During visual recognition, the feature combination of the low-layer and high-level features plays an important role in context modulation. If directly combining the high-layer and low-layer features, the background clutter and semantic ambiguity may be caused due to the introduction of detailed information. In this paper, we propose a general network architecture to concatenate CNN features of different layers in a simple and effective way, called Selective Feature Connection Mechanism (SFCM). Low-level features are selectively linked to high-level features with a feature selector which is generated by high-level features. The proposed connection mechanism can effectively overcome the above-mentioned drawbacks. We demonstrate the effectiveness, superiority, and universal applicability of this method on multiple challenging computer vision tasks, including image classification, scene text detection, and image-to-image translation.
[ { "created": "Thu, 15 Nov 2018 10:58:21 GMT", "version": "v1" }, { "created": "Wed, 17 Apr 2019 05:53:51 GMT", "version": "v2" }, { "created": "Sun, 21 Apr 2019 07:44:13 GMT", "version": "v3" } ]
2019-04-23
[ [ "Du", "Chen", "" ], [ "Wang", "Chunheng", "" ], [ "Wang", "Yanna", "" ], [ "Shi", "Cunzhao", "" ], [ "Xiao", "Baihua", "" ] ]
Different layers of deep convolutional neural networks(CNNs) can encode different-level information. High-layer features always contain more semantic information, and low-layer features contain more detail information. However, low-layer features suffer from the background clutter and semantic ambiguity. During visual recognition, the feature combination of the low-layer and high-level features plays an important role in context modulation. If directly combining the high-layer and low-layer features, the background clutter and semantic ambiguity may be caused due to the introduction of detailed information. In this paper, we propose a general network architecture to concatenate CNN features of different layers in a simple and effective way, called Selective Feature Connection Mechanism (SFCM). Low-level features are selectively linked to high-level features with a feature selector which is generated by high-level features. The proposed connection mechanism can effectively overcome the above-mentioned drawbacks. We demonstrate the effectiveness, superiority, and universal applicability of this method on multiple challenging computer vision tasks, including image classification, scene text detection, and image-to-image translation.
1902.09759
Shuai Wang
Shuai Wang, Minghua Xia, and Yik-Chung Wu
Joint Communication and Motion Energy Minimization in UGV Backscatter Communication
Proc. IEEE ICC'19, Shanghai, China, May 2019, 6 pages
null
null
null
cs.IT math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
While backscatter communication emerges as a promising solution to reduce power consumption at IoT devices, the transmission range of backscatter communication is short. To this end, this work integrates unmanned ground vehicles (UGVs) into the backscatter system. With such a scheme, the UGV could facilitate the communication by approaching various IoT devices. However, moving also costs energy consumption and a fundamental question is: what is the right balance between spending energy on moving versus on communication? To answer this question, this paper proposes a joint graph mobility and backscatter communication model. With the proposed model, the total energy minimization at UGV is formulated as a mixed integer nonlinear programming (MINLP) problem. Furthermore, an efficient algorithm that achieves a local optimal solution is derived, and it leads to automatic trade-off between spending energy on moving versus on communication. Numerical results are provided to validate the performance of the proposed algorithm.
[ { "created": "Tue, 26 Feb 2019 06:55:37 GMT", "version": "v1" } ]
2019-02-27
[ [ "Wang", "Shuai", "" ], [ "Xia", "Minghua", "" ], [ "Wu", "Yik-Chung", "" ] ]
While backscatter communication emerges as a promising solution to reduce power consumption at IoT devices, the transmission range of backscatter communication is short. To this end, this work integrates unmanned ground vehicles (UGVs) into the backscatter system. With such a scheme, the UGV could facilitate the communication by approaching various IoT devices. However, moving also costs energy consumption and a fundamental question is: what is the right balance between spending energy on moving versus on communication? To answer this question, this paper proposes a joint graph mobility and backscatter communication model. With the proposed model, the total energy minimization at UGV is formulated as a mixed integer nonlinear programming (MINLP) problem. Furthermore, an efficient algorithm that achieves a local optimal solution is derived, and it leads to automatic trade-off between spending energy on moving versus on communication. Numerical results are provided to validate the performance of the proposed algorithm.
2210.09347
Huy Ha
Alper Canberk, Cheng Chi, Huy Ha, Benjamin Burchfiel, Eric Cousineau, Siyuan Feng, Shuran Song
Cloth Funnels: Canonicalized-Alignment for Multi-Purpose Garment Manipulation
8 pages, 8 figures, website at https://clothfunnels.cs.columbia.edu/
null
null
null
cs.RO
http://creativecommons.org/licenses/by/4.0/
Automating garment manipulation is challenging due to extremely high variability in object configurations. To reduce this intrinsic variation, we introduce the task of "canonicalized-alignment" that simplifies downstream applications by reducing the possible garment configurations. This task can be considered as "cloth state funnel" that manipulates arbitrarily configured clothing items into a predefined deformable configuration (i.e. canonicalization) at an appropriate rigid pose (i.e. alignment). In the end, the cloth items will result in a compact set of structured and highly visible configurations - which are desirable for downstream manipulation skills. To enable this task, we propose a novel canonicalized-alignment objective that effectively guides learning to avoid adverse local minima during learning. Using this objective, we learn a multi-arm, multi-primitive policy that strategically chooses between dynamic flings and quasi-static pick and place actions to achieve efficient canonicalized-alignment. We evaluate this approach on a real-world ironing and folding system that relies on this learned policy as the common first step. Empirically, we demonstrate that our task-agnostic canonicalized-alignment can enable even simple manually-designed policies to work well where they were previously inadequate, thus bridging the gap between automated non-deformable manufacturing and deformable manipulation. Code and qualitative visualizations are available at https://clothfunnels.cs.columbia.edu/. Video can be found at https://www.youtube.com/watch?v=TkUn0b7mbj0.
[ { "created": "Mon, 17 Oct 2022 18:36:59 GMT", "version": "v1" } ]
2022-10-19
[ [ "Canberk", "Alper", "" ], [ "Chi", "Cheng", "" ], [ "Ha", "Huy", "" ], [ "Burchfiel", "Benjamin", "" ], [ "Cousineau", "Eric", "" ], [ "Feng", "Siyuan", "" ], [ "Song", "Shuran", "" ] ]
Automating garment manipulation is challenging due to extremely high variability in object configurations. To reduce this intrinsic variation, we introduce the task of "canonicalized-alignment" that simplifies downstream applications by reducing the possible garment configurations. This task can be considered as "cloth state funnel" that manipulates arbitrarily configured clothing items into a predefined deformable configuration (i.e. canonicalization) at an appropriate rigid pose (i.e. alignment). In the end, the cloth items will result in a compact set of structured and highly visible configurations - which are desirable for downstream manipulation skills. To enable this task, we propose a novel canonicalized-alignment objective that effectively guides learning to avoid adverse local minima during learning. Using this objective, we learn a multi-arm, multi-primitive policy that strategically chooses between dynamic flings and quasi-static pick and place actions to achieve efficient canonicalized-alignment. We evaluate this approach on a real-world ironing and folding system that relies on this learned policy as the common first step. Empirically, we demonstrate that our task-agnostic canonicalized-alignment can enable even simple manually-designed policies to work well where they were previously inadequate, thus bridging the gap between automated non-deformable manufacturing and deformable manipulation. Code and qualitative visualizations are available at https://clothfunnels.cs.columbia.edu/. Video can be found at https://www.youtube.com/watch?v=TkUn0b7mbj0.
2201.01391
Ademola Okerinde
Ademola Okerinde and Sam Hoggatt and Divya Vani Lakkireddy and Nolan Brubaker and William Hsu and Lior Shamir and Brian Spiesman
Self-Supervised Approach to Addressing Zero-Shot Learning Problem
null
The 4th International Conference on Computing and Data Science (CONF-CDS 2022)
null
null
cs.CV cs.LG
http://creativecommons.org/licenses/by/4.0/
In recent years, self-supervised learning has had significant success in applications involving computer vision and natural language processing. The type of pretext task is important to this boost in performance. One common pretext task is the measure of similarity and dissimilarity between pairs of images. In this scenario, the two images that make up the negative pair are visibly different to humans. However, in entomology, species are nearly indistinguishable and thus hard to differentiate. In this study, we explored the performance of a Siamese neural network using contrastive loss by learning to push apart embeddings of bumblebee species pair that are dissimilar, and pull together similar embeddings. Our experimental results show a 61% F1-score on zero-shot instances, a performance showing 11% improvement on samples of classes that share intersections with the training set.
[ { "created": "Wed, 5 Jan 2022 00:08:36 GMT", "version": "v1" }, { "created": "Fri, 21 Jan 2022 14:09:29 GMT", "version": "v2" } ]
2022-01-24
[ [ "Okerinde", "Ademola", "" ], [ "Hoggatt", "Sam", "" ], [ "Lakkireddy", "Divya Vani", "" ], [ "Brubaker", "Nolan", "" ], [ "Hsu", "William", "" ], [ "Shamir", "Lior", "" ], [ "Spiesman", "Brian", "" ] ]
In recent years, self-supervised learning has had significant success in applications involving computer vision and natural language processing. The type of pretext task is important to this boost in performance. One common pretext task is the measure of similarity and dissimilarity between pairs of images. In this scenario, the two images that make up the negative pair are visibly different to humans. However, in entomology, species are nearly indistinguishable and thus hard to differentiate. In this study, we explored the performance of a Siamese neural network using contrastive loss by learning to push apart embeddings of bumblebee species pair that are dissimilar, and pull together similar embeddings. Our experimental results show a 61% F1-score on zero-shot instances, a performance showing 11% improvement on samples of classes that share intersections with the training set.
2205.03651
Manjanna Basappa
Vishwanath R. Singireddy and Manjanna Basappa
Dispersing Facilities on Planar Segment and Circle Amidst Repulsion
16 figures
null
null
null
cs.CG
http://creativecommons.org/licenses/by/4.0/
In this paper we consider the problem of locating $k$ obnoxious facilities (congruent disks of maximum radius) amidst $n$ demand points (existing repulsive facility sites) ordered from left to right in the plane so that none of the existing facility sites are affected (no demand point falls in the interior of the disks). We study this problem in two restricted settings: (i) the obnoxious facilities are constrained to be centered on along a predetermined horizontal line segment $\bar{pq}$, and (ii) the obnoxious facilities are constrained to lie on the boundary arc of a predetermined disk $\cal C$. An $(1-\epsilon)$-approximation algorithm was given recently to solve the constrained problem in (i) in time $O((n+k)\log{\frac{||pq||}{2(k-1)\epsilon}})$, where $\epsilon>0$ \cite{Sing2021}. Here, for the problem in (i), we first propose an exact polynomial-time algorithm based on a binary search on all candidate radii computed explicitly. This algorithm runs in $O((nk)^2\log{(nk)}+(n+k)\log{(nk)})$ time. We then show that using the parametric search technique of Megiddo \cite{MG1983}; we can solve the problem exactly in $O((n+k)^2)$ time, which is faster than the latter. Continuing further, using the improved parametric technique we give an $O(n\log^2 n)$-time algorithm for $k=2$. We finally show that the above $(1-\epsilon)$-approximation algorithm of \cite{Sing2021} can be easily adapted to solve the circular constrained problem of (ii) with an extra multiplicative factor of $n$ in the running time.
[ { "created": "Sat, 7 May 2022 13:16:04 GMT", "version": "v1" }, { "created": "Thu, 12 May 2022 11:17:01 GMT", "version": "v2" } ]
2022-05-13
[ [ "Singireddy", "Vishwanath R.", "" ], [ "Basappa", "Manjanna", "" ] ]
In this paper we consider the problem of locating $k$ obnoxious facilities (congruent disks of maximum radius) amidst $n$ demand points (existing repulsive facility sites) ordered from left to right in the plane so that none of the existing facility sites are affected (no demand point falls in the interior of the disks). We study this problem in two restricted settings: (i) the obnoxious facilities are constrained to be centered on along a predetermined horizontal line segment $\bar{pq}$, and (ii) the obnoxious facilities are constrained to lie on the boundary arc of a predetermined disk $\cal C$. An $(1-\epsilon)$-approximation algorithm was given recently to solve the constrained problem in (i) in time $O((n+k)\log{\frac{||pq||}{2(k-1)\epsilon}})$, where $\epsilon>0$ \cite{Sing2021}. Here, for the problem in (i), we first propose an exact polynomial-time algorithm based on a binary search on all candidate radii computed explicitly. This algorithm runs in $O((nk)^2\log{(nk)}+(n+k)\log{(nk)})$ time. We then show that using the parametric search technique of Megiddo \cite{MG1983}; we can solve the problem exactly in $O((n+k)^2)$ time, which is faster than the latter. Continuing further, using the improved parametric technique we give an $O(n\log^2 n)$-time algorithm for $k=2$. We finally show that the above $(1-\epsilon)$-approximation algorithm of \cite{Sing2021} can be easily adapted to solve the circular constrained problem of (ii) with an extra multiplicative factor of $n$ in the running time.
2401.06826
Jialiang Tang
Jialiang Tang, Shuo Chen, Gang Niu, Hongyuan Zhu, Joey Tianyi Zhou, Chen Gong, Masashi Sugiyama
Direct Distillation between Different Domains
null
null
null
null
cs.LG cs.AI cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Knowledge Distillation (KD) aims to learn a compact student network using knowledge from a large pre-trained teacher network, where both networks are trained on data from the same distribution. However, in practical applications, the student network may be required to perform in a new scenario (i.e., the target domain), which usually exhibits significant differences from the known scenario of the teacher network (i.e., the source domain). The traditional domain adaptation techniques can be integrated with KD in a two-stage process to bridge the domain gap, but the ultimate reliability of two-stage approaches tends to be limited due to the high computational consumption and the additional errors accumulated from both stages. To solve this problem, we propose a new one-stage method dubbed ``Direct Distillation between Different Domains" (4Ds). We first design a learnable adapter based on the Fourier transform to separate the domain-invariant knowledge from the domain-specific knowledge. Then, we build a fusion-activation mechanism to transfer the valuable domain-invariant knowledge to the student network, while simultaneously encouraging the adapter within the teacher network to learn the domain-specific knowledge of the target data. As a result, the teacher network can effectively transfer categorical knowledge that aligns with the target domain of the student network. Intensive experiments on various benchmark datasets demonstrate that our proposed 4Ds method successfully produces reliable student networks and outperforms state-of-the-art approaches.
[ { "created": "Fri, 12 Jan 2024 02:48:51 GMT", "version": "v1" } ]
2024-01-17
[ [ "Tang", "Jialiang", "" ], [ "Chen", "Shuo", "" ], [ "Niu", "Gang", "" ], [ "Zhu", "Hongyuan", "" ], [ "Zhou", "Joey Tianyi", "" ], [ "Gong", "Chen", "" ], [ "Sugiyama", "Masashi", "" ] ]
Knowledge Distillation (KD) aims to learn a compact student network using knowledge from a large pre-trained teacher network, where both networks are trained on data from the same distribution. However, in practical applications, the student network may be required to perform in a new scenario (i.e., the target domain), which usually exhibits significant differences from the known scenario of the teacher network (i.e., the source domain). The traditional domain adaptation techniques can be integrated with KD in a two-stage process to bridge the domain gap, but the ultimate reliability of two-stage approaches tends to be limited due to the high computational consumption and the additional errors accumulated from both stages. To solve this problem, we propose a new one-stage method dubbed ``Direct Distillation between Different Domains" (4Ds). We first design a learnable adapter based on the Fourier transform to separate the domain-invariant knowledge from the domain-specific knowledge. Then, we build a fusion-activation mechanism to transfer the valuable domain-invariant knowledge to the student network, while simultaneously encouraging the adapter within the teacher network to learn the domain-specific knowledge of the target data. As a result, the teacher network can effectively transfer categorical knowledge that aligns with the target domain of the student network. Intensive experiments on various benchmark datasets demonstrate that our proposed 4Ds method successfully produces reliable student networks and outperforms state-of-the-art approaches.
2312.16174
Yujiao Hu
Yujiao Hu, Qingmin Jia, Yuao Yao, Yong Lee, Mengjie Lee, Chenyi Wang, Xiaomao Zhou, Renchao Xie, F. Richard Yu
Industrial Internet of Things Intelligence Empowering Smart Manufacturing: A Literature Review
Accepted by IoTJ
IEEE Internet of Things Journal,2024
10.1109/JIOT.2024.3367692
null
cs.AI cs.CY
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The fiercely competitive business environment and increasingly personalized customization needs are driving the digital transformation and upgrading of the manufacturing industry. IIoT intelligence, which can provide innovative and efficient solutions for various aspects of the manufacturing value chain, illuminates the path of transformation for the manufacturing industry. It's time to provide a systematic vision of IIoT intelligence. However, existing surveys often focus on specific areas of IIoT intelligence, leading researchers and readers to have biases in their understanding of IIoT intelligence, that is, believing that research in one direction is the most important for the development of IIoT intelligence, while ignoring contributions from other directions. Therefore, this paper provides a comprehensive overview of IIoT intelligence. We first conduct an in-depth analysis of the inevitability of manufacturing transformation and study the successful experiences from the practices of Chinese enterprises. Then we give our definition of IIoT intelligence and demonstrate the value of IIoT intelligence for industries in fucntions, operations, deployments, and application. Afterwards, we propose a hierarchical development architecture for IIoT intelligence, which consists of five layers. The practical values of technical upgrades at each layer are illustrated by a close look on lighthouse factories. Following that, we identify seven kinds of technologies that accelerate the transformation of manufacturing, and clarify their contributions. The ethical implications and environmental impacts of adopting IIoT intelligence in manufacturing are analyzed as well. Finally, we explore the open challenges and development trends from four aspects to inspire future researches.
[ { "created": "Sat, 2 Dec 2023 06:08:39 GMT", "version": "v1" }, { "created": "Thu, 22 Feb 2024 02:28:57 GMT", "version": "v2" } ]
2024-02-23
[ [ "Hu", "Yujiao", "" ], [ "Jia", "Qingmin", "" ], [ "Yao", "Yuao", "" ], [ "Lee", "Yong", "" ], [ "Lee", "Mengjie", "" ], [ "Wang", "Chenyi", "" ], [ "Zhou", "Xiaomao", "" ], [ "Xie", "Renchao", "" ], [ "Yu", "F. Richard", "" ] ]
The fiercely competitive business environment and increasingly personalized customization needs are driving the digital transformation and upgrading of the manufacturing industry. IIoT intelligence, which can provide innovative and efficient solutions for various aspects of the manufacturing value chain, illuminates the path of transformation for the manufacturing industry. It's time to provide a systematic vision of IIoT intelligence. However, existing surveys often focus on specific areas of IIoT intelligence, leading researchers and readers to have biases in their understanding of IIoT intelligence, that is, believing that research in one direction is the most important for the development of IIoT intelligence, while ignoring contributions from other directions. Therefore, this paper provides a comprehensive overview of IIoT intelligence. We first conduct an in-depth analysis of the inevitability of manufacturing transformation and study the successful experiences from the practices of Chinese enterprises. Then we give our definition of IIoT intelligence and demonstrate the value of IIoT intelligence for industries in fucntions, operations, deployments, and application. Afterwards, we propose a hierarchical development architecture for IIoT intelligence, which consists of five layers. The practical values of technical upgrades at each layer are illustrated by a close look on lighthouse factories. Following that, we identify seven kinds of technologies that accelerate the transformation of manufacturing, and clarify their contributions. The ethical implications and environmental impacts of adopting IIoT intelligence in manufacturing are analyzed as well. Finally, we explore the open challenges and development trends from four aspects to inspire future researches.
1907.05205
Vidyasagar Sadhu
Vidyasagar Sadhu, Sanjana Devaraj, Dario Pompili
Towards Ultra-low-power Realization of Analog Joint Source-Channel Coding using MOSFETs
5 pages, IEEE ISCAS 2019. arXiv admin note: text overlap with arXiv:1907.00968
2019 IEEE International Symposium on Circuits and Systems (ISCAS), Sapporo, Japan, 2019, pp. 1-5
10.1109/ISCAS.2019.8702302
null
cs.ET cs.NI eess.SP
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Certain sensing applications such as Internet of Things (IoTs), where the sensing phenomenon may change rapidly in both time and space, requires sensors that consume ultra-low power (so that they do not need to be put to sleep leading to loss of temporal and spatial resolution) and have low costs (for high density deployment). A novel encoding based on Metal Oxide Semiconductor Field Effect Transistors (MOSFETs) is proposed to realize Analog Joint Source Channel Coding (AJSCC), a low-complexity technique to compress two (or more) signals into one with controlled distortion. In AJSCC, the y-axis is quantized while the x-axis is continuously captured. A power-efficient design to support multiple quantization levels is presented so that the digital receiver can decide the optimum quantization and the analog transmitter circuit is able to realize that. The approach is verified via Spice and MATLAB simulations.
[ { "created": "Sun, 30 Jun 2019 06:46:58 GMT", "version": "v1" } ]
2019-07-12
[ [ "Sadhu", "Vidyasagar", "" ], [ "Devaraj", "Sanjana", "" ], [ "Pompili", "Dario", "" ] ]
Certain sensing applications such as Internet of Things (IoTs), where the sensing phenomenon may change rapidly in both time and space, requires sensors that consume ultra-low power (so that they do not need to be put to sleep leading to loss of temporal and spatial resolution) and have low costs (for high density deployment). A novel encoding based on Metal Oxide Semiconductor Field Effect Transistors (MOSFETs) is proposed to realize Analog Joint Source Channel Coding (AJSCC), a low-complexity technique to compress two (or more) signals into one with controlled distortion. In AJSCC, the y-axis is quantized while the x-axis is continuously captured. A power-efficient design to support multiple quantization levels is presented so that the digital receiver can decide the optimum quantization and the analog transmitter circuit is able to realize that. The approach is verified via Spice and MATLAB simulations.
1902.04742
Vaishnavh Nagarajan
Vaishnavh Nagarajan, J. Zico Kolter
Uniform convergence may be unable to explain generalization in deep learning
null
null
null
null
cs.LG stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Aimed at explaining the surprisingly good generalization behavior of overparameterized deep networks, recent works have developed a variety of generalization bounds for deep learning, all based on the fundamental learning-theoretic technique of uniform convergence. While it is well-known that many of these existing bounds are numerically large, through numerous experiments, we bring to light a more concerning aspect of these bounds: in practice, these bounds can {\em increase} with the training dataset size. Guided by our observations, we then present examples of overparameterized linear classifiers and neural networks trained by gradient descent (GD) where uniform convergence provably cannot "explain generalization" -- even if we take into account the implicit bias of GD {\em to the fullest extent possible}. More precisely, even if we consider only the set of classifiers output by GD, which have test errors less than some small $\epsilon$ in our settings, we show that applying (two-sided) uniform convergence on this set of classifiers will yield only a vacuous generalization guarantee larger than $1-\epsilon$. Through these findings, we cast doubt on the power of uniform convergence-based generalization bounds to provide a complete picture of why overparameterized deep networks generalize well.
[ { "created": "Wed, 13 Feb 2019 05:09:38 GMT", "version": "v1" }, { "created": "Tue, 2 Apr 2019 17:20:02 GMT", "version": "v2" }, { "created": "Thu, 19 Dec 2019 20:09:41 GMT", "version": "v3" }, { "created": "Sun, 17 Oct 2021 20:04:09 GMT", "version": "v4" } ]
2021-10-19
[ [ "Nagarajan", "Vaishnavh", "" ], [ "Kolter", "J. Zico", "" ] ]
Aimed at explaining the surprisingly good generalization behavior of overparameterized deep networks, recent works have developed a variety of generalization bounds for deep learning, all based on the fundamental learning-theoretic technique of uniform convergence. While it is well-known that many of these existing bounds are numerically large, through numerous experiments, we bring to light a more concerning aspect of these bounds: in practice, these bounds can {\em increase} with the training dataset size. Guided by our observations, we then present examples of overparameterized linear classifiers and neural networks trained by gradient descent (GD) where uniform convergence provably cannot "explain generalization" -- even if we take into account the implicit bias of GD {\em to the fullest extent possible}. More precisely, even if we consider only the set of classifiers output by GD, which have test errors less than some small $\epsilon$ in our settings, we show that applying (two-sided) uniform convergence on this set of classifiers will yield only a vacuous generalization guarantee larger than $1-\epsilon$. Through these findings, we cast doubt on the power of uniform convergence-based generalization bounds to provide a complete picture of why overparameterized deep networks generalize well.
2311.10372
Kaiwen Ning
Zibin Zheng and Kaiwen Ning and Yanlin Wang and Jingwen Zhang and Dewu Zheng and Mingxi Ye and Jiachi Chen
A Survey of Large Language Models for Code: Evolution, Benchmarking, and Future Trends
null
null
null
null
cs.SE
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
General large language models (LLMs), represented by ChatGPT, have demonstrated significant potential in tasks such as code generation in software engineering. This has led to the development of specialized LLMs for software engineering, known as Code LLMs. A considerable portion of Code LLMs is derived from general LLMs through model fine-tuning. As a result, Code LLMs are often updated frequently and their performance can be influenced by the base LLMs. However, there is currently a lack of systematic investigation into Code LLMs and their performance. In this study, we conduct a comprehensive survey and analysis of the types of Code LLMs and their differences in performance compared to general LLMs. We aim to address three questions: (1) What LLMs are specifically designed for software engineering tasks, and what is the relationship between these Code LLMs? (2) Do Code LLMs really outperform general LLMs in software engineering tasks? (3) Which LLMs are more proficient in different software engineering tasks? To answer these questions, we first collect relevant literature and work from five major databases and open-source communities, resulting in 134 works for analysis. Next, we categorize the Code LLMs based on their publishers and examine their relationships with general LLMs and among themselves. Furthermore, we investigate the performance differences between general LLMs and Code LLMs in various software engineering tasks to demonstrate the impact of base models and Code LLMs. Finally, we comprehensively maintained the performance of LLMs across multiple mainstream benchmarks to identify the best-performing LLMs for each software engineering task. Our research not only assists developers of Code LLMs in choosing base models for the development of more advanced LLMs but also provides insights for practitioners to better understand key improvement directions for Code LLMs.
[ { "created": "Fri, 17 Nov 2023 07:55:16 GMT", "version": "v1" }, { "created": "Mon, 8 Jan 2024 05:41:51 GMT", "version": "v2" } ]
2024-01-09
[ [ "Zheng", "Zibin", "" ], [ "Ning", "Kaiwen", "" ], [ "Wang", "Yanlin", "" ], [ "Zhang", "Jingwen", "" ], [ "Zheng", "Dewu", "" ], [ "Ye", "Mingxi", "" ], [ "Chen", "Jiachi", "" ] ]
General large language models (LLMs), represented by ChatGPT, have demonstrated significant potential in tasks such as code generation in software engineering. This has led to the development of specialized LLMs for software engineering, known as Code LLMs. A considerable portion of Code LLMs is derived from general LLMs through model fine-tuning. As a result, Code LLMs are often updated frequently and their performance can be influenced by the base LLMs. However, there is currently a lack of systematic investigation into Code LLMs and their performance. In this study, we conduct a comprehensive survey and analysis of the types of Code LLMs and their differences in performance compared to general LLMs. We aim to address three questions: (1) What LLMs are specifically designed for software engineering tasks, and what is the relationship between these Code LLMs? (2) Do Code LLMs really outperform general LLMs in software engineering tasks? (3) Which LLMs are more proficient in different software engineering tasks? To answer these questions, we first collect relevant literature and work from five major databases and open-source communities, resulting in 134 works for analysis. Next, we categorize the Code LLMs based on their publishers and examine their relationships with general LLMs and among themselves. Furthermore, we investigate the performance differences between general LLMs and Code LLMs in various software engineering tasks to demonstrate the impact of base models and Code LLMs. Finally, we comprehensively maintained the performance of LLMs across multiple mainstream benchmarks to identify the best-performing LLMs for each software engineering task. Our research not only assists developers of Code LLMs in choosing base models for the development of more advanced LLMs but also provides insights for practitioners to better understand key improvement directions for Code LLMs.
1806.00801
Gui-Song Xia
Pu Jin, Gui-Song Xia, Fan Hu, Qikai Lu, Liangpei Zhang
AID++: An Updated Version of AID on Scene Classification
IGARSS'18 conference paper
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Aerial image scene classification is a fundamental problem for understanding high-resolution remote sensing images and has become an active research task in the field of remote sensing due to its important role in a wide range of applications. However, the limitations of existing datasets for scene classification, such as the small scale and low-diversity, severely hamper the potential usage of the new generation deep convolutional neural networks (CNNs). Although huge efforts have been made in building large-scale datasets very recently, e.g., the Aerial Image Dataset (AID) which contains 10,000 image samples, they are still far from sufficient to fully train a high-capacity deep CNN model. To this end, we present a larger-scale dataset in this paper, named as AID++, for aerial scene classification based on the AID dataset. The proposed AID++ consists of more than 400,000 image samples that are semi-automatically annotated by using the existing the geographical data. We evaluate several prevalent CNN models on the proposed dataset, and the results show that our dataset can be used as a promising benchmark for scene classification.
[ { "created": "Sun, 3 Jun 2018 14:40:20 GMT", "version": "v1" } ]
2018-06-05
[ [ "Jin", "Pu", "" ], [ "Xia", "Gui-Song", "" ], [ "Hu", "Fan", "" ], [ "Lu", "Qikai", "" ], [ "Zhang", "Liangpei", "" ] ]
Aerial image scene classification is a fundamental problem for understanding high-resolution remote sensing images and has become an active research task in the field of remote sensing due to its important role in a wide range of applications. However, the limitations of existing datasets for scene classification, such as the small scale and low-diversity, severely hamper the potential usage of the new generation deep convolutional neural networks (CNNs). Although huge efforts have been made in building large-scale datasets very recently, e.g., the Aerial Image Dataset (AID) which contains 10,000 image samples, they are still far from sufficient to fully train a high-capacity deep CNN model. To this end, we present a larger-scale dataset in this paper, named as AID++, for aerial scene classification based on the AID dataset. The proposed AID++ consists of more than 400,000 image samples that are semi-automatically annotated by using the existing the geographical data. We evaluate several prevalent CNN models on the proposed dataset, and the results show that our dataset can be used as a promising benchmark for scene classification.
1809.02768
Yifan Gao
Yifan Gao, Lidong Bing, Piji Li, Irwin King, Michael R. Lyu
Generating Distractors for Reading Comprehension Questions from Real Examinations
AAAI2019
null
null
null
cs.CL
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We investigate the task of distractor generation for multiple choice reading comprehension questions from examinations. In contrast to all previous works, we do not aim at preparing words or short phrases distractors, instead, we endeavor to generate longer and semantic-rich distractors which are closer to distractors in real reading comprehension from examinations. Taking a reading comprehension article, a pair of question and its correct option as input, our goal is to generate several distractors which are somehow related to the answer, consistent with the semantic context of the question and have some trace in the article. We propose a hierarchical encoder-decoder framework with static and dynamic attention mechanisms to tackle this task. Specifically, the dynamic attention can combine sentence-level and word-level attention varying at each recurrent time step to generate a more readable sequence. The static attention is to modulate the dynamic attention not to focus on question irrelevant sentences or sentences which contribute to the correct option. Our proposed framework outperforms several strong baselines on the first prepared distractor generation dataset of real reading comprehension questions. For human evaluation, compared with those distractors generated by baselines, our generated distractors are more functional to confuse the annotators.
[ { "created": "Sat, 8 Sep 2018 07:11:15 GMT", "version": "v1" }, { "created": "Tue, 18 Dec 2018 07:04:50 GMT", "version": "v2" } ]
2018-12-19
[ [ "Gao", "Yifan", "" ], [ "Bing", "Lidong", "" ], [ "Li", "Piji", "" ], [ "King", "Irwin", "" ], [ "Lyu", "Michael R.", "" ] ]
We investigate the task of distractor generation for multiple choice reading comprehension questions from examinations. In contrast to all previous works, we do not aim at preparing words or short phrases distractors, instead, we endeavor to generate longer and semantic-rich distractors which are closer to distractors in real reading comprehension from examinations. Taking a reading comprehension article, a pair of question and its correct option as input, our goal is to generate several distractors which are somehow related to the answer, consistent with the semantic context of the question and have some trace in the article. We propose a hierarchical encoder-decoder framework with static and dynamic attention mechanisms to tackle this task. Specifically, the dynamic attention can combine sentence-level and word-level attention varying at each recurrent time step to generate a more readable sequence. The static attention is to modulate the dynamic attention not to focus on question irrelevant sentences or sentences which contribute to the correct option. Our proposed framework outperforms several strong baselines on the first prepared distractor generation dataset of real reading comprehension questions. For human evaluation, compared with those distractors generated by baselines, our generated distractors are more functional to confuse the annotators.
2308.16822
Arthur Leroy
Chunchao Ma, Arthur Leroy, Mauricio Alvarez
Latent Variable Multi-output Gaussian Processes for Hierarchical Datasets
29 pages
null
null
null
cs.LG cs.AI
http://creativecommons.org/licenses/by-nc-sa/4.0/
Multi-output Gaussian processes (MOGPs) have been introduced to deal with multiple tasks by exploiting the correlations between different outputs. Generally, MOGPs models assume a flat correlation structure between the outputs. However, such a formulation does not account for more elaborate relationships, for instance, if several replicates were observed for each output (which is a typical setting in biological experiments). This paper proposes an extension of MOGPs for hierarchical datasets (i.e. datasets for which the relationships between observations can be represented within a tree structure). Our model defines a tailored kernel function accounting for hierarchical structures in the data to capture different levels of correlations while leveraging the introduction of latent variables to express the underlying dependencies between outputs through a dedicated kernel. This latter feature is expected to significantly improve scalability as the number of tasks increases. An extensive experimental study involving both synthetic and real-world data from genomics and motion capture is proposed to support our claims.
[ { "created": "Thu, 31 Aug 2023 15:52:35 GMT", "version": "v1" } ]
2023-09-01
[ [ "Ma", "Chunchao", "" ], [ "Leroy", "Arthur", "" ], [ "Alvarez", "Mauricio", "" ] ]
Multi-output Gaussian processes (MOGPs) have been introduced to deal with multiple tasks by exploiting the correlations between different outputs. Generally, MOGPs models assume a flat correlation structure between the outputs. However, such a formulation does not account for more elaborate relationships, for instance, if several replicates were observed for each output (which is a typical setting in biological experiments). This paper proposes an extension of MOGPs for hierarchical datasets (i.e. datasets for which the relationships between observations can be represented within a tree structure). Our model defines a tailored kernel function accounting for hierarchical structures in the data to capture different levels of correlations while leveraging the introduction of latent variables to express the underlying dependencies between outputs through a dedicated kernel. This latter feature is expected to significantly improve scalability as the number of tasks increases. An extensive experimental study involving both synthetic and real-world data from genomics and motion capture is proposed to support our claims.
1909.03723
Marco Virgolin
Marco Virgolin, Ziyuan Wang, Tanja Alderliesten, Peter A. N. Bosman
Machine learning for automatic construction of pseudo-realistic pediatric abdominal phantoms
Currently submitted to SPIE Medical Imaging journal
null
null
null
cs.LG physics.med-ph stat.AP stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Machine Learning (ML) is proving extremely beneficial in many healthcare applications. In pediatric oncology, retrospective studies that investigate the relationship between treatment and late adverse effects still rely on simple heuristics. To assess the effects of radiation therapy, treatment plans are typically simulated on phantoms, i.e., virtual surrogates of patient anatomy. Currently, phantoms are built according to reasonable, yet simple, human-designed criteria. This often results in a lack of individualization. We present a novel approach that combines imaging and ML to build individualized phantoms automatically. Given the features of a patient treated historically (only 2D radiographs available), and a database of 3D Computed Tomography (CT) imaging with organ segmentations and relative patient features, our approach uses ML to predict how to assemble a patient-specific phantom automatically. Experiments on 60 abdominal CTs of pediatric patients show that our approach constructs significantly more representative phantoms than using current phantom building criteria, in terms of location and shape of the abdomen and of two considered organs, the liver and the spleen. Among several ML algorithms considered, the Gene-pool Optimal Mixing Evolutionary Algorithm for Genetic Programming (GP-GOMEA) is found to deliver the best performing models, which are, moreover, transparent and interpretable mathematical expressions.
[ { "created": "Mon, 9 Sep 2019 09:38:22 GMT", "version": "v1" } ]
2019-09-10
[ [ "Virgolin", "Marco", "" ], [ "Wang", "Ziyuan", "" ], [ "Alderliesten", "Tanja", "" ], [ "Bosman", "Peter A. N.", "" ] ]
Machine Learning (ML) is proving extremely beneficial in many healthcare applications. In pediatric oncology, retrospective studies that investigate the relationship between treatment and late adverse effects still rely on simple heuristics. To assess the effects of radiation therapy, treatment plans are typically simulated on phantoms, i.e., virtual surrogates of patient anatomy. Currently, phantoms are built according to reasonable, yet simple, human-designed criteria. This often results in a lack of individualization. We present a novel approach that combines imaging and ML to build individualized phantoms automatically. Given the features of a patient treated historically (only 2D radiographs available), and a database of 3D Computed Tomography (CT) imaging with organ segmentations and relative patient features, our approach uses ML to predict how to assemble a patient-specific phantom automatically. Experiments on 60 abdominal CTs of pediatric patients show that our approach constructs significantly more representative phantoms than using current phantom building criteria, in terms of location and shape of the abdomen and of two considered organs, the liver and the spleen. Among several ML algorithms considered, the Gene-pool Optimal Mixing Evolutionary Algorithm for Genetic Programming (GP-GOMEA) is found to deliver the best performing models, which are, moreover, transparent and interpretable mathematical expressions.
2004.07798
Neil Lutz
Jack H. Lutz, Neil Lutz, Elvira Mayordomo
Extending the Reach of the Point-to-Set Principle
null
null
null
null
cs.CC math.CA math.MG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The point-to-set principle of J. Lutz and N. Lutz (2018) has recently enabled the theory of computing to be used to answer open questions about fractal geometry in Euclidean spaces $\mathbb{R}^n$. These are classical questions, meaning that their statements do not involve computation or related aspects of logic. In this paper we extend the reach of the point-to-set principle from Euclidean spaces to arbitrary separable metric spaces $X$. We first extend two fractal dimensions--computability-theoretic versions of classical Hausdorff and packing dimensions that assign dimensions $\dim(x)$ and $\textrm{Dim}(x)$ to individual points $x\in X$--to arbitrary separable metric spaces and to arbitrary gauge families. Our first two main results then extend the point-to-set principle to arbitrary separable metric spaces and to a large class of gauge families. We demonstrate the power of our extended point-to-set principle by using it to prove new theorems about classical fractal dimensions in hyperspaces. (For a concrete computational example, the stages $E_0, E_1, E_2, \ldots$ used to construct a self-similar fractal $E$ in the plane are elements of the hyperspace of the plane, and they converge to $E$ in the hyperspace.) Our third main result, proven via our extended point-to-set principle, states that, under a wide variety of gauge families, the classical packing dimension agrees with the classical upper Minkowski dimension on all hyperspaces of compact sets. We use this theorem to give, for all sets $E$ that are analytic, i.e., $\mathbf{\Sigma}^1_1$, a tight bound on the packing dimension of the hyperspace of $E$ in terms of the packing dimension of $E$ itself.
[ { "created": "Thu, 16 Apr 2020 17:43:37 GMT", "version": "v1" }, { "created": "Tue, 6 Oct 2020 19:28:01 GMT", "version": "v2" }, { "created": "Sat, 13 Feb 2021 22:53:18 GMT", "version": "v3" } ]
2021-02-16
[ [ "Lutz", "Jack H.", "" ], [ "Lutz", "Neil", "" ], [ "Mayordomo", "Elvira", "" ] ]
The point-to-set principle of J. Lutz and N. Lutz (2018) has recently enabled the theory of computing to be used to answer open questions about fractal geometry in Euclidean spaces $\mathbb{R}^n$. These are classical questions, meaning that their statements do not involve computation or related aspects of logic. In this paper we extend the reach of the point-to-set principle from Euclidean spaces to arbitrary separable metric spaces $X$. We first extend two fractal dimensions--computability-theoretic versions of classical Hausdorff and packing dimensions that assign dimensions $\dim(x)$ and $\textrm{Dim}(x)$ to individual points $x\in X$--to arbitrary separable metric spaces and to arbitrary gauge families. Our first two main results then extend the point-to-set principle to arbitrary separable metric spaces and to a large class of gauge families. We demonstrate the power of our extended point-to-set principle by using it to prove new theorems about classical fractal dimensions in hyperspaces. (For a concrete computational example, the stages $E_0, E_1, E_2, \ldots$ used to construct a self-similar fractal $E$ in the plane are elements of the hyperspace of the plane, and they converge to $E$ in the hyperspace.) Our third main result, proven via our extended point-to-set principle, states that, under a wide variety of gauge families, the classical packing dimension agrees with the classical upper Minkowski dimension on all hyperspaces of compact sets. We use this theorem to give, for all sets $E$ that are analytic, i.e., $\mathbf{\Sigma}^1_1$, a tight bound on the packing dimension of the hyperspace of $E$ in terms of the packing dimension of $E$ itself.
1104.0919
Benjamin Burton
Benjamin A. Burton and Mathias Hiron
Locating regions in a sequence under density constraints
17 pages, 8 figures; v2: minor revisions, additional explanations; to appear in SIAM Journal on Computing
SIAM Journal on Computing 42 (2013), no. 3, 1201-1215
10.1137/110830605
null
cs.DS
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Several biological problems require the identification of regions in a sequence where some feature occurs within a target density range: examples including the location of GC-rich regions, identification of CpG islands, and sequence matching. Mathematically, this corresponds to searching a string of 0s and 1s for a substring whose relative proportion of 1s lies between given lower and upper bounds. We consider the algorithmic problem of locating the longest such substring, as well as other related problems (such as finding the shortest substring or a maximal set of disjoint substrings). For locating the longest such substring, we develop an algorithm that runs in O(n) time, improving upon the previous best-known O(n log n) result. For the related problems we develop O(n log log n) algorithms, again improving upon the best-known O(n log n) results. Practical testing verifies that our new algorithms enjoy significantly smaller time and memory footprints, and can process sequences that are orders of magnitude longer as a result.
[ { "created": "Tue, 5 Apr 2011 19:42:00 GMT", "version": "v1" }, { "created": "Tue, 2 Apr 2013 21:14:45 GMT", "version": "v2" } ]
2013-08-15
[ [ "Burton", "Benjamin A.", "" ], [ "Hiron", "Mathias", "" ] ]
Several biological problems require the identification of regions in a sequence where some feature occurs within a target density range: examples including the location of GC-rich regions, identification of CpG islands, and sequence matching. Mathematically, this corresponds to searching a string of 0s and 1s for a substring whose relative proportion of 1s lies between given lower and upper bounds. We consider the algorithmic problem of locating the longest such substring, as well as other related problems (such as finding the shortest substring or a maximal set of disjoint substrings). For locating the longest such substring, we develop an algorithm that runs in O(n) time, improving upon the previous best-known O(n log n) result. For the related problems we develop O(n log log n) algorithms, again improving upon the best-known O(n log n) results. Practical testing verifies that our new algorithms enjoy significantly smaller time and memory footprints, and can process sequences that are orders of magnitude longer as a result.
1902.06531
Yansong Gao Dr
Yansong Gao, Chang Xu, Derui Wang, Shiping Chen, Damith C.Ranasinghe, Surya Nepal
STRIP: A Defence Against Trojan Attacks on Deep Neural Networks
13 pages
In 2019 Annual Computer Security Applications Conference (ACSAC 19), December 9-13, 2019, San Juan, PR, USA. ACM, New York, NY, USA
null
null
cs.CR
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
A recent trojan attack on deep neural network (DNN) models is one insidious variant of data poisoning attacks. Trojan attacks exploit an effective backdoor created in a DNN model by leveraging the difficulty in interpretability of the learned model to misclassify any inputs signed with the attacker's chosen trojan trigger. Since the trojan trigger is a secret guarded and exploited by the attacker, detecting such trojan inputs is a challenge, especially at run-time when models are in active operation. This work builds STRong Intentional Perturbation (STRIP) based run-time trojan attack detection system and focuses on vision system. We intentionally perturb the incoming input, for instance by superimposing various image patterns, and observe the randomness of predicted classes for perturbed inputs from a given deployed model---malicious or benign. A low entropy in predicted classes violates the input-dependence property of a benign model and implies the presence of a malicious input---a characteristic of a trojaned input. The high efficacy of our method is validated through case studies on three popular and contrasting datasets: MNIST, CIFAR10 and GTSRB. We achieve an overall false acceptance rate (FAR) of less than 1%, given a preset false rejection rate (FRR) of 1%, for different types of triggers. Using CIFAR10 and GTSRB, we have empirically achieved result of 0% for both FRR and FAR. We have also evaluated STRIP robustness against a number of trojan attack variants and adaptive attacks.
[ { "created": "Mon, 18 Feb 2019 11:49:33 GMT", "version": "v1" }, { "created": "Fri, 17 Jan 2020 03:27:26 GMT", "version": "v2" } ]
2020-01-20
[ [ "Gao", "Yansong", "" ], [ "Xu", "Chang", "" ], [ "Wang", "Derui", "" ], [ "Chen", "Shiping", "" ], [ "Ranasinghe", "Damith C.", "" ], [ "Nepal", "Surya", "" ] ]
A recent trojan attack on deep neural network (DNN) models is one insidious variant of data poisoning attacks. Trojan attacks exploit an effective backdoor created in a DNN model by leveraging the difficulty in interpretability of the learned model to misclassify any inputs signed with the attacker's chosen trojan trigger. Since the trojan trigger is a secret guarded and exploited by the attacker, detecting such trojan inputs is a challenge, especially at run-time when models are in active operation. This work builds STRong Intentional Perturbation (STRIP) based run-time trojan attack detection system and focuses on vision system. We intentionally perturb the incoming input, for instance by superimposing various image patterns, and observe the randomness of predicted classes for perturbed inputs from a given deployed model---malicious or benign. A low entropy in predicted classes violates the input-dependence property of a benign model and implies the presence of a malicious input---a characteristic of a trojaned input. The high efficacy of our method is validated through case studies on three popular and contrasting datasets: MNIST, CIFAR10 and GTSRB. We achieve an overall false acceptance rate (FAR) of less than 1%, given a preset false rejection rate (FRR) of 1%, for different types of triggers. Using CIFAR10 and GTSRB, we have empirically achieved result of 0% for both FRR and FAR. We have also evaluated STRIP robustness against a number of trojan attack variants and adaptive attacks.
1905.04849
Cai Shaofeng
Shaofeng Cai, Yao Shu, Wei Wang, Beng Chin Ooi
Dynamic Routing Networks
10 pages, 3 figures, 3 tables
null
null
null
cs.LG cs.CV stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The deployment of deep neural networks in real-world applications is mostly restricted by their high inference costs. Extensive efforts have been made to improve the accuracy with expert-designed or algorithm-searched architectures. However, the incremental improvement is typically achieved with increasingly more expensive models that only a small portion of input instances really need. Inference with a static architecture that processes all input instances via the same transformation would thus incur unnecessary computational costs. Therefore, customizing the model capacity in an instance-aware manner is much needed for higher inference efficiency. In this paper, we propose Dynamic Routing Networks (DRNets), which support efficient instance-aware inference by routing the input instance to only necessary transformation branches selected from a candidate set of branches for each connection between transformation nodes. The branch selection is dynamically determined via the corresponding branch importance weights, which are first generated from lightweight hypernetworks (RouterNets) and then recalibrated with Gumbel-Softmax before the selection. Extensive experiments show that DRNets can reduce a substantial amount of parameter size and FLOPs during inference with prediction performance comparable to state-of-the-art architectures.
[ { "created": "Mon, 13 May 2019 03:45:42 GMT", "version": "v1" }, { "created": "Thu, 23 May 2019 16:45:18 GMT", "version": "v2" }, { "created": "Thu, 24 Oct 2019 04:47:36 GMT", "version": "v3" }, { "created": "Tue, 28 Jul 2020 16:26:29 GMT", "version": "v4" }, { "created": "Sun, 8 Nov 2020 13:11:45 GMT", "version": "v5" } ]
2020-11-10
[ [ "Cai", "Shaofeng", "" ], [ "Shu", "Yao", "" ], [ "Wang", "Wei", "" ], [ "Ooi", "Beng Chin", "" ] ]
The deployment of deep neural networks in real-world applications is mostly restricted by their high inference costs. Extensive efforts have been made to improve the accuracy with expert-designed or algorithm-searched architectures. However, the incremental improvement is typically achieved with increasingly more expensive models that only a small portion of input instances really need. Inference with a static architecture that processes all input instances via the same transformation would thus incur unnecessary computational costs. Therefore, customizing the model capacity in an instance-aware manner is much needed for higher inference efficiency. In this paper, we propose Dynamic Routing Networks (DRNets), which support efficient instance-aware inference by routing the input instance to only necessary transformation branches selected from a candidate set of branches for each connection between transformation nodes. The branch selection is dynamically determined via the corresponding branch importance weights, which are first generated from lightweight hypernetworks (RouterNets) and then recalibrated with Gumbel-Softmax before the selection. Extensive experiments show that DRNets can reduce a substantial amount of parameter size and FLOPs during inference with prediction performance comparable to state-of-the-art architectures.
1405.4100
Cristian Prisacariu
Cristian Prisacariu
Higher Dimensional Modal Logic
null
null
null
null
cs.LO
http://creativecommons.org/licenses/by-nc-sa/3.0/
Higher dimensional automata (HDA) are a model of concurrency that can express most of the traditional partial order models like Mazurkiewicz traces, pomsets, event structures, or Petri nets. Modal logics, interpreted over Kripke structures, are the logics for reasoning about sequential behavior and interleaved concurrency. Modal logic is a well behaved subset of first-order logic; many variants of modal logic are decidable. However, there are no modal-like logics for the more expressive HDA models. In this paper we introduce and investigate a modal logic over HDAs which incorporates two modalities for reasoning about "during" and "after". We prove that this general higher dimensional modal logic (HDML) is decidable and we define an axiomatic system for it. We also show how, when the HDA model is restricted to Kripke structures, a syntactic restriction of HDML becomes the standard modal logic. Then we isolate the class of HDAs that encode Mazurkiewicz traces and show how HDML, with natural definitions of corresponding Until operators, can be restricted to LTrL (the linear time temporal logic over Mazurkiewicz traces) or the branching time ISTL. We also study the expressiveness of the basic HDML language wrt. bisimulations and conclude that HDML captures the split-bisimulation.
[ { "created": "Fri, 16 May 2014 09:11:38 GMT", "version": "v1" } ]
2014-05-19
[ [ "Prisacariu", "Cristian", "" ] ]
Higher dimensional automata (HDA) are a model of concurrency that can express most of the traditional partial order models like Mazurkiewicz traces, pomsets, event structures, or Petri nets. Modal logics, interpreted over Kripke structures, are the logics for reasoning about sequential behavior and interleaved concurrency. Modal logic is a well behaved subset of first-order logic; many variants of modal logic are decidable. However, there are no modal-like logics for the more expressive HDA models. In this paper we introduce and investigate a modal logic over HDAs which incorporates two modalities for reasoning about "during" and "after". We prove that this general higher dimensional modal logic (HDML) is decidable and we define an axiomatic system for it. We also show how, when the HDA model is restricted to Kripke structures, a syntactic restriction of HDML becomes the standard modal logic. Then we isolate the class of HDAs that encode Mazurkiewicz traces and show how HDML, with natural definitions of corresponding Until operators, can be restricted to LTrL (the linear time temporal logic over Mazurkiewicz traces) or the branching time ISTL. We also study the expressiveness of the basic HDML language wrt. bisimulations and conclude that HDML captures the split-bisimulation.
1704.08045
Quynh Nguyen
Quynh Nguyen and Matthias Hein
The loss surface of deep and wide neural networks
ICML 2017. Main results now hold for larger classes of loss functions
null
null
null
cs.LG cs.AI cs.CV cs.NE stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
While the optimization problem behind deep neural networks is highly non-convex, it is frequently observed in practice that training deep networks seems possible without getting stuck in suboptimal points. It has been argued that this is the case as all local minima are close to being globally optimal. We show that this is (almost) true, in fact almost all local minima are globally optimal, for a fully connected network with squared loss and analytic activation function given that the number of hidden units of one layer of the network is larger than the number of training points and the network structure from this layer on is pyramidal.
[ { "created": "Wed, 26 Apr 2017 10:24:54 GMT", "version": "v1" }, { "created": "Mon, 12 Jun 2017 19:43:39 GMT", "version": "v2" } ]
2017-06-14
[ [ "Nguyen", "Quynh", "" ], [ "Hein", "Matthias", "" ] ]
While the optimization problem behind deep neural networks is highly non-convex, it is frequently observed in practice that training deep networks seems possible without getting stuck in suboptimal points. It has been argued that this is the case as all local minima are close to being globally optimal. We show that this is (almost) true, in fact almost all local minima are globally optimal, for a fully connected network with squared loss and analytic activation function given that the number of hidden units of one layer of the network is larger than the number of training points and the network structure from this layer on is pyramidal.
1911.09845
Jun Gao
Jun Gao, Wei Bi, Xiaojiang Liu, Junhui Li, Guodong Zhou, Shuming Shi
A Discrete CVAE for Response Generation on Short-Text Conversation
Accepted for publication at EMNLP 2019
null
null
null
cs.CL
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Neural conversation models such as encoder-decoder models are easy to generate bland and generic responses. Some researchers propose to use the conditional variational autoencoder(CVAE) which maximizes the lower bound on the conditional log-likelihood on a continuous latent variable. With different sampled la-tent variables, the model is expected to generate diverse responses. Although the CVAE-based models have shown tremendous potential, their improvement of generating high-quality responses is still unsatisfactory. In this paper, we introduce a discrete latent variable with an explicit semantic meaning to improve the CVAE on short-text conversation. A major advantage of our model is that we can exploit the semantic distance between the latent variables to maintain good diversity between the sampled latent variables. Accordingly, we pro-pose a two-stage sampling approach to enable efficient diverse variable selection from a large latent space assumed in the short-text conversation task. Experimental results indicate that our model outperforms various kinds of generation models under both automatic and human evaluations and generates more diverse and in-formative responses.
[ { "created": "Fri, 22 Nov 2019 04:14:31 GMT", "version": "v1" } ]
2019-11-25
[ [ "Gao", "Jun", "" ], [ "Bi", "Wei", "" ], [ "Liu", "Xiaojiang", "" ], [ "Li", "Junhui", "" ], [ "Zhou", "Guodong", "" ], [ "Shi", "Shuming", "" ] ]
Neural conversation models such as encoder-decoder models are easy to generate bland and generic responses. Some researchers propose to use the conditional variational autoencoder(CVAE) which maximizes the lower bound on the conditional log-likelihood on a continuous latent variable. With different sampled la-tent variables, the model is expected to generate diverse responses. Although the CVAE-based models have shown tremendous potential, their improvement of generating high-quality responses is still unsatisfactory. In this paper, we introduce a discrete latent variable with an explicit semantic meaning to improve the CVAE on short-text conversation. A major advantage of our model is that we can exploit the semantic distance between the latent variables to maintain good diversity between the sampled latent variables. Accordingly, we pro-pose a two-stage sampling approach to enable efficient diverse variable selection from a large latent space assumed in the short-text conversation task. Experimental results indicate that our model outperforms various kinds of generation models under both automatic and human evaluations and generates more diverse and in-formative responses.
1701.04551
Mingchao Yu
Mingchao Yu, Parastoo Sadeghi
Approximating Throughput and Packet Decoding Delay in Linear Network Coded Wireless Broadcast
5 pages, 2 figures, 1 table, submitted to ISIT2017
null
null
null
cs.IT math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In this paper, we study a wireless packet broadcast system that uses linear network coding (LNC) to help receivers recover data packets that are missing due to packet erasures. We study two intertwined performance metrics, namely throughput and average packet decoding delay (APDD) and establish strong/weak approximation relations based on whether the approximation holds for the performance of every receiver (strong) or for the average performance across all receivers (weak). We prove an equivalence between strong throughput approximation and strong APDD approximation. We prove that throughput-optimal LNC techniques can strongly approximate APDD, and partition-based LNC techniques may weakly approximate throughput. We also prove that memoryless LNC techniques, including instantly decodable network coding techniques, are not strong throughput and APDD approximation nor weak throughput approximation techniques.
[ { "created": "Tue, 17 Jan 2017 07:26:00 GMT", "version": "v1" } ]
2017-01-18
[ [ "Yu", "Mingchao", "" ], [ "Sadeghi", "Parastoo", "" ] ]
In this paper, we study a wireless packet broadcast system that uses linear network coding (LNC) to help receivers recover data packets that are missing due to packet erasures. We study two intertwined performance metrics, namely throughput and average packet decoding delay (APDD) and establish strong/weak approximation relations based on whether the approximation holds for the performance of every receiver (strong) or for the average performance across all receivers (weak). We prove an equivalence between strong throughput approximation and strong APDD approximation. We prove that throughput-optimal LNC techniques can strongly approximate APDD, and partition-based LNC techniques may weakly approximate throughput. We also prove that memoryless LNC techniques, including instantly decodable network coding techniques, are not strong throughput and APDD approximation nor weak throughput approximation techniques.
2404.00367
Yan Zhang Main
Bin Wang, Yan Zhang, Yan Ma, Yaohui Jin, Yanyan Xu
SA-LSPL:Sequence-Aware Long- and Short- Term Preference Learning for next POI recommendation
null
null
null
null
cs.CY
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The next Point of Interest (POI) recommendation aims to recommend the next POI for users at a specific time. As users' check-in records can be viewed as a long sequence, methods based on Recurrent Neural Networks (RNNs) have recently shown good applicability to this task. However, existing methods often struggle to fully explore the spatio-temporal correlations and dependencies at the sequence level, and don't take full consideration for various factors influencing users' preferences. To address these issues, we propose a novel approach called Sequence-Aware Long- and Short-Term Preference Learning (SA-LSPL) for next-POI recommendation. We combine various information features to effectively model users' long-term preferences. Specifically, our proposed model uses a multi-modal embedding module to embed diverse check-in details, taking into account both user's personalized preferences and social influences comprehensively. Additionally, we consider explicit spatio-temporal correlations at the sequence level and implicit sequence dependencies. Furthermore, SA-LSPL learns the spatio-temporal correlations of consecutive and non-consecutive visits in the current check-in sequence, as well as transition dependencies between categories, providing a comprehensive capture of user's short-term preferences. Extensive experiments on two real-world datasets demonstrate the superiority of SA-LSPL over state-of-the-art baseline methods.
[ { "created": "Sat, 30 Mar 2024 13:40:25 GMT", "version": "v1" } ]
2024-04-02
[ [ "Wang", "Bin", "" ], [ "Zhang", "Yan", "" ], [ "Ma", "Yan", "" ], [ "Jin", "Yaohui", "" ], [ "Xu", "Yanyan", "" ] ]
The next Point of Interest (POI) recommendation aims to recommend the next POI for users at a specific time. As users' check-in records can be viewed as a long sequence, methods based on Recurrent Neural Networks (RNNs) have recently shown good applicability to this task. However, existing methods often struggle to fully explore the spatio-temporal correlations and dependencies at the sequence level, and don't take full consideration for various factors influencing users' preferences. To address these issues, we propose a novel approach called Sequence-Aware Long- and Short-Term Preference Learning (SA-LSPL) for next-POI recommendation. We combine various information features to effectively model users' long-term preferences. Specifically, our proposed model uses a multi-modal embedding module to embed diverse check-in details, taking into account both user's personalized preferences and social influences comprehensively. Additionally, we consider explicit spatio-temporal correlations at the sequence level and implicit sequence dependencies. Furthermore, SA-LSPL learns the spatio-temporal correlations of consecutive and non-consecutive visits in the current check-in sequence, as well as transition dependencies between categories, providing a comprehensive capture of user's short-term preferences. Extensive experiments on two real-world datasets demonstrate the superiority of SA-LSPL over state-of-the-art baseline methods.
1303.5759
Hong Xu
Hong Xu
An Efficient Implementation of Belief Function Propagation
Appears in Proceedings of the Seventh Conference on Uncertainty in Artificial Intelligence (UAI1991)
null
null
UAI-P-1991-PG-425-432
cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The local computation technique (Shafer et al. 1987, Shafer and Shenoy 1988, Shenoy and Shafer 1986) is used for propagating belief functions in so called a Markov Tree. In this paper, we describe an efficient implementation of belief function propagation on the basis of the local computation technique. The presented method avoids all the redundant computations in the propagation process, and so makes the computational complexity decrease with respect to other existing implementations (Hsia and Shenoy 1989, Zarley et al. 1988). We also give a combined algorithm for both propagation and re-propagation which makes the re-propagation process more efficient when one or more of the prior belief functions is changed.
[ { "created": "Wed, 20 Mar 2013 15:34:07 GMT", "version": "v1" } ]
2013-03-26
[ [ "Xu", "Hong", "" ] ]
The local computation technique (Shafer et al. 1987, Shafer and Shenoy 1988, Shenoy and Shafer 1986) is used for propagating belief functions in so called a Markov Tree. In this paper, we describe an efficient implementation of belief function propagation on the basis of the local computation technique. The presented method avoids all the redundant computations in the propagation process, and so makes the computational complexity decrease with respect to other existing implementations (Hsia and Shenoy 1989, Zarley et al. 1988). We also give a combined algorithm for both propagation and re-propagation which makes the re-propagation process more efficient when one or more of the prior belief functions is changed.
2204.07936
Jessica Leu
Jessica Leu, Yujiao Cheng, Changliu Liu, Masayoshi Tomizuka
Robust Task Planning for Assembly Lines with Human-Robot Collaboration
null
null
null
null
cs.RO
http://creativecommons.org/licenses/by-nc-nd/4.0/
Efficient and robust task planning for a human-robot collaboration (HRC) system remains challenging. The human-aware task planner needs to assign jobs to both robots and human workers so that they can work collaboratively to achieve better time efficiency. However, the complexity of the tasks and the stochastic nature of the human collaborators bring challenges to such task planning. To reduce the complexity of the planning problem, we utilize the hierarchical task model, which explicitly captures the sequential and parallel relationships of the task. We model human movements with the sigma-lognormal functions to account for human-induced uncertainties. A human action model adaptation scheme is applied during run-time, and it provides a measure for modeling the human-induced uncertainties. We propose a sampling-based method to estimate human job completion time uncertainties. Next, we propose a robust task planner, which formulates the planning problem as a robust optimization problem by considering the task structure and the uncertainties. We conduct simulations of a robot arm collaborating with a human worker in an electronics assembly setting. The results show that our proposed planner can reduce task completion time when human-induced uncertainties occur compared to the baseline planner.
[ { "created": "Sun, 17 Apr 2022 05:54:01 GMT", "version": "v1" } ]
2022-04-19
[ [ "Leu", "Jessica", "" ], [ "Cheng", "Yujiao", "" ], [ "Liu", "Changliu", "" ], [ "Tomizuka", "Masayoshi", "" ] ]
Efficient and robust task planning for a human-robot collaboration (HRC) system remains challenging. The human-aware task planner needs to assign jobs to both robots and human workers so that they can work collaboratively to achieve better time efficiency. However, the complexity of the tasks and the stochastic nature of the human collaborators bring challenges to such task planning. To reduce the complexity of the planning problem, we utilize the hierarchical task model, which explicitly captures the sequential and parallel relationships of the task. We model human movements with the sigma-lognormal functions to account for human-induced uncertainties. A human action model adaptation scheme is applied during run-time, and it provides a measure for modeling the human-induced uncertainties. We propose a sampling-based method to estimate human job completion time uncertainties. Next, we propose a robust task planner, which formulates the planning problem as a robust optimization problem by considering the task structure and the uncertainties. We conduct simulations of a robot arm collaborating with a human worker in an electronics assembly setting. The results show that our proposed planner can reduce task completion time when human-induced uncertainties occur compared to the baseline planner.
2104.10715
Rishab Khincha
Utkarsh Sarawgi, Rishab Khincha, Wazeer Zulfikar, Satrajit Ghosh, Pattie Maes
Uncertainty-Aware Boosted Ensembling in Multi-Modal Settings
Accepted at IJCNN 2021, to appear in IEEE proceedings. Equal contributions from US, RK and WZ
null
null
null
cs.LG cs.AI
http://creativecommons.org/licenses/by/4.0/
Reliability of machine learning (ML) systems is crucial in safety-critical applications such as healthcare, and uncertainty estimation is a widely researched method to highlight the confidence of ML systems in deployment. Sequential and parallel ensemble techniques have shown improved performance of ML systems in multi-modal settings by leveraging the feature sets together. We propose an uncertainty-aware boosting technique for multi-modal ensembling in order to focus on the data points with higher associated uncertainty estimates, rather than the ones with higher loss values. We evaluate this method on healthcare tasks related to Dementia and Parkinson's disease which involve real-world multi-modal speech and text data, wherein our method shows an improved performance. Additional analysis suggests that introducing uncertainty-awareness into the boosted ensembles decreases the overall entropy of the system, making it more robust to heteroscedasticity in the data, as well as better calibrating each of the modalities along with high quality prediction intervals. We open-source our entire codebase at https://github.com/usarawgi911/Uncertainty-aware-boosting
[ { "created": "Wed, 21 Apr 2021 18:28:13 GMT", "version": "v1" } ]
2021-04-23
[ [ "Sarawgi", "Utkarsh", "" ], [ "Khincha", "Rishab", "" ], [ "Zulfikar", "Wazeer", "" ], [ "Ghosh", "Satrajit", "" ], [ "Maes", "Pattie", "" ] ]
Reliability of machine learning (ML) systems is crucial in safety-critical applications such as healthcare, and uncertainty estimation is a widely researched method to highlight the confidence of ML systems in deployment. Sequential and parallel ensemble techniques have shown improved performance of ML systems in multi-modal settings by leveraging the feature sets together. We propose an uncertainty-aware boosting technique for multi-modal ensembling in order to focus on the data points with higher associated uncertainty estimates, rather than the ones with higher loss values. We evaluate this method on healthcare tasks related to Dementia and Parkinson's disease which involve real-world multi-modal speech and text data, wherein our method shows an improved performance. Additional analysis suggests that introducing uncertainty-awareness into the boosted ensembles decreases the overall entropy of the system, making it more robust to heteroscedasticity in the data, as well as better calibrating each of the modalities along with high quality prediction intervals. We open-source our entire codebase at https://github.com/usarawgi911/Uncertainty-aware-boosting
2005.03210
Dylan Losey
Hong Jun Jeon, Dylan P. Losey, Dorsa Sadigh
Shared Autonomy with Learned Latent Actions
null
null
null
null
cs.RO
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Assistive robots enable people with disabilities to conduct everyday tasks on their own. However, these tasks can be complex, containing both coarse reaching motions and fine-grained manipulation. For example, when eating, not only does one need to move to the correct food item, but they must also precisely manipulate the food in different ways (e.g., cutting, stabbing, scooping). Shared autonomy methods make robot teleoperation safer and more precise by arbitrating user inputs with robot controls. However, these works have focused mainly on the high-level task of reaching a goal from a discrete set, while largely ignoring manipulation of objects at that goal. Meanwhile, dimensionality reduction techniques for teleoperation map useful high-dimensional robot actions into an intuitive low-dimensional controller, but it is unclear if these methods can achieve the requisite precision for tasks like eating. Our insight is that---by combining intuitive embeddings from learned latent actions with robotic assistance from shared autonomy---we can enable precise assistive manipulation. In this work, we adopt learned latent actions for shared autonomy by proposing a new model structure that changes the meaning of the human's input based on the robot's confidence of the goal. We show convergence bounds on the robot's distance to the most likely goal, and develop a training procedure to learn a controller that is able to move between goals even in the presence of shared autonomy. We evaluate our method in simulations and an eating user study. See videos of our experiments here: https://youtu.be/7BouKojzVyk
[ { "created": "Thu, 7 May 2020 02:39:56 GMT", "version": "v1" }, { "created": "Mon, 11 May 2020 16:50:25 GMT", "version": "v2" } ]
2020-05-12
[ [ "Jeon", "Hong Jun", "" ], [ "Losey", "Dylan P.", "" ], [ "Sadigh", "Dorsa", "" ] ]
Assistive robots enable people with disabilities to conduct everyday tasks on their own. However, these tasks can be complex, containing both coarse reaching motions and fine-grained manipulation. For example, when eating, not only does one need to move to the correct food item, but they must also precisely manipulate the food in different ways (e.g., cutting, stabbing, scooping). Shared autonomy methods make robot teleoperation safer and more precise by arbitrating user inputs with robot controls. However, these works have focused mainly on the high-level task of reaching a goal from a discrete set, while largely ignoring manipulation of objects at that goal. Meanwhile, dimensionality reduction techniques for teleoperation map useful high-dimensional robot actions into an intuitive low-dimensional controller, but it is unclear if these methods can achieve the requisite precision for tasks like eating. Our insight is that---by combining intuitive embeddings from learned latent actions with robotic assistance from shared autonomy---we can enable precise assistive manipulation. In this work, we adopt learned latent actions for shared autonomy by proposing a new model structure that changes the meaning of the human's input based on the robot's confidence of the goal. We show convergence bounds on the robot's distance to the most likely goal, and develop a training procedure to learn a controller that is able to move between goals even in the presence of shared autonomy. We evaluate our method in simulations and an eating user study. See videos of our experiments here: https://youtu.be/7BouKojzVyk
cs/0605060
Rajiv Ranjan Mr.
Rajiv Ranjan, Aaron Harwood and Rajkumar Buyya
A Case for Cooperative and Incentive-Based Coupling of Distributed Clusters
22 pages, extended version of the conference paper published at IEEE Cluster'05, Boston, MA
In Proceedings of the 7th IEEE International Conference on Cluster Computing (Cluster 2005), IEEE Computer Society Press, September 27 - 30, 2005, Boston, Massachusetts, USA.
10.1109/CLUSTR.2005.347038
null
cs.DC
null
Research interest in Grid computing has grown significantly over the past five years. Management of distributed resources is one of the key issues in Grid computing. Central to management of resources is the effectiveness of resource allocation as it determines the overall utility of the system. The current approaches to superscheduling in a grid environment are non-coordinated since application level schedulers or brokers make scheduling decisions independently of the others in the system. Clearly, this can exacerbate the load sharing and utilization problems of distributed resources due to suboptimal schedules that are likely to occur. To overcome these limitations, we propose a mechanism for coordinated sharing of distributed clusters based on computational economy. The resulting environment, called \emph{Grid-Federation}, allows the transparent use of resources from the federation when local resources are insufficient to meet its users' requirements. The use of computational economy methodology in coordinating resource allocation not only facilitates the QoS based scheduling, but also enhances utility delivered by resources.
[ { "created": "Mon, 15 May 2006 10:21:22 GMT", "version": "v1" } ]
2016-11-18
[ [ "Ranjan", "Rajiv", "" ], [ "Harwood", "Aaron", "" ], [ "Buyya", "Rajkumar", "" ] ]
Research interest in Grid computing has grown significantly over the past five years. Management of distributed resources is one of the key issues in Grid computing. Central to management of resources is the effectiveness of resource allocation as it determines the overall utility of the system. The current approaches to superscheduling in a grid environment are non-coordinated since application level schedulers or brokers make scheduling decisions independently of the others in the system. Clearly, this can exacerbate the load sharing and utilization problems of distributed resources due to suboptimal schedules that are likely to occur. To overcome these limitations, we propose a mechanism for coordinated sharing of distributed clusters based on computational economy. The resulting environment, called \emph{Grid-Federation}, allows the transparent use of resources from the federation when local resources are insufficient to meet its users' requirements. The use of computational economy methodology in coordinating resource allocation not only facilitates the QoS based scheduling, but also enhances utility delivered by resources.
1507.06462
Matteo Sammartino
Roberto Bruni, Ugo Montanari, Matteo Sammartino
A coalgebraic semantics for causality in Petri nets
Accepted by Journal of Logical and Algebraic Methods in Programming
null
10.1016/j.jlamp.2015.07.003
null
cs.LO
http://creativecommons.org/licenses/by/4.0/
In this paper we revisit some pioneering efforts to equip Petri nets with compact operational models for expressing causality. The models we propose have a bisimilarity relation and a minimal representative for each equivalence class, and they can be fully explained as coalgebras on a presheaf category on an index category of partial orders. First, we provide a set-theoretic model in the form of a a causal case graph, that is a labeled transition system where states and transitions represent markings and firings of the net, respectively, and are equipped with causal information. Most importantly, each state has a poset representing causal dependencies among past events. Our first result shows the correspondence with behavior structure semantics as proposed by Trakhtenbrot and Rabinovich. Causal case graphs may be infinitely-branching and have infinitely many states, but we show how they can be refined to get an equivalent finitely-branching model. In it, states are equipped with symmetries, which are essential for the existence of a minimal, often finite-state, model. The next step is constructing a coalgebraic model. We exploit the fact that events can be represented as names, and event generation as name generation. Thus we can apply the Fiore-Turi framework: we model causal relations as a suitable category of posets with action labels, and generation of new events with causal dependencies as an endofunctor on this category. Then we define a well-behaved category of coalgebras. Our coalgebraic model is still infinite-state, but we exploit the equivalence between coalgebras over a class of presheaves and History Dependent automata to derive a compact representation, which is equivalent to our set-theoretical compact model. Remarkably, state reduction is automatically performed along the equivalence.
[ { "created": "Thu, 23 Jul 2015 12:08:22 GMT", "version": "v1" } ]
2015-07-24
[ [ "Bruni", "Roberto", "" ], [ "Montanari", "Ugo", "" ], [ "Sammartino", "Matteo", "" ] ]
In this paper we revisit some pioneering efforts to equip Petri nets with compact operational models for expressing causality. The models we propose have a bisimilarity relation and a minimal representative for each equivalence class, and they can be fully explained as coalgebras on a presheaf category on an index category of partial orders. First, we provide a set-theoretic model in the form of a a causal case graph, that is a labeled transition system where states and transitions represent markings and firings of the net, respectively, and are equipped with causal information. Most importantly, each state has a poset representing causal dependencies among past events. Our first result shows the correspondence with behavior structure semantics as proposed by Trakhtenbrot and Rabinovich. Causal case graphs may be infinitely-branching and have infinitely many states, but we show how they can be refined to get an equivalent finitely-branching model. In it, states are equipped with symmetries, which are essential for the existence of a minimal, often finite-state, model. The next step is constructing a coalgebraic model. We exploit the fact that events can be represented as names, and event generation as name generation. Thus we can apply the Fiore-Turi framework: we model causal relations as a suitable category of posets with action labels, and generation of new events with causal dependencies as an endofunctor on this category. Then we define a well-behaved category of coalgebras. Our coalgebraic model is still infinite-state, but we exploit the equivalence between coalgebras over a class of presheaves and History Dependent automata to derive a compact representation, which is equivalent to our set-theoretical compact model. Remarkably, state reduction is automatically performed along the equivalence.
2205.01643
Jinze Yu
Jinze Yu, Jiaming Liu, Xiaobao Wei, Haoyi Zhou, Yohei Nakata, Denis Gudovskiy, Tomoyuki Okuno, Jianxin Li, Kurt Keutzer, Shanghang Zhang
MTTrans: Cross-Domain Object Detection with Mean-Teacher Transformer
Accepted by ECCV 2022
null
null
null
cs.CV
http://creativecommons.org/licenses/by/4.0/
Recently, DEtection TRansformer (DETR), an end-to-end object detection pipeline, has achieved promising performance. However, it requires large-scale labeled data and suffers from domain shift, especially when no labeled data is available in the target domain. To solve this problem, we propose an end-to-end cross-domain detection Transformer based on the mean teacher framework, MTTrans, which can fully exploit unlabeled target domain data in object detection training and transfer knowledge between domains via pseudo labels. We further propose the comprehensive multi-level feature alignment to improve the pseudo labels generated by the mean teacher framework taking advantage of the cross-scale self-attention mechanism in Deformable DETR. Image and object features are aligned at the local, global, and instance levels with domain query-based feature alignment (DQFA), bi-level graph-based prototype alignment (BGPA), and token-wise image feature alignment (TIFA). On the other hand, the unlabeled target domain data pseudo-labeled and available for the object detection training by the mean teacher framework can lead to better feature extraction and alignment. Thus, the mean teacher framework and the comprehensive multi-level feature alignment can be optimized iteratively and mutually based on the architecture of Transformers. Extensive experiments demonstrate that our proposed method achieves state-of-the-art performance in three domain adaptation scenarios, especially the result of Sim10k to Cityscapes scenario is remarkably improved from 52.6 mAP to 57.9 mAP. Code will be released.
[ { "created": "Tue, 3 May 2022 17:11:55 GMT", "version": "v1" }, { "created": "Tue, 16 Aug 2022 09:55:23 GMT", "version": "v2" } ]
2022-08-17
[ [ "Yu", "Jinze", "" ], [ "Liu", "Jiaming", "" ], [ "Wei", "Xiaobao", "" ], [ "Zhou", "Haoyi", "" ], [ "Nakata", "Yohei", "" ], [ "Gudovskiy", "Denis", "" ], [ "Okuno", "Tomoyuki", "" ], [ "Li", "Jianxin", "" ], [ "Keutzer", "Kurt", "" ], [ "Zhang", "Shanghang", "" ] ]
Recently, DEtection TRansformer (DETR), an end-to-end object detection pipeline, has achieved promising performance. However, it requires large-scale labeled data and suffers from domain shift, especially when no labeled data is available in the target domain. To solve this problem, we propose an end-to-end cross-domain detection Transformer based on the mean teacher framework, MTTrans, which can fully exploit unlabeled target domain data in object detection training and transfer knowledge between domains via pseudo labels. We further propose the comprehensive multi-level feature alignment to improve the pseudo labels generated by the mean teacher framework taking advantage of the cross-scale self-attention mechanism in Deformable DETR. Image and object features are aligned at the local, global, and instance levels with domain query-based feature alignment (DQFA), bi-level graph-based prototype alignment (BGPA), and token-wise image feature alignment (TIFA). On the other hand, the unlabeled target domain data pseudo-labeled and available for the object detection training by the mean teacher framework can lead to better feature extraction and alignment. Thus, the mean teacher framework and the comprehensive multi-level feature alignment can be optimized iteratively and mutually based on the architecture of Transformers. Extensive experiments demonstrate that our proposed method achieves state-of-the-art performance in three domain adaptation scenarios, especially the result of Sim10k to Cityscapes scenario is remarkably improved from 52.6 mAP to 57.9 mAP. Code will be released.
2006.01304
Kyungmi Lee
Kyungmi Lee, Anantha P. Chandrakasan
Rethinking Empirical Evaluation of Adversarial Robustness Using First-Order Attack Methods
null
null
null
null
cs.LG cs.AI stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We identify three common cases that lead to overestimation of adversarial accuracy against bounded first-order attack methods, which is popularly used as a proxy for adversarial robustness in empirical studies. For each case, we propose compensation methods that either address sources of inaccurate gradient computation, such as numerical instability near zero and non-differentiability, or reduce the total number of back-propagations for iterative attacks by approximating second-order information. These compensation methods can be combined with existing attack methods for a more precise empirical evaluation metric. We illustrate the impact of these three cases with examples of practical interest, such as benchmarking model capacity and regularization techniques for robustness. Overall, our work shows that overestimated adversarial accuracy that is not indicative of robustness is prevalent even for conventionally trained deep neural networks, and highlights cautions of using empirical evaluation without guaranteed bounds.
[ { "created": "Mon, 1 Jun 2020 22:55:09 GMT", "version": "v1" } ]
2020-06-03
[ [ "Lee", "Kyungmi", "" ], [ "Chandrakasan", "Anantha P.", "" ] ]
We identify three common cases that lead to overestimation of adversarial accuracy against bounded first-order attack methods, which is popularly used as a proxy for adversarial robustness in empirical studies. For each case, we propose compensation methods that either address sources of inaccurate gradient computation, such as numerical instability near zero and non-differentiability, or reduce the total number of back-propagations for iterative attacks by approximating second-order information. These compensation methods can be combined with existing attack methods for a more precise empirical evaluation metric. We illustrate the impact of these three cases with examples of practical interest, such as benchmarking model capacity and regularization techniques for robustness. Overall, our work shows that overestimated adversarial accuracy that is not indicative of robustness is prevalent even for conventionally trained deep neural networks, and highlights cautions of using empirical evaluation without guaranteed bounds.
1511.03518
Ya-Hui An
Ya-Hui An, Qiang Dong, Chong-Jing Sun, Da-Cheng Nie and Yan Fu
Diffusion-like recommendation with enhanced similarity of objects
null
Physica A: Statistical Mechanics and its Applications 461 (2016) 708-715
10.1016/j.physa.2016.06.027
null
cs.IR cs.SI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In last decades, diversity and accuracy have been regarded as two important measures in evaluating a recommendation model. However, a clear concern is that a model focusing excessively on one measure will put the other one at risk, thus it is not easy to greatly improve diversity and accuracy simultaneously. In this paper, we propose to enhance the Resource-Allocation (RA) similarity in resource transfer equations of diffusion-like models, by giving a tunable exponent to the RA similarity, and traversing the value of the exponent to achieve the optimal recommendation results. In this way, we can increase the recommendation scores (allocated resource) of many unpopular objects. Experiments on three benchmark data sets, MovieLens, Netflix, and RateYourMusic show that the modified models can yield remarkable performance improvement compared with the original ones.
[ { "created": "Wed, 11 Nov 2015 14:43:32 GMT", "version": "v1" }, { "created": "Thu, 11 Oct 2018 06:13:15 GMT", "version": "v2" } ]
2018-10-17
[ [ "An", "Ya-Hui", "" ], [ "Dong", "Qiang", "" ], [ "Sun", "Chong-Jing", "" ], [ "Nie", "Da-Cheng", "" ], [ "Fu", "Yan", "" ] ]
In last decades, diversity and accuracy have been regarded as two important measures in evaluating a recommendation model. However, a clear concern is that a model focusing excessively on one measure will put the other one at risk, thus it is not easy to greatly improve diversity and accuracy simultaneously. In this paper, we propose to enhance the Resource-Allocation (RA) similarity in resource transfer equations of diffusion-like models, by giving a tunable exponent to the RA similarity, and traversing the value of the exponent to achieve the optimal recommendation results. In this way, we can increase the recommendation scores (allocated resource) of many unpopular objects. Experiments on three benchmark data sets, MovieLens, Netflix, and RateYourMusic show that the modified models can yield remarkable performance improvement compared with the original ones.
2003.09986
Yao Qiang
Yao Qiang, Xin Li, Dongxiao Zhu
Toward Tag-free Aspect Based Sentiment Analysis: A Multiple Attention Network Approach
to appear in the proceedings of IJCNN'20
null
null
null
cs.CL cs.IR
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Existing aspect based sentiment analysis (ABSA) approaches leverage various neural network models to extract the aspect sentiments via learning aspect-specific feature representations. However, these approaches heavily rely on manual tagging of user reviews according to the predefined aspects as the input, a laborious and time-consuming process. Moreover, the underlying methods do not explain how and why the opposing aspect level polarities in a user review lead to the overall polarity. In this paper, we tackle these two problems by designing and implementing a new Multiple-Attention Network (MAN) approach for more powerful ABSA without the need for aspect tags using two new tag-free data sets crawled directly from TripAdvisor ({https://www.tripadvisor.com}). With the Self- and Position-Aware attention mechanism, MAN is capable of extracting both aspect level and overall sentiments from the text reviews using the aspect level and overall customer ratings, and it can also detect the vital aspect(s) leading to the overall sentiment polarity among different aspects via a new aspect ranking scheme. We carry out extensive experiments to demonstrate the strong performance of MAN compared to other state-of-the-art ABSA approaches and the explainability of our approach by visualizing and interpreting attention weights in case studies.
[ { "created": "Sun, 22 Mar 2020 20:18:20 GMT", "version": "v1" } ]
2020-03-24
[ [ "Qiang", "Yao", "" ], [ "Li", "Xin", "" ], [ "Zhu", "Dongxiao", "" ] ]
Existing aspect based sentiment analysis (ABSA) approaches leverage various neural network models to extract the aspect sentiments via learning aspect-specific feature representations. However, these approaches heavily rely on manual tagging of user reviews according to the predefined aspects as the input, a laborious and time-consuming process. Moreover, the underlying methods do not explain how and why the opposing aspect level polarities in a user review lead to the overall polarity. In this paper, we tackle these two problems by designing and implementing a new Multiple-Attention Network (MAN) approach for more powerful ABSA without the need for aspect tags using two new tag-free data sets crawled directly from TripAdvisor ({https://www.tripadvisor.com}). With the Self- and Position-Aware attention mechanism, MAN is capable of extracting both aspect level and overall sentiments from the text reviews using the aspect level and overall customer ratings, and it can also detect the vital aspect(s) leading to the overall sentiment polarity among different aspects via a new aspect ranking scheme. We carry out extensive experiments to demonstrate the strong performance of MAN compared to other state-of-the-art ABSA approaches and the explainability of our approach by visualizing and interpreting attention weights in case studies.
2103.10241
Ke Lai
Ke Lai, Jing Lei, Yansha Deng, Lei Wen, Gaojie Chen
Analyzing Uplink Grant-free Sparse Code Multiple Access System in Massive IoT Networks
null
null
null
null
cs.IT eess.SP math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Grant-free sparse code multiple access (GF-SCMA) is considered to be a promising multiple access candidate for future wireless networks. In this paper, we focus on characterizing the performance of uplink GF-SCMA schemes in a network with ubiquitous connections, such as the Internet of Things (IoT) networks. To provide a tractable approach to evaluate the performance of GF-SCMA, we first develop a theoretical model taking into account the property of multi-user detection (MUD) in the SCMA system. We then analyze the error rate performance of GF-SCMA in the case of codebook collision to investigate the reliability of GF-SCMA when reusing codebook in massive IoT networks. For performance evaluation, accurate approximations for both success probability and average symbol error probability (ASEP) are derived. To elaborate further, we utilize the analytical results to discuss the impact of codeword sparse degree in GFSCMA. After that, we conduct a comparative study between SCMA and its variant, dense code multiple access (DCMA), with GF transmission to offer insights into the effectiveness of these two schemes. This facilitates the GF-SCMA system design in practical implementation. Simulation results show that denser codebooks can help to support more UEs and increase the reliability of data transmission in a GF-SCMA network. Moreover, a higher success probability can be achieved by GFSCMA with denser UE deployment at low detection thresholds since SCMA can achieve overloading gain.
[ { "created": "Thu, 18 Mar 2021 13:23:23 GMT", "version": "v1" } ]
2021-03-19
[ [ "Lai", "Ke", "" ], [ "Lei", "Jing", "" ], [ "Deng", "Yansha", "" ], [ "Wen", "Lei", "" ], [ "Chen", "Gaojie", "" ] ]
Grant-free sparse code multiple access (GF-SCMA) is considered to be a promising multiple access candidate for future wireless networks. In this paper, we focus on characterizing the performance of uplink GF-SCMA schemes in a network with ubiquitous connections, such as the Internet of Things (IoT) networks. To provide a tractable approach to evaluate the performance of GF-SCMA, we first develop a theoretical model taking into account the property of multi-user detection (MUD) in the SCMA system. We then analyze the error rate performance of GF-SCMA in the case of codebook collision to investigate the reliability of GF-SCMA when reusing codebook in massive IoT networks. For performance evaluation, accurate approximations for both success probability and average symbol error probability (ASEP) are derived. To elaborate further, we utilize the analytical results to discuss the impact of codeword sparse degree in GFSCMA. After that, we conduct a comparative study between SCMA and its variant, dense code multiple access (DCMA), with GF transmission to offer insights into the effectiveness of these two schemes. This facilitates the GF-SCMA system design in practical implementation. Simulation results show that denser codebooks can help to support more UEs and increase the reliability of data transmission in a GF-SCMA network. Moreover, a higher success probability can be achieved by GFSCMA with denser UE deployment at low detection thresholds since SCMA can achieve overloading gain.
0710.3279
Meixia Tao
Meixia Tao, Ying-Chang Liang and Fan Zhang
Resource Allocation for Delay Differentiated Traffic in Multiuser OFDM Systems
29 pages, 8 figures, submitted to IEEE Transactions on Wireless Communications
null
10.1109/ICC.2006.255331
null
cs.NI cs.IT math.IT
null
Most existing work on adaptive allocation of subcarriers and power in multiuser orthogonal frequency division multiplexing (OFDM) systems has focused on homogeneous traffic consisting solely of either delay-constrained data (guaranteed service) or non-delay-constrained data (best-effort service). In this paper, we investigate the resource allocation problem in a heterogeneous multiuser OFDM system with both delay-constrained (DC) and non-delay-constrained (NDC) traffic. The objective is to maximize the sum-rate of all the users with NDC traffic while maintaining guaranteed rates for the users with DC traffic under a total transmit power constraint. Through our analysis we show that the optimal power allocation over subcarriers follows a multi-level water-filling principle; moreover, the valid candidates competing for each subcarrier include only one NDC user but all DC users. By converting this combinatorial problem with exponential complexity into a convex problem or showing that it can be solved in the dual domain, efficient iterative algorithms are proposed to find the optimal solutions. To further reduce the computational cost, a low-complexity suboptimal algorithm is also developed. Numerical studies are conducted to evaluate the performance the proposed algorithms in terms of service outage probability, achievable transmission rate pairs for DC and NDC traffic, and multiuser diversity.
[ { "created": "Wed, 17 Oct 2007 12:04:34 GMT", "version": "v1" } ]
2016-11-15
[ [ "Tao", "Meixia", "" ], [ "Liang", "Ying-Chang", "" ], [ "Zhang", "Fan", "" ] ]
Most existing work on adaptive allocation of subcarriers and power in multiuser orthogonal frequency division multiplexing (OFDM) systems has focused on homogeneous traffic consisting solely of either delay-constrained data (guaranteed service) or non-delay-constrained data (best-effort service). In this paper, we investigate the resource allocation problem in a heterogeneous multiuser OFDM system with both delay-constrained (DC) and non-delay-constrained (NDC) traffic. The objective is to maximize the sum-rate of all the users with NDC traffic while maintaining guaranteed rates for the users with DC traffic under a total transmit power constraint. Through our analysis we show that the optimal power allocation over subcarriers follows a multi-level water-filling principle; moreover, the valid candidates competing for each subcarrier include only one NDC user but all DC users. By converting this combinatorial problem with exponential complexity into a convex problem or showing that it can be solved in the dual domain, efficient iterative algorithms are proposed to find the optimal solutions. To further reduce the computational cost, a low-complexity suboptimal algorithm is also developed. Numerical studies are conducted to evaluate the performance the proposed algorithms in terms of service outage probability, achievable transmission rate pairs for DC and NDC traffic, and multiuser diversity.
cs/0110052
Nandlal L. Sarda
N. L. Sarda and Ankur Jain
Mragyati : A System for Keyword-based Searching in Databases
null
null
null
null
cs.DB
null
The web, through many search engine sites, has popularized the keyword-based search paradigm, where a user can specify a string of keywords and expect to retrieve relevant documents, possibly ranked by their relevance to the query. Since a lot of information is stored in databases (and not as HTML documents), it is important to provide a similar search paradigm for databases, where users can query a database without knowing the database schema and database query languages such as SQL. In this paper, we propose such a database search system, which accepts a free-form query as a collection of keywords, translates it into queries on the database using the database metadata, and presents query results in a well-structured and browsable form. Th eysytem maps keywords onto the database schema and uses inter-relationships (i.e., data semantics) among the referred tables to generate meaningful query results. We also describe our prototype for database search, called Mragyati. Th eapproach proposed here is scalable, as it does not build an in-memory graph of the entire database for searching for relationships among the objects selected by the user's query.
[ { "created": "Thu, 25 Oct 2001 08:55:57 GMT", "version": "v1" } ]
2007-05-23
[ [ "Sarda", "N. L.", "" ], [ "Jain", "Ankur", "" ] ]
The web, through many search engine sites, has popularized the keyword-based search paradigm, where a user can specify a string of keywords and expect to retrieve relevant documents, possibly ranked by their relevance to the query. Since a lot of information is stored in databases (and not as HTML documents), it is important to provide a similar search paradigm for databases, where users can query a database without knowing the database schema and database query languages such as SQL. In this paper, we propose such a database search system, which accepts a free-form query as a collection of keywords, translates it into queries on the database using the database metadata, and presents query results in a well-structured and browsable form. Th eysytem maps keywords onto the database schema and uses inter-relationships (i.e., data semantics) among the referred tables to generate meaningful query results. We also describe our prototype for database search, called Mragyati. Th eapproach proposed here is scalable, as it does not build an in-memory graph of the entire database for searching for relationships among the objects selected by the user's query.
2407.12101
Marc Pickett
Marc Pickett, Jeremy Hartman, Ayan Kumar Bhowmick, Raquib-ul Alam, Aditya Vempaty
Better RAG using Relevant Information Gain
4 page paper submitted to EMNLP
null
null
null
cs.CL cs.AI
http://creativecommons.org/licenses/by/4.0/
A common way to extend the memory of large language models (LLMs) is by retrieval augmented generation (RAG), which inserts text retrieved from a larger memory into an LLM's context window. However, the context window is typically limited to several thousand tokens, which limits the number of retrieved passages that can inform a model's response. For this reason, it's important to avoid occupying context window space with redundant information by ensuring a degree of diversity among retrieved passages. At the same time, the information should also be relevant to the current task. Most prior methods that encourage diversity among retrieved results, such as Maximal Marginal Relevance (MMR), do so by incorporating an objective that explicitly trades off diversity and relevance. We propose a novel simple optimization metric based on relevant information gain, a probabilistic measure of the total information relevant to a query for a set of retrieved results. By optimizing this metric, diversity organically emerges from our system. When used as a drop-in replacement for the retrieval component of a RAG system, this method yields state-of-the-art performance on question answering tasks from the Retrieval Augmented Generation Benchmark (RGB), outperforming existing metrics that directly optimize for relevance and diversity.
[ { "created": "Tue, 16 Jul 2024 18:09:21 GMT", "version": "v1" } ]
2024-07-18
[ [ "Pickett", "Marc", "" ], [ "Hartman", "Jeremy", "" ], [ "Bhowmick", "Ayan Kumar", "" ], [ "Alam", "Raquib-ul", "" ], [ "Vempaty", "Aditya", "" ] ]
A common way to extend the memory of large language models (LLMs) is by retrieval augmented generation (RAG), which inserts text retrieved from a larger memory into an LLM's context window. However, the context window is typically limited to several thousand tokens, which limits the number of retrieved passages that can inform a model's response. For this reason, it's important to avoid occupying context window space with redundant information by ensuring a degree of diversity among retrieved passages. At the same time, the information should also be relevant to the current task. Most prior methods that encourage diversity among retrieved results, such as Maximal Marginal Relevance (MMR), do so by incorporating an objective that explicitly trades off diversity and relevance. We propose a novel simple optimization metric based on relevant information gain, a probabilistic measure of the total information relevant to a query for a set of retrieved results. By optimizing this metric, diversity organically emerges from our system. When used as a drop-in replacement for the retrieval component of a RAG system, this method yields state-of-the-art performance on question answering tasks from the Retrieval Augmented Generation Benchmark (RGB), outperforming existing metrics that directly optimize for relevance and diversity.
2002.06075
David Aparicio
David Apar\'icio, Ricardo Barata, Jo\~ao Bravo, Jo\~ao Tiago Ascens\~ao, Pedro Bizarro
ARMS: Automated rules management system for fraud detection
11 pages, 12 figures, submitted to KDD '20 Applied Data Science Track
null
null
null
cs.LG cs.AI cs.DB stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Fraud detection is essential in financial services, with the potential of greatly reducing criminal activities and saving considerable resources for businesses and customers. We address online fraud detection, which consists of classifying incoming transactions as either legitimate or fraudulent in real-time. Modern fraud detection systems consist of a machine learning model and rules defined by human experts. Often, the rules performance degrades over time due to concept drift, especially of adversarial nature. Furthermore, they can be costly to maintain, either because they are computationally expensive or because they send transactions for manual review. We propose ARMS, an automated rules management system that evaluates the contribution of individual rules and optimizes the set of active rules using heuristic search and a user-defined loss-function. It complies with critical domain-specific requirements, such as handling different actions (e.g., accept, alert, and decline), priorities, blacklists, and large datasets (i.e., hundreds of rules and millions of transactions). We use ARMS to optimize the rule-based systems of two real-world clients. Results show that it can maintain the original systems' performance (e.g., recall, or false-positive rate) using only a fraction of the original rules (~ 50% in one case, and ~ 20% in the other).
[ { "created": "Fri, 14 Feb 2020 15:29:59 GMT", "version": "v1" } ]
2020-02-17
[ [ "Aparício", "David", "" ], [ "Barata", "Ricardo", "" ], [ "Bravo", "João", "" ], [ "Ascensão", "João Tiago", "" ], [ "Bizarro", "Pedro", "" ] ]
Fraud detection is essential in financial services, with the potential of greatly reducing criminal activities and saving considerable resources for businesses and customers. We address online fraud detection, which consists of classifying incoming transactions as either legitimate or fraudulent in real-time. Modern fraud detection systems consist of a machine learning model and rules defined by human experts. Often, the rules performance degrades over time due to concept drift, especially of adversarial nature. Furthermore, they can be costly to maintain, either because they are computationally expensive or because they send transactions for manual review. We propose ARMS, an automated rules management system that evaluates the contribution of individual rules and optimizes the set of active rules using heuristic search and a user-defined loss-function. It complies with critical domain-specific requirements, such as handling different actions (e.g., accept, alert, and decline), priorities, blacklists, and large datasets (i.e., hundreds of rules and millions of transactions). We use ARMS to optimize the rule-based systems of two real-world clients. Results show that it can maintain the original systems' performance (e.g., recall, or false-positive rate) using only a fraction of the original rules (~ 50% in one case, and ~ 20% in the other).
1505.00346
Azra Abtahi
Azra Abtahi, M. Modarres-Hashemi, Farokh Marvasti, and Foroogh S. Tabataba
Power Allocation and Measurement Matrix Design for Block CS-Based Distributed MIMO Radars
The paper is accepted in Elseveir Aerospace Science and Technology
null
null
null
cs.IT math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Multiple-input multiple-output (MIMO) radars offer higher resolution, better target detection, and more accurate target parameter estimation. Due to the sparsity of the targets in space-velocity domain, we can exploit Compressive Sensing (CS) to improve the performance of MIMO radars when the sampling rate is much less than the Nyquist rate. In distributed MIMO radars, block CS methods can be used instead of classical CS ones for more performance improvement, because the received signal in this group of MIMO radars is a block sparse signal in a basis. In this paper, two new methods are proposed to improve the performance of the block CS-based distributed MIMO radars. The first one is a new method for optimal energy allocation to the transmitters, and the other one is a new method for optimal design of the measurement matrix. These methods are based on the minimization of an upper bound of the sensing matrix block-coherence. Simulation results show an increase in the accuracy of multiple targets parameters estimation for both proposed methods.
[ { "created": "Sat, 2 May 2015 14:49:50 GMT", "version": "v1" }, { "created": "Tue, 20 Oct 2015 15:33:49 GMT", "version": "v2" }, { "created": "Tue, 8 Mar 2016 19:21:16 GMT", "version": "v3" } ]
2016-03-09
[ [ "Abtahi", "Azra", "" ], [ "Modarres-Hashemi", "M.", "" ], [ "Marvasti", "Farokh", "" ], [ "Tabataba", "Foroogh S.", "" ] ]
Multiple-input multiple-output (MIMO) radars offer higher resolution, better target detection, and more accurate target parameter estimation. Due to the sparsity of the targets in space-velocity domain, we can exploit Compressive Sensing (CS) to improve the performance of MIMO radars when the sampling rate is much less than the Nyquist rate. In distributed MIMO radars, block CS methods can be used instead of classical CS ones for more performance improvement, because the received signal in this group of MIMO radars is a block sparse signal in a basis. In this paper, two new methods are proposed to improve the performance of the block CS-based distributed MIMO radars. The first one is a new method for optimal energy allocation to the transmitters, and the other one is a new method for optimal design of the measurement matrix. These methods are based on the minimization of an upper bound of the sensing matrix block-coherence. Simulation results show an increase in the accuracy of multiple targets parameters estimation for both proposed methods.
1304.5940
Emil Bj\"ornson
Nafiseh Shariati and Emil Bj\"ornson and Mats Bengtsson and M\'erouane Debbah
Low-Complexity Channel Estimation in Large-Scale MIMO using Polynomial Expansion
Published at IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC 2013), 8-11 September 2013, 6 pages, 4 figures, 1 table
null
null
null
cs.IT math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This paper considers pilot-based channel estimation in large-scale multiple-input multiple-output (MIMO) communication systems, also known as "massive MIMO". Unlike previous works on this topic, which mainly considered the impact of inter-cell disturbance due to pilot reuse (so-called pilot contamination), we are concerned with the computational complexity. The conventional minimum mean square error (MMSE) and minimum variance unbiased (MVU) channel estimators rely on inverting covariance matrices, which has cubic complexity in the multiplication of number of antennas at each side. Since this is extremely expensive when there are hundreds of antennas, we propose to approximate the inversion by an L-order matrix polynomial. A set of low-complexity Bayesian channel estimators, coined Polynomial ExpAnsion CHannel (PEACH) estimators, are introduced. The coefficients of the polynomials are optimized to yield small mean square error (MSE). We show numerically that near-optimal performance is achieved with low polynomial orders. In practice, the order L can be selected to balance between complexity and MSE. Interestingly, pilot contamination is beneficial to the PEACH estimators in the sense that smaller L can be used to achieve near-optimal MSEs.
[ { "created": "Mon, 22 Apr 2013 13:17:32 GMT", "version": "v1" }, { "created": "Wed, 17 Jul 2013 07:42:18 GMT", "version": "v2" } ]
2013-07-18
[ [ "Shariati", "Nafiseh", "" ], [ "Björnson", "Emil", "" ], [ "Bengtsson", "Mats", "" ], [ "Debbah", "Mérouane", "" ] ]
This paper considers pilot-based channel estimation in large-scale multiple-input multiple-output (MIMO) communication systems, also known as "massive MIMO". Unlike previous works on this topic, which mainly considered the impact of inter-cell disturbance due to pilot reuse (so-called pilot contamination), we are concerned with the computational complexity. The conventional minimum mean square error (MMSE) and minimum variance unbiased (MVU) channel estimators rely on inverting covariance matrices, which has cubic complexity in the multiplication of number of antennas at each side. Since this is extremely expensive when there are hundreds of antennas, we propose to approximate the inversion by an L-order matrix polynomial. A set of low-complexity Bayesian channel estimators, coined Polynomial ExpAnsion CHannel (PEACH) estimators, are introduced. The coefficients of the polynomials are optimized to yield small mean square error (MSE). We show numerically that near-optimal performance is achieved with low polynomial orders. In practice, the order L can be selected to balance between complexity and MSE. Interestingly, pilot contamination is beneficial to the PEACH estimators in the sense that smaller L can be used to achieve near-optimal MSEs.
2007.08147
Michel Rigo
E. Charlier, A. Massuir, M. Rigo, E. Rowland
Ultimate periodicity problem for linear numeration systems
39 pages, 2 figures. This is an improved version of the original submission. It clarifies some arguments taking into account several comments from reviews
International Journal of Algebra and Computation 32 (2022) 561-596
10.1142/S0218196722500254
null
cs.DM math.CO math.NT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We address the following decision problem. Given a numeration system $U$ and a $U$-recognizable set $X\subseteq\mathbb{N}$, i.e. the set of its greedy $U$-representations is recognized by a finite automaton, decide whether or not $X$ is ultimately periodic. We prove that this problem is decidable for a large class of numeration systems built on linearly recurrent sequences. Based on arithmetical considerations about the recurrence equation and on $p$-adic methods, the DFA given as input provides a bound on the admissible periods to test.
[ { "created": "Thu, 16 Jul 2020 07:12:39 GMT", "version": "v1" }, { "created": "Fri, 10 Dec 2021 12:08:35 GMT", "version": "v2" } ]
2023-09-04
[ [ "Charlier", "E.", "" ], [ "Massuir", "A.", "" ], [ "Rigo", "M.", "" ], [ "Rowland", "E.", "" ] ]
We address the following decision problem. Given a numeration system $U$ and a $U$-recognizable set $X\subseteq\mathbb{N}$, i.e. the set of its greedy $U$-representations is recognized by a finite automaton, decide whether or not $X$ is ultimately periodic. We prove that this problem is decidable for a large class of numeration systems built on linearly recurrent sequences. Based on arithmetical considerations about the recurrence equation and on $p$-adic methods, the DFA given as input provides a bound on the admissible periods to test.
2103.01169
Luca Maria Aiello
Sanja Scepanovic, Luca Maria Aiello, Ke Zhou, Sagar Joglekar, Daniele Quercia
The Healthy States of America: Creating a Health Taxonomy with Social Media
In proceedings of the International Conference on Web and Social Media (ICWSM'21)
null
null
null
cs.CY cs.AI cs.CL
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Since the uptake of social media, researchers have mined online discussions to track the outbreak and evolution of specific diseases or chronic conditions such as influenza or depression. To broaden the set of diseases under study, we developed a Deep Learning tool for Natural Language Processing that extracts mentions of virtually any medical condition or disease from unstructured social media text. With that tool at hand, we processed Reddit and Twitter posts, analyzed the clusters of the two resulting co-occurrence networks of conditions, and discovered that they correspond to well-defined categories of medical conditions. This resulted in the creation of the first comprehensive taxonomy of medical conditions automatically derived from online discussions. We validated the structure of our taxonomy against the official International Statistical Classification of Diseases and Related Health Problems (ICD-11), finding matches of our clusters with 20 official categories, out of 22. Based on the mentions of our taxonomy's sub-categories on Reddit posts geo-referenced in the U.S., we were then able to compute disease-specific health scores. As opposed to counts of disease mentions or counts with no knowledge of our taxonomy's structure, we found that our disease-specific health scores are causally linked with the officially reported prevalence of 18 conditions.
[ { "created": "Mon, 1 Mar 2021 18:07:47 GMT", "version": "v1" } ]
2021-03-02
[ [ "Scepanovic", "Sanja", "" ], [ "Aiello", "Luca Maria", "" ], [ "Zhou", "Ke", "" ], [ "Joglekar", "Sagar", "" ], [ "Quercia", "Daniele", "" ] ]
Since the uptake of social media, researchers have mined online discussions to track the outbreak and evolution of specific diseases or chronic conditions such as influenza or depression. To broaden the set of diseases under study, we developed a Deep Learning tool for Natural Language Processing that extracts mentions of virtually any medical condition or disease from unstructured social media text. With that tool at hand, we processed Reddit and Twitter posts, analyzed the clusters of the two resulting co-occurrence networks of conditions, and discovered that they correspond to well-defined categories of medical conditions. This resulted in the creation of the first comprehensive taxonomy of medical conditions automatically derived from online discussions. We validated the structure of our taxonomy against the official International Statistical Classification of Diseases and Related Health Problems (ICD-11), finding matches of our clusters with 20 official categories, out of 22. Based on the mentions of our taxonomy's sub-categories on Reddit posts geo-referenced in the U.S., we were then able to compute disease-specific health scores. As opposed to counts of disease mentions or counts with no knowledge of our taxonomy's structure, we found that our disease-specific health scores are causally linked with the officially reported prevalence of 18 conditions.
2406.04829
Zijia An
Zijia An, Boyu Diao, Libo Huang, Ruiqi Liu, Zhulin An, Yongjun Xu
EGOR: Efficient Generated Objects Replay for incremental object detection
null
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Incremental object detection aims to simultaneously maintain old-class accuracy and detect emerging new-class objects in incremental data. Most existing distillation-based methods underperform when unlabeled old-class objects are absent in the incremental dataset. While the absence can be mitigated by generating old-class samples, it also incurs high computational costs. In this paper, we argue that the extra computational cost stems from the inconsistency between the detector and the generative model, along with redundant generation. To overcome this problem, we propose Efficient Generated Object Replay (EGOR). Specifically, we generate old-class samples by inversing the original detectors, thus eliminating the necessity of training and storing additional generative models. We also propose augmented replay to reuse the objects in generated samples, thereby reducing the redundant generation. In addition, we propose high-response knowledge distillation focusing on the knowledge related to the old class, which transfers the knowledge in generated objects to the incremental detector. With the addition of the generated objects and losses, we observe a bias towards old classes in the detector. We balance the losses for old and new classes to alleviate the bias, thereby increasing the overall detection accuracy. Extensive experiments conducted on MS COCO 2017 demonstrate that our method can efficiently improve detection performance in the absence of old-class objects.
[ { "created": "Fri, 7 Jun 2024 10:54:40 GMT", "version": "v1" } ]
2024-06-10
[ [ "An", "Zijia", "" ], [ "Diao", "Boyu", "" ], [ "Huang", "Libo", "" ], [ "Liu", "Ruiqi", "" ], [ "An", "Zhulin", "" ], [ "Xu", "Yongjun", "" ] ]
Incremental object detection aims to simultaneously maintain old-class accuracy and detect emerging new-class objects in incremental data. Most existing distillation-based methods underperform when unlabeled old-class objects are absent in the incremental dataset. While the absence can be mitigated by generating old-class samples, it also incurs high computational costs. In this paper, we argue that the extra computational cost stems from the inconsistency between the detector and the generative model, along with redundant generation. To overcome this problem, we propose Efficient Generated Object Replay (EGOR). Specifically, we generate old-class samples by inversing the original detectors, thus eliminating the necessity of training and storing additional generative models. We also propose augmented replay to reuse the objects in generated samples, thereby reducing the redundant generation. In addition, we propose high-response knowledge distillation focusing on the knowledge related to the old class, which transfers the knowledge in generated objects to the incremental detector. With the addition of the generated objects and losses, we observe a bias towards old classes in the detector. We balance the losses for old and new classes to alleviate the bias, thereby increasing the overall detection accuracy. Extensive experiments conducted on MS COCO 2017 demonstrate that our method can efficiently improve detection performance in the absence of old-class objects.
2209.02298
Dipankar Chaki
Manan Choksi, Dipankar Chaki, Abdallah Lakhdari, Athman Bouguettaya
You Are What You Use: Usage-based Profiling in IoT Environments
4 pages, 2 figures
null
null
null
cs.HC
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Habit extraction is essential to automate services and provide appliance usage insights in the smart home environment. However, habit extraction comes with plenty of challenges in viewing typical start and end times for particular activities. This paper introduces a novel way of identifying habits using an ensemble of unsupervised clustering techniques. We use different clustering algorithms to extract habits based on how static or dynamic they are. Silhouette coefficients and a novel noise metric are utilized to extract habits appropriately. Furthermore, we associate the extracted habits with time intervals and a confidence score to denote how confident we are that a habit is likely to occur at that time.
[ { "created": "Tue, 6 Sep 2022 08:51:08 GMT", "version": "v1" } ]
2022-09-07
[ [ "Choksi", "Manan", "" ], [ "Chaki", "Dipankar", "" ], [ "Lakhdari", "Abdallah", "" ], [ "Bouguettaya", "Athman", "" ] ]
Habit extraction is essential to automate services and provide appliance usage insights in the smart home environment. However, habit extraction comes with plenty of challenges in viewing typical start and end times for particular activities. This paper introduces a novel way of identifying habits using an ensemble of unsupervised clustering techniques. We use different clustering algorithms to extract habits based on how static or dynamic they are. Silhouette coefficients and a novel noise metric are utilized to extract habits appropriately. Furthermore, we associate the extracted habits with time intervals and a confidence score to denote how confident we are that a habit is likely to occur at that time.
2111.01909
Wang Guangjun
Guangjun Wang
Geometrical holographic display
null
null
null
null
cs.GR physics.optics
http://creativecommons.org/licenses/by/4.0/
Is it possible to realize a holographic display with commercial available components and devices? Is it possible to manipulate light to reconstruct light field without using coherent light and complicated optical components? Is it possible to minimize the amount of date involved in building 3D scenes? Is it possible to design a holographic video display that doesn't require huge computational cost? Is it possible to realize a holographic video display with portable form like a flat-panel display commercialized nowadays? This research gives yes answers to all the above questions. A novel geometrical holographic display was proposed, which uses geometrical optical principle to reproduce realistic 3D images with all human visual cues and without visual side effects. In addition, a least necessary light field representation was introduced which can provide guidance for how to minimize the amount of data involved when designing a FP3D or building 3D scenes. Finally, a proof-of-concept prototype is set up which can provide true 3D scenes with depth range larger than 5m.
[ { "created": "Mon, 1 Nov 2021 11:56:23 GMT", "version": "v1" } ]
2021-11-04
[ [ "Wang", "Guangjun", "" ] ]
Is it possible to realize a holographic display with commercial available components and devices? Is it possible to manipulate light to reconstruct light field without using coherent light and complicated optical components? Is it possible to minimize the amount of date involved in building 3D scenes? Is it possible to design a holographic video display that doesn't require huge computational cost? Is it possible to realize a holographic video display with portable form like a flat-panel display commercialized nowadays? This research gives yes answers to all the above questions. A novel geometrical holographic display was proposed, which uses geometrical optical principle to reproduce realistic 3D images with all human visual cues and without visual side effects. In addition, a least necessary light field representation was introduced which can provide guidance for how to minimize the amount of data involved when designing a FP3D or building 3D scenes. Finally, a proof-of-concept prototype is set up which can provide true 3D scenes with depth range larger than 5m.
2405.11210
Emilio Mart\'inez-Pa\~neda
C. Cui, P. Bortot, M. Ortolani, E. Mart\'inez-Pa\~neda
Computational predictions of hydrogen-assisted fatigue crack growth
null
null
null
null
cs.CE cond-mat.mtrl-sci physics.app-ph physics.chem-ph
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
A new model is presented to predict hydrogen-assisted fatigue. The model combines a phase field description of fracture and fatigue, stress-assisted hydrogen diffusion, and a toughness degradation formulation with cyclic and hydrogen contributions. Hydrogen-assisted fatigue crack growth predictions exhibit an excellent agreement with experiments over all the scenarios considered, spanning multiple load ratios, H2 pressures and loading frequencies. These are obtained without any calibration with hydrogen-assisted fatigue data, taking as input only mechanical and hydrogen transport material properties, the material's fatigue characteristics (from a single test in air), and the sensitivity of fracture toughness to hydrogen content. Furthermore, the model is used to determine: (i) what are suitable test loading frequencies to obtain conservative data, and (ii) the underestimation made when not pre-charging samples. The model can handle both laboratory specimens and large-scale engineering components, enabling the Virtual Testing paradigm in infrastructure exposed to hydrogen environments and cyclic loading.
[ { "created": "Sat, 18 May 2024 07:34:48 GMT", "version": "v1" } ]
2024-05-21
[ [ "Cui", "C.", "" ], [ "Bortot", "P.", "" ], [ "Ortolani", "M.", "" ], [ "Martínez-Pañeda", "E.", "" ] ]
A new model is presented to predict hydrogen-assisted fatigue. The model combines a phase field description of fracture and fatigue, stress-assisted hydrogen diffusion, and a toughness degradation formulation with cyclic and hydrogen contributions. Hydrogen-assisted fatigue crack growth predictions exhibit an excellent agreement with experiments over all the scenarios considered, spanning multiple load ratios, H2 pressures and loading frequencies. These are obtained without any calibration with hydrogen-assisted fatigue data, taking as input only mechanical and hydrogen transport material properties, the material's fatigue characteristics (from a single test in air), and the sensitivity of fracture toughness to hydrogen content. Furthermore, the model is used to determine: (i) what are suitable test loading frequencies to obtain conservative data, and (ii) the underestimation made when not pre-charging samples. The model can handle both laboratory specimens and large-scale engineering components, enabling the Virtual Testing paradigm in infrastructure exposed to hydrogen environments and cyclic loading.
1006.5686
Steven Weber
Nan Xie, Steven Weber
Geometric Approximations of Some Aloha-like Stability Regions
Presented at IEEE ISIT 2010 (Austin, TX)
null
10.1109/ISIT.2010.5513425
null
cs.IT math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Most bounds on the stability region of Aloha give necessary and sufficient conditions for the stability of an arrival rate vector under a specific contention probability (control) vector. But such results do not yield easy-to-check bounds on the overall Aloha stability region because they potentially require checking membership in an uncountably infinite number of sets parameterized by each possible control vector. In this paper we consider an important specific inner bound on Aloha that has this property of difficulty to check membership in the set. We provide ellipsoids (for which membership is easy-to-check) that we conjecture are inner and outer bounds on this set. We also study the set of controls that stabilize a fixed arrival rate vector; this set is shown to be a convex set.
[ { "created": "Tue, 29 Jun 2010 17:32:13 GMT", "version": "v1" } ]
2016-11-18
[ [ "Xie", "Nan", "" ], [ "Weber", "Steven", "" ] ]
Most bounds on the stability region of Aloha give necessary and sufficient conditions for the stability of an arrival rate vector under a specific contention probability (control) vector. But such results do not yield easy-to-check bounds on the overall Aloha stability region because they potentially require checking membership in an uncountably infinite number of sets parameterized by each possible control vector. In this paper we consider an important specific inner bound on Aloha that has this property of difficulty to check membership in the set. We provide ellipsoids (for which membership is easy-to-check) that we conjecture are inner and outer bounds on this set. We also study the set of controls that stabilize a fixed arrival rate vector; this set is shown to be a convex set.
1507.04438
Akbar Rafiey
Binay Bhattacharya, Ante \'Custi\'c, Akbar Rafiey, Arash Rafiey, Vladyslav Sokol
Approximation Algorithms for Generalized MST and TSP in Grid Clusters
null
null
null
null
cs.DM cs.CG cs.DS
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We consider a special case of the generalized minimum spanning tree problem (GMST) and the generalized travelling salesman problem (GTSP) where we are given a set of points inside the integer grid (in Euclidean plane) where each grid cell is $1 \times 1$. In the MST version of the problem, the goal is to find a minimum tree that contains exactly one point from each non-empty grid cell (cluster). Similarly, in the TSP version of the problem, the goal is to find a minimum weight cycle containing one point from each non-empty grid cell. We give a $(1+4\sqrt{2}+\epsilon)$ and $(1.5+8\sqrt{2}+\epsilon)$-approximation algorithm for these two problems in the described setting, respectively. Our motivation is based on the problem posed in [7] for a constant approximation algorithm. The authors designed a PTAS for the more special case of the GMST where non-empty cells are connected end dense enough. However, their algorithm heavily relies on this connectivity restriction and is unpractical. Our results develop the topic further.
[ { "created": "Thu, 16 Jul 2015 03:00:41 GMT", "version": "v1" } ]
2015-07-17
[ [ "Bhattacharya", "Binay", "" ], [ "Ćustić", "Ante", "" ], [ "Rafiey", "Akbar", "" ], [ "Rafiey", "Arash", "" ], [ "Sokol", "Vladyslav", "" ] ]
We consider a special case of the generalized minimum spanning tree problem (GMST) and the generalized travelling salesman problem (GTSP) where we are given a set of points inside the integer grid (in Euclidean plane) where each grid cell is $1 \times 1$. In the MST version of the problem, the goal is to find a minimum tree that contains exactly one point from each non-empty grid cell (cluster). Similarly, in the TSP version of the problem, the goal is to find a minimum weight cycle containing one point from each non-empty grid cell. We give a $(1+4\sqrt{2}+\epsilon)$ and $(1.5+8\sqrt{2}+\epsilon)$-approximation algorithm for these two problems in the described setting, respectively. Our motivation is based on the problem posed in [7] for a constant approximation algorithm. The authors designed a PTAS for the more special case of the GMST where non-empty cells are connected end dense enough. However, their algorithm heavily relies on this connectivity restriction and is unpractical. Our results develop the topic further.
2404.03225
Xu Wang
Xu Wang, Tian Ye, Rajgopal Kannan, Viktor Prasanna
FACTUAL: A Novel Framework for Contrastive Learning Based Robust SAR Image Classification
2024 IEEE Radar Conference
null
null
null
cs.CV cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Deep Learning (DL) Models for Synthetic Aperture Radar (SAR) Automatic Target Recognition (ATR), while delivering improved performance, have been shown to be quite vulnerable to adversarial attacks. Existing works improve robustness by training models on adversarial samples. However, by focusing mostly on attacks that manipulate images randomly, they neglect the real-world feasibility of such attacks. In this paper, we propose FACTUAL, a novel Contrastive Learning framework for Adversarial Training and robust SAR classification. FACTUAL consists of two components: (1) Differing from existing works, a novel perturbation scheme that incorporates realistic physical adversarial attacks (such as OTSA) to build a supervised adversarial pre-training network. This network utilizes class labels for clustering clean and perturbed images together into a more informative feature space. (2) A linear classifier cascaded after the encoder to use the computed representations to predict the target labels. By pre-training and fine-tuning our model on both clean and adversarial samples, we show that our model achieves high prediction accuracy on both cases. Our model achieves 99.7% accuracy on clean samples, and 89.6% on perturbed samples, both outperforming previous state-of-the-art methods.
[ { "created": "Thu, 4 Apr 2024 06:20:22 GMT", "version": "v1" } ]
2024-04-05
[ [ "Wang", "Xu", "" ], [ "Ye", "Tian", "" ], [ "Kannan", "Rajgopal", "" ], [ "Prasanna", "Viktor", "" ] ]
Deep Learning (DL) Models for Synthetic Aperture Radar (SAR) Automatic Target Recognition (ATR), while delivering improved performance, have been shown to be quite vulnerable to adversarial attacks. Existing works improve robustness by training models on adversarial samples. However, by focusing mostly on attacks that manipulate images randomly, they neglect the real-world feasibility of such attacks. In this paper, we propose FACTUAL, a novel Contrastive Learning framework for Adversarial Training and robust SAR classification. FACTUAL consists of two components: (1) Differing from existing works, a novel perturbation scheme that incorporates realistic physical adversarial attacks (such as OTSA) to build a supervised adversarial pre-training network. This network utilizes class labels for clustering clean and perturbed images together into a more informative feature space. (2) A linear classifier cascaded after the encoder to use the computed representations to predict the target labels. By pre-training and fine-tuning our model on both clean and adversarial samples, we show that our model achieves high prediction accuracy on both cases. Our model achieves 99.7% accuracy on clean samples, and 89.6% on perturbed samples, both outperforming previous state-of-the-art methods.
2003.09148
Nico Messikommer
Nico Messikommer, Daniel Gehrig, Antonio Loquercio, Davide Scaramuzza
Event-based Asynchronous Sparse Convolutional Networks
null
European Conference on Computer Vision (ECCV), 2020
null
null
cs.CV cs.LG eess.SP
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Event cameras are bio-inspired sensors that respond to per-pixel brightness changes in the form of asynchronous and sparse "events". Recently, pattern recognition algorithms, such as learning-based methods, have made significant progress with event cameras by converting events into synchronous dense, image-like representations and applying traditional machine learning methods developed for standard cameras. However, these approaches discard the spatial and temporal sparsity inherent in event data at the cost of higher computational complexity and latency. In this work, we present a general framework for converting models trained on synchronous image-like event representations into asynchronous models with identical output, thus directly leveraging the intrinsic asynchronous and sparse nature of the event data. We show both theoretically and experimentally that this drastically reduces the computational complexity and latency of high-capacity, synchronous neural networks without sacrificing accuracy. In addition, our framework has several desirable characteristics: (i) it exploits spatio-temporal sparsity of events explicitly, (ii) it is agnostic to the event representation, network architecture, and task, and (iii) it does not require any train-time change, since it is compatible with the standard neural networks' training process. We thoroughly validate the proposed framework on two computer vision tasks: object detection and object recognition. In these tasks, we reduce the computational complexity up to 20 times with respect to high-latency neural networks. At the same time, we outperform state-of-the-art asynchronous approaches up to 24% in prediction accuracy.
[ { "created": "Fri, 20 Mar 2020 08:39:49 GMT", "version": "v1" }, { "created": "Fri, 17 Jul 2020 15:52:12 GMT", "version": "v2" } ]
2020-07-20
[ [ "Messikommer", "Nico", "" ], [ "Gehrig", "Daniel", "" ], [ "Loquercio", "Antonio", "" ], [ "Scaramuzza", "Davide", "" ] ]
Event cameras are bio-inspired sensors that respond to per-pixel brightness changes in the form of asynchronous and sparse "events". Recently, pattern recognition algorithms, such as learning-based methods, have made significant progress with event cameras by converting events into synchronous dense, image-like representations and applying traditional machine learning methods developed for standard cameras. However, these approaches discard the spatial and temporal sparsity inherent in event data at the cost of higher computational complexity and latency. In this work, we present a general framework for converting models trained on synchronous image-like event representations into asynchronous models with identical output, thus directly leveraging the intrinsic asynchronous and sparse nature of the event data. We show both theoretically and experimentally that this drastically reduces the computational complexity and latency of high-capacity, synchronous neural networks without sacrificing accuracy. In addition, our framework has several desirable characteristics: (i) it exploits spatio-temporal sparsity of events explicitly, (ii) it is agnostic to the event representation, network architecture, and task, and (iii) it does not require any train-time change, since it is compatible with the standard neural networks' training process. We thoroughly validate the proposed framework on two computer vision tasks: object detection and object recognition. In these tasks, we reduce the computational complexity up to 20 times with respect to high-latency neural networks. At the same time, we outperform state-of-the-art asynchronous approaches up to 24% in prediction accuracy.
2203.12412
Ahmet Caner Y\"uz\"ug\"uler
Ahmet Caner Y\"uz\"ug\"uler, Nikolaos Dimitriadis, Pascal Frossard
U-Boost NAS: Utilization-Boosted Differentiable Neural Architecture Search
null
null
null
null
cs.LG cs.AI cs.CV
http://creativecommons.org/licenses/by/4.0/
Optimizing resource utilization in target platforms is key to achieving high performance during DNN inference. While optimizations have been proposed for inference latency, memory footprint, and energy consumption, prior hardware-aware neural architecture search (NAS) methods have omitted resource utilization, preventing DNNs to take full advantage of the target inference platforms. Modeling resource utilization efficiently and accurately is challenging, especially for widely-used array-based inference accelerators such as Google TPU. In this work, we propose a novel hardware-aware NAS framework that does not only optimize for task accuracy and inference latency, but also for resource utilization. We also propose and validate a new computational model for resource utilization in inference accelerators. By using the proposed NAS framework and the proposed resource utilization model, we achieve 2.8 - 4x speedup for DNN inference compared to prior hardware-aware NAS methods while attaining similar or improved accuracy in image classification on CIFAR-10 and Imagenet-100 datasets.
[ { "created": "Wed, 23 Mar 2022 13:44:15 GMT", "version": "v1" } ]
2022-03-24
[ [ "Yüzügüler", "Ahmet Caner", "" ], [ "Dimitriadis", "Nikolaos", "" ], [ "Frossard", "Pascal", "" ] ]
Optimizing resource utilization in target platforms is key to achieving high performance during DNN inference. While optimizations have been proposed for inference latency, memory footprint, and energy consumption, prior hardware-aware neural architecture search (NAS) methods have omitted resource utilization, preventing DNNs to take full advantage of the target inference platforms. Modeling resource utilization efficiently and accurately is challenging, especially for widely-used array-based inference accelerators such as Google TPU. In this work, we propose a novel hardware-aware NAS framework that does not only optimize for task accuracy and inference latency, but also for resource utilization. We also propose and validate a new computational model for resource utilization in inference accelerators. By using the proposed NAS framework and the proposed resource utilization model, we achieve 2.8 - 4x speedup for DNN inference compared to prior hardware-aware NAS methods while attaining similar or improved accuracy in image classification on CIFAR-10 and Imagenet-100 datasets.
1608.04309
Yasin Yazicioglu
A. Yasin Yazicioglu, Waseem Abbas, and Magnus Egerstedt
Graph Distances and Controllability of Networks
Accepted to the IEEE Transactions on Automatic Control
null
10.1109/TAC.2016.2546180
null
cs.SY cs.SI math.OC
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In this technical note, we study the controllability of diffusively coupled networks from a graph theoretic perspective. We consider leader-follower networks, where the external control inputs are injected to only some of the agents, namely the leaders. Our main result relates the controllability of such systems to the graph distances between the agents. More specifically, we present a graph topological lower bound on the rank of the controllability matrix. This lower bound is tight, and it is applicable to systems with arbitrary network topologies, coupling weights, and number of leaders. An algorithm for computing the lower bound is also provided. Furthermore, as a prominent application, we present how the proposed bound can be utilized to select a minimal set of leaders for achieving controllability, even when the coupling weights are unknown.
[ { "created": "Mon, 15 Aug 2016 15:51:42 GMT", "version": "v1" } ]
2016-08-17
[ [ "Yazicioglu", "A. Yasin", "" ], [ "Abbas", "Waseem", "" ], [ "Egerstedt", "Magnus", "" ] ]
In this technical note, we study the controllability of diffusively coupled networks from a graph theoretic perspective. We consider leader-follower networks, where the external control inputs are injected to only some of the agents, namely the leaders. Our main result relates the controllability of such systems to the graph distances between the agents. More specifically, we present a graph topological lower bound on the rank of the controllability matrix. This lower bound is tight, and it is applicable to systems with arbitrary network topologies, coupling weights, and number of leaders. An algorithm for computing the lower bound is also provided. Furthermore, as a prominent application, we present how the proposed bound can be utilized to select a minimal set of leaders for achieving controllability, even when the coupling weights are unknown.
1911.06489
Yanjie Gou
Yanjie Gou, Yinjie Lei, Lingqiao Liu, Pingping Zhang, Xi Peng
Improving Distant Supervised Relation Extraction by Dynamic Neural Network
29 pages, 8 figures
null
null
null
cs.CL cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Distant Supervised Relation Extraction (DSRE) is usually formulated as a problem of classifying a bag of sentences that contain two query entities, into the predefined relation classes. Most existing methods consider those relation classes as distinct semantic categories while ignoring their potential connection to query entities. In this paper, we propose to leverage this connection to improve the relation extraction accuracy. Our key ideas are twofold: (1) For sentences belonging to the same relation class, the expression style, i.e. words choice, can vary according to the query entities. To account for this style shift, the model should adjust its parameters in accordance with entity types. (2) Some relation classes are semantically similar, and the entity types appear in one relation may also appear in others. Therefore, it can be trained cross different relation classes and further enhance those classes with few samples, i.e., long-tail classes. To unify these two arguments, we developed a novel Dynamic Neural Network for Relation Extraction (DNNRE). The network adopts a novel dynamic parameter generator that dynamically generates the network parameters according to the query entity types and relation classes. By using this mechanism, the network can simultaneously handle the style shift problem and enhance the prediction accuracy for long-tail classes. Through our experimental study, we demonstrate the effectiveness of the proposed method and show that it can achieve superior performance over the state-of-the-art methods.
[ { "created": "Fri, 15 Nov 2019 06:31:13 GMT", "version": "v1" }, { "created": "Fri, 13 Dec 2019 04:29:41 GMT", "version": "v2" } ]
2019-12-16
[ [ "Gou", "Yanjie", "" ], [ "Lei", "Yinjie", "" ], [ "Liu", "Lingqiao", "" ], [ "Zhang", "Pingping", "" ], [ "Peng", "Xi", "" ] ]
Distant Supervised Relation Extraction (DSRE) is usually formulated as a problem of classifying a bag of sentences that contain two query entities, into the predefined relation classes. Most existing methods consider those relation classes as distinct semantic categories while ignoring their potential connection to query entities. In this paper, we propose to leverage this connection to improve the relation extraction accuracy. Our key ideas are twofold: (1) For sentences belonging to the same relation class, the expression style, i.e. words choice, can vary according to the query entities. To account for this style shift, the model should adjust its parameters in accordance with entity types. (2) Some relation classes are semantically similar, and the entity types appear in one relation may also appear in others. Therefore, it can be trained cross different relation classes and further enhance those classes with few samples, i.e., long-tail classes. To unify these two arguments, we developed a novel Dynamic Neural Network for Relation Extraction (DNNRE). The network adopts a novel dynamic parameter generator that dynamically generates the network parameters according to the query entity types and relation classes. By using this mechanism, the network can simultaneously handle the style shift problem and enhance the prediction accuracy for long-tail classes. Through our experimental study, we demonstrate the effectiveness of the proposed method and show that it can achieve superior performance over the state-of-the-art methods.
2303.01277
Meng Zhang
Meng Zhang, Qinghao Hu, Peng Sun, Yonggang Wen, Tianwei Zhang
Boosting Distributed Full-graph GNN Training with Asynchronous One-bit Communication
null
null
null
null
cs.DC cs.AI cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Training Graph Neural Networks (GNNs) on large graphs is challenging due to the conflict between the high memory demand and limited GPU memory. Recently, distributed full-graph GNN training has been widely adopted to tackle this problem. However, the substantial inter-GPU communication overhead can cause severe throughput degradation. Existing communication compression techniques mainly focus on traditional DNN training, whose bottleneck lies in synchronizing gradients and parameters. We find they do not work well in distributed GNN training as the barrier is the layer-wise communication of features during the forward pass & feature gradients during the backward pass. To this end, we propose an efficient distributed GNN training framework Sylvie, which employs one-bit quantization technique in GNNs and further pipelines the curtailed communication with computation to enormously shrink the overhead while maintaining the model quality. In detail, Sylvie provides a lightweight Low-bit Module to quantize the sent data and dequantize the received data back to full precision values in each layer. Additionally, we propose a Bounded Staleness Adaptor to control the introduced staleness to achieve further performance enhancement. We conduct theoretical convergence analysis and extensive experiments on various models & datasets to demonstrate Sylvie can considerably boost the training throughput by up to 28.1x.
[ { "created": "Thu, 2 Mar 2023 14:02:39 GMT", "version": "v1" } ]
2023-03-03
[ [ "Zhang", "Meng", "" ], [ "Hu", "Qinghao", "" ], [ "Sun", "Peng", "" ], [ "Wen", "Yonggang", "" ], [ "Zhang", "Tianwei", "" ] ]
Training Graph Neural Networks (GNNs) on large graphs is challenging due to the conflict between the high memory demand and limited GPU memory. Recently, distributed full-graph GNN training has been widely adopted to tackle this problem. However, the substantial inter-GPU communication overhead can cause severe throughput degradation. Existing communication compression techniques mainly focus on traditional DNN training, whose bottleneck lies in synchronizing gradients and parameters. We find they do not work well in distributed GNN training as the barrier is the layer-wise communication of features during the forward pass & feature gradients during the backward pass. To this end, we propose an efficient distributed GNN training framework Sylvie, which employs one-bit quantization technique in GNNs and further pipelines the curtailed communication with computation to enormously shrink the overhead while maintaining the model quality. In detail, Sylvie provides a lightweight Low-bit Module to quantize the sent data and dequantize the received data back to full precision values in each layer. Additionally, we propose a Bounded Staleness Adaptor to control the introduced staleness to achieve further performance enhancement. We conduct theoretical convergence analysis and extensive experiments on various models & datasets to demonstrate Sylvie can considerably boost the training throughput by up to 28.1x.
2001.00182
Francesco Malandrino
Christian Vitale and Carla Fabiana Chiasserini and Francesco Malandrino and Senay Semu Tadesse
Characterizing Delay and Control Traffic of the Cellular MME with IoT Support
null
null
null
null
cs.NI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
One of the main use cases for advanced cellular networks is represented by massive Internet-of-things (MIoT), i.e., an enormous number of IoT devices that transmit data toward the cellular network infrastructure. To make cellular MIoT a reality, data transfer and control procedures specifically designed for the support of IoT are needed. For this reason, 3GPP has introduced the Control Plane Cellular IoT optimization, which foresees a simplified bearer instantiation, with the Mobility Management Entity (MME) handling both control and data traffic. The performance of the MME has therefore become critical, and properly scaling its computational capability can determine the ability of the whole network to tackle MIoT effectively. In particular, considering virtualized networks and the need for an efficient allocation of computing resources, it is paramount to characterize the MME performance as the MIoT traffic load changes. We address this need by presenting compact, closed-form expressions linking the number of IoT sources with the rate at which bearers are requested, and such a rate with the delay incurred by the IoT data. We show that our analysis, supported by testbed experiments and verified through large-scale simulations, represents a valuable tool to make effective scaling decisions in virtualized cellular core networks.
[ { "created": "Wed, 1 Jan 2020 10:14:54 GMT", "version": "v1" } ]
2020-01-03
[ [ "Vitale", "Christian", "" ], [ "Chiasserini", "Carla Fabiana", "" ], [ "Malandrino", "Francesco", "" ], [ "Tadesse", "Senay Semu", "" ] ]
One of the main use cases for advanced cellular networks is represented by massive Internet-of-things (MIoT), i.e., an enormous number of IoT devices that transmit data toward the cellular network infrastructure. To make cellular MIoT a reality, data transfer and control procedures specifically designed for the support of IoT are needed. For this reason, 3GPP has introduced the Control Plane Cellular IoT optimization, which foresees a simplified bearer instantiation, with the Mobility Management Entity (MME) handling both control and data traffic. The performance of the MME has therefore become critical, and properly scaling its computational capability can determine the ability of the whole network to tackle MIoT effectively. In particular, considering virtualized networks and the need for an efficient allocation of computing resources, it is paramount to characterize the MME performance as the MIoT traffic load changes. We address this need by presenting compact, closed-form expressions linking the number of IoT sources with the rate at which bearers are requested, and such a rate with the delay incurred by the IoT data. We show that our analysis, supported by testbed experiments and verified through large-scale simulations, represents a valuable tool to make effective scaling decisions in virtualized cellular core networks.
1412.2824
Yu Zhang
Vignesh Narayanan, Yu Zhang, Nathaniel Mendoza and Subbarao Kambhampati
Plan or not: Remote Human-robot Teaming with Incomplete Task Information
null
null
null
null
cs.AI cs.RO
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Human-robot interaction can be divided into two categories based on the physical distance between the human and robot: remote and proximal. In proximal interaction, the human and robot often engage in close coordination; in remote interaction, the human and robot are less coupled due to communication constraints. As a result, providing automation for the robot in remote interaction becomes more important. Thus far, human factor studies on automation in remote human-robot interaction have been restricted to various forms of supervision, in which the robot is essentially being used as a smart mobile manipulation platform with sensing capabilities. In this paper, we investigate the incorporation of general planning capability into the robot to facilitate peer-to-peer human-robot teaming, in which the human and robot are viewed as teammates that are physically separated. The human and robot share the same global goal and collaborate to achieve it. Note that humans may feel uncomfortable at such robot autonomy, which can potentially reduce teaming performance. One important difference between peer-to-peer teaming and supervised teaming is that an autonomous robot in peer-to-peer teaming can achieve the goal alone when the task information is completely specified. However, incompleteness often exists, which implies information asymmetry. While information asymmetry can be desirable sometimes, it may also lead to the robot choosing improper actions that negatively influence the teaming performance. We aim to investigate the various trade-offs, e.g., mental workload and situation awareness, between these two types of remote human-robot teaming.
[ { "created": "Tue, 9 Dec 2014 01:05:59 GMT", "version": "v1" } ]
2014-12-10
[ [ "Narayanan", "Vignesh", "" ], [ "Zhang", "Yu", "" ], [ "Mendoza", "Nathaniel", "" ], [ "Kambhampati", "Subbarao", "" ] ]
Human-robot interaction can be divided into two categories based on the physical distance between the human and robot: remote and proximal. In proximal interaction, the human and robot often engage in close coordination; in remote interaction, the human and robot are less coupled due to communication constraints. As a result, providing automation for the robot in remote interaction becomes more important. Thus far, human factor studies on automation in remote human-robot interaction have been restricted to various forms of supervision, in which the robot is essentially being used as a smart mobile manipulation platform with sensing capabilities. In this paper, we investigate the incorporation of general planning capability into the robot to facilitate peer-to-peer human-robot teaming, in which the human and robot are viewed as teammates that are physically separated. The human and robot share the same global goal and collaborate to achieve it. Note that humans may feel uncomfortable at such robot autonomy, which can potentially reduce teaming performance. One important difference between peer-to-peer teaming and supervised teaming is that an autonomous robot in peer-to-peer teaming can achieve the goal alone when the task information is completely specified. However, incompleteness often exists, which implies information asymmetry. While information asymmetry can be desirable sometimes, it may also lead to the robot choosing improper actions that negatively influence the teaming performance. We aim to investigate the various trade-offs, e.g., mental workload and situation awareness, between these two types of remote human-robot teaming.
2405.00251
Dylan Green
Dylan Green, William Harvey, Saeid Naderiparizi, Matthew Niedoba, Yunpeng Liu, Xiaoxuan Liang, Jonathan Lavington, Ke Zhang, Vasileios Lioutas, Setareh Dabiri, Adam Scibior, Berend Zwartsenberg, Frank Wood
Semantically Consistent Video Inpainting with Conditional Diffusion Models
null
null
null
null
cs.CV cs.LG
http://creativecommons.org/licenses/by-sa/4.0/
Current state-of-the-art methods for video inpainting typically rely on optical flow or attention-based approaches to inpaint masked regions by propagating visual information across frames. While such approaches have led to significant progress on standard benchmarks, they struggle with tasks that require the synthesis of novel content that is not present in other frames. In this paper we reframe video inpainting as a conditional generative modeling problem and present a framework for solving such problems with conditional video diffusion models. We highlight the advantages of using a generative approach for this task, showing that our method is capable of generating diverse, high-quality inpaintings and synthesizing new content that is spatially, temporally, and semantically consistent with the provided context.
[ { "created": "Tue, 30 Apr 2024 23:49:26 GMT", "version": "v1" } ]
2024-05-02
[ [ "Green", "Dylan", "" ], [ "Harvey", "William", "" ], [ "Naderiparizi", "Saeid", "" ], [ "Niedoba", "Matthew", "" ], [ "Liu", "Yunpeng", "" ], [ "Liang", "Xiaoxuan", "" ], [ "Lavington", "Jonathan", "" ], [ "Zhang", "Ke", "" ], [ "Lioutas", "Vasileios", "" ], [ "Dabiri", "Setareh", "" ], [ "Scibior", "Adam", "" ], [ "Zwartsenberg", "Berend", "" ], [ "Wood", "Frank", "" ] ]
Current state-of-the-art methods for video inpainting typically rely on optical flow or attention-based approaches to inpaint masked regions by propagating visual information across frames. While such approaches have led to significant progress on standard benchmarks, they struggle with tasks that require the synthesis of novel content that is not present in other frames. In this paper we reframe video inpainting as a conditional generative modeling problem and present a framework for solving such problems with conditional video diffusion models. We highlight the advantages of using a generative approach for this task, showing that our method is capable of generating diverse, high-quality inpaintings and synthesizing new content that is spatially, temporally, and semantically consistent with the provided context.
2405.07481
Xiaoyi Zhang
Tianci Bi, Xiaoyi Zhang, Zhizheng Zhang, Wenxuan Xie, Cuiling Lan, Yan Lu and Nanning Zheng
Text Grouping Adapter: Adapting Pre-trained Text Detector for Layout Analysis
Accepted to CVPR 2024
null
null
null
cs.CV
http://creativecommons.org/licenses/by/4.0/
Significant progress has been made in scene text detection models since the rise of deep learning, but scene text layout analysis, which aims to group detected text instances as paragraphs, has not kept pace. Previous works either treated text detection and grouping using separate models, or train a model from scratch while using a unified one. All of them have not yet made full use of the already well-trained text detectors and easily obtainable detection datasets. In this paper, we present Text Grouping Adapter (TGA), a module that can enable the utilization of various pre-trained text detectors to learn layout analysis, allowing us to adopt a well-trained text detector right off the shelf or just fine-tune it efficiently. Designed to be compatible with various text detector architectures, TGA takes detected text regions and image features as universal inputs to assemble text instance features. To capture broader contextual information for layout analysis, we propose to predict text group masks from text instance features by one-to-many assignment. Our comprehensive experiments demonstrate that, even with frozen pre-trained models, incorporating our TGA into various pre-trained text detectors and text spotters can achieve superior layout analysis performance, simultaneously inheriting generalized text detection ability from pre-training. In the case of full parameter fine-tuning, we can further improve layout analysis performance.
[ { "created": "Mon, 13 May 2024 05:48:35 GMT", "version": "v1" } ]
2024-05-14
[ [ "Bi", "Tianci", "" ], [ "Zhang", "Xiaoyi", "" ], [ "Zhang", "Zhizheng", "" ], [ "Xie", "Wenxuan", "" ], [ "Lan", "Cuiling", "" ], [ "Lu", "Yan", "" ], [ "Zheng", "Nanning", "" ] ]
Significant progress has been made in scene text detection models since the rise of deep learning, but scene text layout analysis, which aims to group detected text instances as paragraphs, has not kept pace. Previous works either treated text detection and grouping using separate models, or train a model from scratch while using a unified one. All of them have not yet made full use of the already well-trained text detectors and easily obtainable detection datasets. In this paper, we present Text Grouping Adapter (TGA), a module that can enable the utilization of various pre-trained text detectors to learn layout analysis, allowing us to adopt a well-trained text detector right off the shelf or just fine-tune it efficiently. Designed to be compatible with various text detector architectures, TGA takes detected text regions and image features as universal inputs to assemble text instance features. To capture broader contextual information for layout analysis, we propose to predict text group masks from text instance features by one-to-many assignment. Our comprehensive experiments demonstrate that, even with frozen pre-trained models, incorporating our TGA into various pre-trained text detectors and text spotters can achieve superior layout analysis performance, simultaneously inheriting generalized text detection ability from pre-training. In the case of full parameter fine-tuning, we can further improve layout analysis performance.
2308.16635
Jin Liu
Jin Liu, Xi Wang, Xiaomeng Fu, Yesheng Chai, Cai Yu, Jiao Dai, Jizhong Han
MFR-Net: Multi-faceted Responsive Listening Head Generation via Denoising Diffusion Model
Accepted by ACM MM 2023
null
null
null
cs.CV
http://creativecommons.org/licenses/by-nc-sa/4.0/
Face-to-face communication is a common scenario including roles of speakers and listeners. Most existing research methods focus on producing speaker videos, while the generation of listener heads remains largely overlooked. Responsive listening head generation is an important task that aims to model face-to-face communication scenarios by generating a listener head video given a speaker video and a listener head image. An ideal generated responsive listening video should respond to the speaker with attitude or viewpoint expressing while maintaining diversity in interaction patterns and accuracy in listener identity information. To achieve this goal, we propose the \textbf{M}ulti-\textbf{F}aceted \textbf{R}esponsive Listening Head Generation Network (MFR-Net). Specifically, MFR-Net employs the probabilistic denoising diffusion model to predict diverse head pose and expression features. In order to perform multi-faceted response to the speaker video, while maintaining accurate listener identity preservation, we design the Feature Aggregation Module to boost listener identity features and fuse them with other speaker-related features. Finally, a renderer finetuned with identity consistency loss produces the final listening head videos. Our extensive experiments demonstrate that MFR-Net not only achieves multi-faceted responses in diversity and speaker identity information but also in attitude and viewpoint expression.
[ { "created": "Thu, 31 Aug 2023 11:10:28 GMT", "version": "v1" } ]
2023-09-01
[ [ "Liu", "Jin", "" ], [ "Wang", "Xi", "" ], [ "Fu", "Xiaomeng", "" ], [ "Chai", "Yesheng", "" ], [ "Yu", "Cai", "" ], [ "Dai", "Jiao", "" ], [ "Han", "Jizhong", "" ] ]
Face-to-face communication is a common scenario including roles of speakers and listeners. Most existing research methods focus on producing speaker videos, while the generation of listener heads remains largely overlooked. Responsive listening head generation is an important task that aims to model face-to-face communication scenarios by generating a listener head video given a speaker video and a listener head image. An ideal generated responsive listening video should respond to the speaker with attitude or viewpoint expressing while maintaining diversity in interaction patterns and accuracy in listener identity information. To achieve this goal, we propose the \textbf{M}ulti-\textbf{F}aceted \textbf{R}esponsive Listening Head Generation Network (MFR-Net). Specifically, MFR-Net employs the probabilistic denoising diffusion model to predict diverse head pose and expression features. In order to perform multi-faceted response to the speaker video, while maintaining accurate listener identity preservation, we design the Feature Aggregation Module to boost listener identity features and fuse them with other speaker-related features. Finally, a renderer finetuned with identity consistency loss produces the final listening head videos. Our extensive experiments demonstrate that MFR-Net not only achieves multi-faceted responses in diversity and speaker identity information but also in attitude and viewpoint expression.
2209.12248
Rui He
Rui He, Zehua Fu, Qingjie Liu, Yunhong Wang, Xunxun Chen
D$^{\bf{3}}$: Duplicate Detection Decontaminator for Multi-Athlete Tracking in Sports Videos
null
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Tracking multiple athletes in sports videos is a very challenging Multi-Object Tracking (MOT) task, since athletes often have the same appearance and are intimately covered with each other, making a common occlusion problem becomes an abhorrent duplicate detection. In this paper, the duplicate detection is newly and precisely defined as occlusion misreporting on the same athlete by multiple detection boxes in one frame. To address this problem, we meticulously design a novel transformer-based Duplicate Detection Decontaminator (D$^3$) for training, and a specific algorithm Rally-Hungarian (RH) for matching. Once duplicate detection occurs, D$^3$ immediately modifies the procedure by generating enhanced boxes losses. RH, triggered by the team sports substitution rules, is exceedingly suitable for sports videos. Moreover, to complement the tracking dataset that without shot changes, we release a new dataset based on sports video named RallyTrack. Extensive experiments on RallyTrack show that combining D$^3$ and RH can dramatically improve the tracking performance with 9.2 in MOTA and 4.5 in HOTA. Meanwhile, experiments on MOT-series and DanceTrack discover that D$^3$ can accelerate convergence during training, especially save up to 80 percent of the original training time on MOT17. Finally, our model, which is trained only with volleyball videos, can be applied directly to basketball and soccer videos for MAT, which shows priority of our method. Our dataset is available at https://github.com/heruihr/rallytrack.
[ { "created": "Sun, 25 Sep 2022 15:46:39 GMT", "version": "v1" } ]
2022-09-27
[ [ "He", "Rui", "" ], [ "Fu", "Zehua", "" ], [ "Liu", "Qingjie", "" ], [ "Wang", "Yunhong", "" ], [ "Chen", "Xunxun", "" ] ]
Tracking multiple athletes in sports videos is a very challenging Multi-Object Tracking (MOT) task, since athletes often have the same appearance and are intimately covered with each other, making a common occlusion problem becomes an abhorrent duplicate detection. In this paper, the duplicate detection is newly and precisely defined as occlusion misreporting on the same athlete by multiple detection boxes in one frame. To address this problem, we meticulously design a novel transformer-based Duplicate Detection Decontaminator (D$^3$) for training, and a specific algorithm Rally-Hungarian (RH) for matching. Once duplicate detection occurs, D$^3$ immediately modifies the procedure by generating enhanced boxes losses. RH, triggered by the team sports substitution rules, is exceedingly suitable for sports videos. Moreover, to complement the tracking dataset that without shot changes, we release a new dataset based on sports video named RallyTrack. Extensive experiments on RallyTrack show that combining D$^3$ and RH can dramatically improve the tracking performance with 9.2 in MOTA and 4.5 in HOTA. Meanwhile, experiments on MOT-series and DanceTrack discover that D$^3$ can accelerate convergence during training, especially save up to 80 percent of the original training time on MOT17. Finally, our model, which is trained only with volleyball videos, can be applied directly to basketball and soccer videos for MAT, which shows priority of our method. Our dataset is available at https://github.com/heruihr/rallytrack.
1802.01958
Philipp Harzig
Philipp Harzig, Stephan Brehm, Rainer Lienhart, Carolin Kaiser, Ren\'e Schallner
Multimodal Image Captioning for Marketing Analysis
4 pages, 1 figure, accepted at MIPR2018
null
10.1109/MIPR.2018.00035
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Automatically captioning images with natural language sentences is an important research topic. State of the art models are able to produce human-like sentences. These models typically describe the depicted scene as a whole and do not target specific objects of interest or emotional relationships between these objects in the image. However, marketing companies require to describe these important attributes of a given scene. In our case, objects of interest are consumer goods, which are usually identifiable by a product logo and are associated with certain brands. From a marketing point of view, it is desirable to also evaluate the emotional context of a trademarked product, i.e., whether it appears in a positive or a negative connotation. We address the problem of finding brands in images and deriving corresponding captions by introducing a modified image captioning network. We also add a third output modality, which simultaneously produces real-valued image ratings. Our network is trained using a classification-aware loss function in order to stimulate the generation of sentences with an emphasis on words identifying the brand of a product. We evaluate our model on a dataset of images depicting interactions between humans and branded products. The introduced network improves mean class accuracy by 24.5 percent. Thanks to adding the third output modality, it also considerably improves the quality of generated captions for images depicting branded products.
[ { "created": "Tue, 6 Feb 2018 14:23:32 GMT", "version": "v1" }, { "created": "Mon, 6 May 2019 10:35:53 GMT", "version": "v2" } ]
2019-08-07
[ [ "Harzig", "Philipp", "" ], [ "Brehm", "Stephan", "" ], [ "Lienhart", "Rainer", "" ], [ "Kaiser", "Carolin", "" ], [ "Schallner", "René", "" ] ]
Automatically captioning images with natural language sentences is an important research topic. State of the art models are able to produce human-like sentences. These models typically describe the depicted scene as a whole and do not target specific objects of interest or emotional relationships between these objects in the image. However, marketing companies require to describe these important attributes of a given scene. In our case, objects of interest are consumer goods, which are usually identifiable by a product logo and are associated with certain brands. From a marketing point of view, it is desirable to also evaluate the emotional context of a trademarked product, i.e., whether it appears in a positive or a negative connotation. We address the problem of finding brands in images and deriving corresponding captions by introducing a modified image captioning network. We also add a third output modality, which simultaneously produces real-valued image ratings. Our network is trained using a classification-aware loss function in order to stimulate the generation of sentences with an emphasis on words identifying the brand of a product. We evaluate our model on a dataset of images depicting interactions between humans and branded products. The introduced network improves mean class accuracy by 24.5 percent. Thanks to adding the third output modality, it also considerably improves the quality of generated captions for images depicting branded products.
2007.11348
Ori Shapira
Ori Shapira and Ran Levy
Massive Multi-Document Summarization of Product Reviews with Weak Supervision
null
null
null
null
cs.CL
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Product reviews summarization is a type of Multi-Document Summarization (MDS) task in which the summarized document sets are often far larger than in traditional MDS (up to tens of thousands of reviews). We highlight this difference and coin the term "Massive Multi-Document Summarization" (MMDS) to denote an MDS task that involves hundreds of documents or more. Prior work on product reviews summarization considered small samples of the reviews, mainly due to the difficulty of handling massive document sets. We show that summarizing small samples can result in loss of important information and provide misleading evaluation results. We propose a schema for summarizing a massive set of reviews on top of a standard summarization algorithm. Since writing large volumes of reference summaries needed for advanced neural network models is impractical, our solution relies on weak supervision. Finally, we propose an evaluation scheme that is based on multiple crowdsourced reference summaries and aims to capture the massive review collection. We show that an initial implementation of our schema significantly improves over several baselines in ROUGE scores, and exhibits strong coherence in a manual linguistic quality assessment.
[ { "created": "Wed, 22 Jul 2020 11:22:57 GMT", "version": "v1" } ]
2020-07-23
[ [ "Shapira", "Ori", "" ], [ "Levy", "Ran", "" ] ]
Product reviews summarization is a type of Multi-Document Summarization (MDS) task in which the summarized document sets are often far larger than in traditional MDS (up to tens of thousands of reviews). We highlight this difference and coin the term "Massive Multi-Document Summarization" (MMDS) to denote an MDS task that involves hundreds of documents or more. Prior work on product reviews summarization considered small samples of the reviews, mainly due to the difficulty of handling massive document sets. We show that summarizing small samples can result in loss of important information and provide misleading evaluation results. We propose a schema for summarizing a massive set of reviews on top of a standard summarization algorithm. Since writing large volumes of reference summaries needed for advanced neural network models is impractical, our solution relies on weak supervision. Finally, we propose an evaluation scheme that is based on multiple crowdsourced reference summaries and aims to capture the massive review collection. We show that an initial implementation of our schema significantly improves over several baselines in ROUGE scores, and exhibits strong coherence in a manual linguistic quality assessment.
1503.05157
Jeremy Debattista
Jeremy Debattista, Santiago Londo\~no, Christoph Lange, S\"oren Auer
Quality Assessment of Linked Datasets using Probabilistic Approximation
15 pages, 2 figures, To appear in ESWC 2015 proceedings
null
null
null
cs.DB cs.DS
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
With the increasing application of Linked Open Data, assessing the quality of datasets by computing quality metrics becomes an issue of crucial importance. For large and evolving datasets, an exact, deterministic computation of the quality metrics is too time consuming or expensive. We employ probabilistic techniques such as Reservoir Sampling, Bloom Filters and Clustering Coefficient estimation for implementing a broad set of data quality metrics in an approximate but sufficiently accurate way. Our implementation is integrated in the comprehensive data quality assessment framework Luzzu. We evaluated its performance and accuracy on Linked Open Datasets of broad relevance.
[ { "created": "Tue, 17 Mar 2015 18:39:22 GMT", "version": "v1" } ]
2015-03-18
[ [ "Debattista", "Jeremy", "" ], [ "Londoño", "Santiago", "" ], [ "Lange", "Christoph", "" ], [ "Auer", "Sören", "" ] ]
With the increasing application of Linked Open Data, assessing the quality of datasets by computing quality metrics becomes an issue of crucial importance. For large and evolving datasets, an exact, deterministic computation of the quality metrics is too time consuming or expensive. We employ probabilistic techniques such as Reservoir Sampling, Bloom Filters and Clustering Coefficient estimation for implementing a broad set of data quality metrics in an approximate but sufficiently accurate way. Our implementation is integrated in the comprehensive data quality assessment framework Luzzu. We evaluated its performance and accuracy on Linked Open Datasets of broad relevance.
2102.12606
Nahyun Kwon
Nahyun Kwon, Chen Liang and Jeeeun Kim
3D4ALL: Toward an Inclusive Pipeline to Classify 3D Contents
9 pages, 2 figures, TExSS, ACM IUI 2021 Workshops
null
null
null
cs.HC
http://creativecommons.org/licenses/by/4.0/
Algorithmic content moderation manages an explosive number of user-created content shared online everyday. Despite a massive number of 3D designs that are free to be downloaded, shared, and 3D printed by the users, detecting sensitivity with transparency and fairness has been controversial. Although sensitive 3D content might have a greater impact than other media due to its possible reproducibility and replicability without restriction, prevailed unawareness resulted in proliferation of sensitive 3D models online and a lack of discussion on transparent and fair 3D content moderation. As the 3D content exists as a document on the web mainly consisting of text and images, we first study the existing algorithmic efforts based on text and images and the prior endeavors to encompass transparency and fairness in moderation, which can also be useful in a 3D printing domain. At the same time, we identify 3D specific features that should be addressed to advance a 3D specialized algorithmic moderation. As a potential solution, we suggest a human-in-the-loop pipeline using augmented learning, powered by various stakeholders with different backgrounds and perspectives in understanding the content. Our pipeline aims to minimize personal biases by enabling diverse stakeholders to be vocal in reflecting various factors to interpret the content. We add our initial proposal for redesigning metadata of open 3D repositories, to invoke users' responsible actions of being granted consent from the subject upon sharing contents for free in the public spaces.
[ { "created": "Wed, 24 Feb 2021 23:58:07 GMT", "version": "v1" } ]
2021-02-26
[ [ "Kwon", "Nahyun", "" ], [ "Liang", "Chen", "" ], [ "Kim", "Jeeeun", "" ] ]
Algorithmic content moderation manages an explosive number of user-created content shared online everyday. Despite a massive number of 3D designs that are free to be downloaded, shared, and 3D printed by the users, detecting sensitivity with transparency and fairness has been controversial. Although sensitive 3D content might have a greater impact than other media due to its possible reproducibility and replicability without restriction, prevailed unawareness resulted in proliferation of sensitive 3D models online and a lack of discussion on transparent and fair 3D content moderation. As the 3D content exists as a document on the web mainly consisting of text and images, we first study the existing algorithmic efforts based on text and images and the prior endeavors to encompass transparency and fairness in moderation, which can also be useful in a 3D printing domain. At the same time, we identify 3D specific features that should be addressed to advance a 3D specialized algorithmic moderation. As a potential solution, we suggest a human-in-the-loop pipeline using augmented learning, powered by various stakeholders with different backgrounds and perspectives in understanding the content. Our pipeline aims to minimize personal biases by enabling diverse stakeholders to be vocal in reflecting various factors to interpret the content. We add our initial proposal for redesigning metadata of open 3D repositories, to invoke users' responsible actions of being granted consent from the subject upon sharing contents for free in the public spaces.
2204.11965
Mohamed Raed El Aoun
Mohamed Raed El aoun, Heng Li, Foutse Khomh, Lionel Tidjon
Bug Characteristics in Quantum Software Ecosystem
null
null
null
null
cs.SE
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
With the advance in quantum computing in recent years, quantum software becomes vital for exploring the full potential of quantum computing systems. Quantum programming is different from classical programming, for example, the state of a quantum program is probabilistic in nature, and a quantum computer is error-prone due to the instability of quantum mechanisms. Therefore, the characteristics of bugs in quantum software projects may be very different from that of classical software projects. This work aims to understand the characteristics of bugs in quantum software projects, in order to provide insights to help devise effective testing and debugging mechanisms. To achieve this goal, we conduct an empirical study on the bug reports of 125 quantum software projects. We observe that quantum software projects are more buggy than classical software projects and that quantum project bugs are more costly to fix than classical project bugs. We also identify the types of the bugs and the quantum programming components where they occurred. Our study shows that the bugs are spread across different components, but quantum-specific bugs particularly appear in the compiler, gate operation, and state preparation components. The three most occurring types of bugs are Program anomaly bugs, Configuration bugs, and Data type and structure bugs. Our study highlights some particularly challenging areas in quantum software development, such as the lack of scientific quantum computation libraries that implement comprehensive mathematical functions for quantum computing. Quantum developers also seek specialized data manipulation libraries for quantum software engineering like Numpy for quantum computing. Our findings also provide insights for future work to advance the quantum program development, testing, and debugging of quantum software, such as providing tooling support for debugging low-level circuits.
[ { "created": "Mon, 25 Apr 2022 20:59:46 GMT", "version": "v1" } ]
2022-04-27
[ [ "aoun", "Mohamed Raed El", "" ], [ "Li", "Heng", "" ], [ "Khomh", "Foutse", "" ], [ "Tidjon", "Lionel", "" ] ]
With the advance in quantum computing in recent years, quantum software becomes vital for exploring the full potential of quantum computing systems. Quantum programming is different from classical programming, for example, the state of a quantum program is probabilistic in nature, and a quantum computer is error-prone due to the instability of quantum mechanisms. Therefore, the characteristics of bugs in quantum software projects may be very different from that of classical software projects. This work aims to understand the characteristics of bugs in quantum software projects, in order to provide insights to help devise effective testing and debugging mechanisms. To achieve this goal, we conduct an empirical study on the bug reports of 125 quantum software projects. We observe that quantum software projects are more buggy than classical software projects and that quantum project bugs are more costly to fix than classical project bugs. We also identify the types of the bugs and the quantum programming components where they occurred. Our study shows that the bugs are spread across different components, but quantum-specific bugs particularly appear in the compiler, gate operation, and state preparation components. The three most occurring types of bugs are Program anomaly bugs, Configuration bugs, and Data type and structure bugs. Our study highlights some particularly challenging areas in quantum software development, such as the lack of scientific quantum computation libraries that implement comprehensive mathematical functions for quantum computing. Quantum developers also seek specialized data manipulation libraries for quantum software engineering like Numpy for quantum computing. Our findings also provide insights for future work to advance the quantum program development, testing, and debugging of quantum software, such as providing tooling support for debugging low-level circuits.
2307.09742
Guanbin Li
Ganlong Zhao, Guanbin Li, Yipeng Qin, Yizhou Yu
Improved Distribution Matching for Dataset Condensation
CVPR2023
null
null
null
cs.LG cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Dataset Condensation aims to condense a large dataset into a smaller one while maintaining its ability to train a well-performing model, thus reducing the storage cost and training effort in deep learning applications. However, conventional dataset condensation methods are optimization-oriented and condense the dataset by performing gradient or parameter matching during model optimization, which is computationally intensive even on small datasets and models. In this paper, we propose a novel dataset condensation method based on distribution matching, which is more efficient and promising. Specifically, we identify two important shortcomings of naive distribution matching (i.e., imbalanced feature numbers and unvalidated embeddings for distance computation) and address them with three novel techniques (i.e., partitioning and expansion augmentation, efficient and enriched model sampling, and class-aware distribution regularization). Our simple yet effective method outperforms most previous optimization-oriented methods with much fewer computational resources, thereby scaling data condensation to larger datasets and models. Extensive experiments demonstrate the effectiveness of our method. Codes are available at https://github.com/uitrbn/IDM
[ { "created": "Wed, 19 Jul 2023 04:07:33 GMT", "version": "v1" } ]
2023-07-20
[ [ "Zhao", "Ganlong", "" ], [ "Li", "Guanbin", "" ], [ "Qin", "Yipeng", "" ], [ "Yu", "Yizhou", "" ] ]
Dataset Condensation aims to condense a large dataset into a smaller one while maintaining its ability to train a well-performing model, thus reducing the storage cost and training effort in deep learning applications. However, conventional dataset condensation methods are optimization-oriented and condense the dataset by performing gradient or parameter matching during model optimization, which is computationally intensive even on small datasets and models. In this paper, we propose a novel dataset condensation method based on distribution matching, which is more efficient and promising. Specifically, we identify two important shortcomings of naive distribution matching (i.e., imbalanced feature numbers and unvalidated embeddings for distance computation) and address them with three novel techniques (i.e., partitioning and expansion augmentation, efficient and enriched model sampling, and class-aware distribution regularization). Our simple yet effective method outperforms most previous optimization-oriented methods with much fewer computational resources, thereby scaling data condensation to larger datasets and models. Extensive experiments demonstrate the effectiveness of our method. Codes are available at https://github.com/uitrbn/IDM
1406.1516
Zhiguo Ding
Zhiguo Ding and Zheng Yang and Pingzhi Fan and H. Vincent Poor
On the Performance of Non-Orthogonal Multiple Access in 5G Systems with Randomly Deployed Users
null
null
10.1109/LSP.2014.2343971
null
cs.IT math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In this letter, the performance of non-orthogonal multiple access (NOMA) is investigated in a cellular downlink scenario with randomly deployed users. The developed analytical results show that NOMA can achieve superior performance in terms of ergodic sum rates; however, the outage performance of NOMA depends critically on the choices of the users' targeted data rates and allocated power. In particular, a wrong choice of the targeted data rates and allocated power can lead to a situation in which the user's outage probability is always one, i.e. the user's targeted quality of service will never be met.
[ { "created": "Thu, 5 Jun 2014 20:36:58 GMT", "version": "v1" } ]
2015-06-19
[ [ "Ding", "Zhiguo", "" ], [ "Yang", "Zheng", "" ], [ "Fan", "Pingzhi", "" ], [ "Poor", "H. Vincent", "" ] ]
In this letter, the performance of non-orthogonal multiple access (NOMA) is investigated in a cellular downlink scenario with randomly deployed users. The developed analytical results show that NOMA can achieve superior performance in terms of ergodic sum rates; however, the outage performance of NOMA depends critically on the choices of the users' targeted data rates and allocated power. In particular, a wrong choice of the targeted data rates and allocated power can lead to a situation in which the user's outage probability is always one, i.e. the user's targeted quality of service will never be met.
1906.05797
Julian Straub
Julian Straub, Thomas Whelan, Lingni Ma, Yufan Chen, Erik Wijmans, Simon Green, Jakob J. Engel, Raul Mur-Artal, Carl Ren, Shobhit Verma, Anton Clarkson, Mingfei Yan, Brian Budge, Yajie Yan, Xiaqing Pan, June Yon, Yuyang Zou, Kimberly Leon, Nigel Carter, Jesus Briales, Tyler Gillingham, Elias Mueggler, Luis Pesqueira, Manolis Savva, Dhruv Batra, Hauke M. Strasdat, Renzo De Nardi, Michael Goesele, Steven Lovegrove, Richard Newcombe
The Replica Dataset: A Digital Replica of Indoor Spaces
null
null
null
null
cs.CV cs.GR eess.IV
http://creativecommons.org/licenses/by/4.0/
We introduce Replica, a dataset of 18 highly photo-realistic 3D indoor scene reconstructions at room and building scale. Each scene consists of a dense mesh, high-resolution high-dynamic-range (HDR) textures, per-primitive semantic class and instance information, and planar mirror and glass reflectors. The goal of Replica is to enable machine learning (ML) research that relies on visually, geometrically, and semantically realistic generative models of the world - for instance, egocentric computer vision, semantic segmentation in 2D and 3D, geometric inference, and the development of embodied agents (virtual robots) performing navigation, instruction following, and question answering. Due to the high level of realism of the renderings from Replica, there is hope that ML systems trained on Replica may transfer directly to real world image and video data. Together with the data, we are releasing a minimal C++ SDK as a starting point for working with the Replica dataset. In addition, Replica is `Habitat-compatible', i.e. can be natively used with AI Habitat for training and testing embodied agents.
[ { "created": "Thu, 13 Jun 2019 16:29:58 GMT", "version": "v1" } ]
2019-06-14
[ [ "Straub", "Julian", "" ], [ "Whelan", "Thomas", "" ], [ "Ma", "Lingni", "" ], [ "Chen", "Yufan", "" ], [ "Wijmans", "Erik", "" ], [ "Green", "Simon", "" ], [ "Engel", "Jakob J.", "" ], [ "Mur-Artal", "Raul", "" ], [ "Ren", "Carl", "" ], [ "Verma", "Shobhit", "" ], [ "Clarkson", "Anton", "" ], [ "Yan", "Mingfei", "" ], [ "Budge", "Brian", "" ], [ "Yan", "Yajie", "" ], [ "Pan", "Xiaqing", "" ], [ "Yon", "June", "" ], [ "Zou", "Yuyang", "" ], [ "Leon", "Kimberly", "" ], [ "Carter", "Nigel", "" ], [ "Briales", "Jesus", "" ], [ "Gillingham", "Tyler", "" ], [ "Mueggler", "Elias", "" ], [ "Pesqueira", "Luis", "" ], [ "Savva", "Manolis", "" ], [ "Batra", "Dhruv", "" ], [ "Strasdat", "Hauke M.", "" ], [ "De Nardi", "Renzo", "" ], [ "Goesele", "Michael", "" ], [ "Lovegrove", "Steven", "" ], [ "Newcombe", "Richard", "" ] ]
We introduce Replica, a dataset of 18 highly photo-realistic 3D indoor scene reconstructions at room and building scale. Each scene consists of a dense mesh, high-resolution high-dynamic-range (HDR) textures, per-primitive semantic class and instance information, and planar mirror and glass reflectors. The goal of Replica is to enable machine learning (ML) research that relies on visually, geometrically, and semantically realistic generative models of the world - for instance, egocentric computer vision, semantic segmentation in 2D and 3D, geometric inference, and the development of embodied agents (virtual robots) performing navigation, instruction following, and question answering. Due to the high level of realism of the renderings from Replica, there is hope that ML systems trained on Replica may transfer directly to real world image and video data. Together with the data, we are releasing a minimal C++ SDK as a starting point for working with the Replica dataset. In addition, Replica is `Habitat-compatible', i.e. can be natively used with AI Habitat for training and testing embodied agents.
2308.07783
Mohammad Baradaran
Mohammad Baradaran, Robert Bergevin
Future Video Prediction from a Single Frame for Video Anomaly Detection
null
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Video anomaly detection (VAD) is an important but challenging task in computer vision. The main challenge rises due to the rarity of training samples to model all anomaly cases. Hence, semi-supervised anomaly detection methods have gotten more attention, since they focus on modeling normals and they detect anomalies by measuring the deviations from normal patterns. Despite impressive advances of these methods in modeling normal motion and appearance, long-term motion modeling has not been effectively explored so far. Inspired by the abilities of the future frame prediction proxy-task, we introduce the task of future video prediction from a single frame, as a novel proxy-task for video anomaly detection. This proxy-task alleviates the challenges of previous methods in learning longer motion patterns. Moreover, we replace the initial and future raw frames with their corresponding semantic segmentation map, which not only makes the method aware of object class but also makes the prediction task less complex for the model. Extensive experiments on the benchmark datasets (ShanghaiTech, UCSD-Ped1, and UCSD-Ped2) show the effectiveness of the method and the superiority of its performance compared to SOTA prediction-based VAD methods.
[ { "created": "Tue, 15 Aug 2023 14:04:50 GMT", "version": "v1" } ]
2023-08-16
[ [ "Baradaran", "Mohammad", "" ], [ "Bergevin", "Robert", "" ] ]
Video anomaly detection (VAD) is an important but challenging task in computer vision. The main challenge rises due to the rarity of training samples to model all anomaly cases. Hence, semi-supervised anomaly detection methods have gotten more attention, since they focus on modeling normals and they detect anomalies by measuring the deviations from normal patterns. Despite impressive advances of these methods in modeling normal motion and appearance, long-term motion modeling has not been effectively explored so far. Inspired by the abilities of the future frame prediction proxy-task, we introduce the task of future video prediction from a single frame, as a novel proxy-task for video anomaly detection. This proxy-task alleviates the challenges of previous methods in learning longer motion patterns. Moreover, we replace the initial and future raw frames with their corresponding semantic segmentation map, which not only makes the method aware of object class but also makes the prediction task less complex for the model. Extensive experiments on the benchmark datasets (ShanghaiTech, UCSD-Ped1, and UCSD-Ped2) show the effectiveness of the method and the superiority of its performance compared to SOTA prediction-based VAD methods.
1809.00251
Adrian Viera
Leonardo Le\'on, Felipe Moreno-Vera, Renato Castro, Jos\'e Nav\'io, Marco Capcha
Car Monitoring System in Apartment Garages by Small Autonomous Car using Deep Learning
13 pages, 12 figures, Version 1 accepted in SimBig 2018. Improving to get better results
null
null
null
cs.CV cs.LG cs.RO
http://creativecommons.org/publicdomain/zero/1.0/
Currently, there is an increase in the number of Peruvian families living in apartments instead of houses for the lots of advantage; However, in some cases there are troubles such as robberies of goods that are usually left at the parking lots or the entrance of strangers that use the tenants parking lots (this last trouble sometimes is related to kidnappings or robberies in building apartments). Due to these problems, the use of a self-driving mini-car is proposed to implement a monitoring system of license plates in an underground garage inside a building using a deep learning model with the aim of recording the vehicles and identifying their owners if they were tenants or not. In addition, the small robot has its own location system using beacons that allow us to identify the position of the parking lot corresponding to each tenant of the building while the mini-car is on its way. Finally, one of the objectives of this work is to build a low-cost mini-robot that would replace expensive cameras or work together in order to keep safe the goods of tenants.
[ { "created": "Sat, 1 Sep 2018 21:00:58 GMT", "version": "v1" }, { "created": "Sat, 29 Sep 2018 01:00:32 GMT", "version": "v2" }, { "created": "Sat, 14 Sep 2019 21:40:42 GMT", "version": "v3" } ]
2019-09-17
[ [ "León", "Leonardo", "" ], [ "Moreno-Vera", "Felipe", "" ], [ "Castro", "Renato", "" ], [ "Navío", "José", "" ], [ "Capcha", "Marco", "" ] ]
Currently, there is an increase in the number of Peruvian families living in apartments instead of houses for the lots of advantage; However, in some cases there are troubles such as robberies of goods that are usually left at the parking lots or the entrance of strangers that use the tenants parking lots (this last trouble sometimes is related to kidnappings or robberies in building apartments). Due to these problems, the use of a self-driving mini-car is proposed to implement a monitoring system of license plates in an underground garage inside a building using a deep learning model with the aim of recording the vehicles and identifying their owners if they were tenants or not. In addition, the small robot has its own location system using beacons that allow us to identify the position of the parking lot corresponding to each tenant of the building while the mini-car is on its way. Finally, one of the objectives of this work is to build a low-cost mini-robot that would replace expensive cameras or work together in order to keep safe the goods of tenants.
2204.06736
EPTCS
Alexander Bolotov (University of Westminster)
On the Expressive Power of the Normal Form for Branching-Time Temporal Logics
In Proceedings NCL 2022, arXiv:2204.06359
EPTCS 358, 2022, pp. 254-269
10.4204/EPTCS.358.19
null
cs.FL cs.LO
http://creativecommons.org/licenses/by/4.0/
With the emerging applications that involve complex distributed systems branching-time specifications are specifically important as they reflect dynamic and non-deterministic nature of such applications. We describe the expressive power of a simple yet powerful branching-time specification framework -- branching-time normal form (BNF), which has been developed as part of clausal resolution for branching-time temporal logics. We show the encoding of Buchi Tree Automata in the language of the normal form, thus representing, syntactically, tree automata in a high-level way. Thus we can treat BNF as a normal form for the latter. These results enable us (1) to translate given problem specifications into the normal form and apply as a verification method a deductive reasoning technique -- the clausal temporal resolution; (2) to apply one of the core components of the resolution method -- the loop searching to extract, syntactically, hidden invariants in a wide range of complex temporal specifications.
[ { "created": "Thu, 14 Apr 2022 03:28:29 GMT", "version": "v1" } ]
2022-04-15
[ [ "Bolotov", "Alexander", "", "University of Westminster" ] ]
With the emerging applications that involve complex distributed systems branching-time specifications are specifically important as they reflect dynamic and non-deterministic nature of such applications. We describe the expressive power of a simple yet powerful branching-time specification framework -- branching-time normal form (BNF), which has been developed as part of clausal resolution for branching-time temporal logics. We show the encoding of Buchi Tree Automata in the language of the normal form, thus representing, syntactically, tree automata in a high-level way. Thus we can treat BNF as a normal form for the latter. These results enable us (1) to translate given problem specifications into the normal form and apply as a verification method a deductive reasoning technique -- the clausal temporal resolution; (2) to apply one of the core components of the resolution method -- the loop searching to extract, syntactically, hidden invariants in a wide range of complex temporal specifications.
2103.08720
Weilong Ren
Weilong Ren, Xiang Lian, Kambiz Ghazinour
Online Topic-Aware Entity Resolution Over Incomplete Data Streams (Technical Report)
Technical report of the paper entitled "Online Topic-Aware Entity Resolution Over Incomplete Data Streams", published on SIGMOD 2021
null
null
null
cs.DB
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In many real applications such as the data integration, social network analysis, and the Semantic Web, the entity resolution (ER) is an important and fundamental problem, which identifies and links the same real-world entities from various data sources. While prior works usually consider ER over static and complete data, in practice, application data are usually collected in a streaming fashion, and often incur missing attributes (due to the inaccuracy of data extraction techniques). Therefore, in this paper, we will formulate and tackle a novel problem, topic-aware entity resolution over incomplete data streams (TER-iDS), which online imputes incomplete tuples and detects pairs of topic-related matching entities from incomplete data streams. In order to effectively and efficiently tackle the TER-iDS problem, we propose an effective imputation strategy, carefully design effective pruning strategies, as well as indexes/synopsis, and develop an efficient TER-iDS algorithm via index joins. Extensive experiments have been conducted to evaluate the effectiveness and efficiency of our proposed TER-iDS approach over real data sets.
[ { "created": "Mon, 15 Mar 2021 21:06:12 GMT", "version": "v1" } ]
2021-03-17
[ [ "Ren", "Weilong", "" ], [ "Lian", "Xiang", "" ], [ "Ghazinour", "Kambiz", "" ] ]
In many real applications such as the data integration, social network analysis, and the Semantic Web, the entity resolution (ER) is an important and fundamental problem, which identifies and links the same real-world entities from various data sources. While prior works usually consider ER over static and complete data, in practice, application data are usually collected in a streaming fashion, and often incur missing attributes (due to the inaccuracy of data extraction techniques). Therefore, in this paper, we will formulate and tackle a novel problem, topic-aware entity resolution over incomplete data streams (TER-iDS), which online imputes incomplete tuples and detects pairs of topic-related matching entities from incomplete data streams. In order to effectively and efficiently tackle the TER-iDS problem, we propose an effective imputation strategy, carefully design effective pruning strategies, as well as indexes/synopsis, and develop an efficient TER-iDS algorithm via index joins. Extensive experiments have been conducted to evaluate the effectiveness and efficiency of our proposed TER-iDS approach over real data sets.
2109.09300
Jieming Zhou
Jieming Zhou, Tong Zhang, Pengfei Fang, Lars Petersson, Mehrtash Harandi
Feature Correlation Aggregation: on the Path to Better Graph Neural Networks
null
null
null
null
cs.LG cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Prior to the introduction of Graph Neural Networks (GNNs), modeling and analyzing irregular data, particularly graphs, was thought to be the Achilles' heel of deep learning. The core concept of GNNs is to find a representation by recursively aggregating the representations of a central node and those of its neighbors. The core concept of GNNs is to find a representation by recursively aggregating the representations of a central node and those of its neighbor, and its success has been demonstrated by many GNNs' designs. However, most of them only focus on using the first-order information between a node and its neighbors. In this paper, we introduce a central node permutation variant function through a frustratingly simple and innocent-looking modification to the core operation of a GNN, namely the Feature cOrrelation aGgregation (FOG) module which learns the second-order information from feature correlation between a node and its neighbors in the pipeline. By adding FOG into existing variants of GNNs, we empirically verify this second-order information complements the features generated by original GNNs across a broad set of benchmarks. A tangible boost in performance of the model is observed where the model surpasses previous state-of-the-art results by a significant margin while employing fewer parameters. (e.g., 33.116% improvement on a real-world molecular dataset using graph convolutional networks).
[ { "created": "Mon, 20 Sep 2021 05:04:26 GMT", "version": "v1" } ]
2021-09-22
[ [ "Zhou", "Jieming", "" ], [ "Zhang", "Tong", "" ], [ "Fang", "Pengfei", "" ], [ "Petersson", "Lars", "" ], [ "Harandi", "Mehrtash", "" ] ]
Prior to the introduction of Graph Neural Networks (GNNs), modeling and analyzing irregular data, particularly graphs, was thought to be the Achilles' heel of deep learning. The core concept of GNNs is to find a representation by recursively aggregating the representations of a central node and those of its neighbors. The core concept of GNNs is to find a representation by recursively aggregating the representations of a central node and those of its neighbor, and its success has been demonstrated by many GNNs' designs. However, most of them only focus on using the first-order information between a node and its neighbors. In this paper, we introduce a central node permutation variant function through a frustratingly simple and innocent-looking modification to the core operation of a GNN, namely the Feature cOrrelation aGgregation (FOG) module which learns the second-order information from feature correlation between a node and its neighbors in the pipeline. By adding FOG into existing variants of GNNs, we empirically verify this second-order information complements the features generated by original GNNs across a broad set of benchmarks. A tangible boost in performance of the model is observed where the model surpasses previous state-of-the-art results by a significant margin while employing fewer parameters. (e.g., 33.116% improvement on a real-world molecular dataset using graph convolutional networks).
2311.14704
Ezequiel Santos
David de Oliveira Lemes, Ezequiel Fran\c{c}a dos Santos, Eduardo Romanek, Celso Fujimoto, Adriano Felix Valente
An\'alise e modelagem de jogos digitais: Relato de uma experi\^encia educacional utlizando PBL em um grupo multidisciplinar
in Portuguese language
null
null
null
cs.CY
http://creativecommons.org/licenses/by/4.0/
Traditional software engineering education generally emphasizes strict collaboration and technical skills However active teaching strategies where students actively engage with the material transitioning from passive observers to active manipulators of realworld tools have shown effectiveness in software engineering The evolving market demands new skills in the context of digital transformation presenting challenges such as modeling complex business scenarios and navigating the interconnections between people systems and technologies Shifting from conventional software engineering instruction to active methodologies like ProblemBased Learning PBL has proven to bring realworld market challenges and realities into the classroom This article details an experience from the Digital Games Analysis and Modeling course in the Digital Games Masters program at Pontifical Catholic University of Sao Paulo It covers the discussed concepts case study rolebased work method and steps of the meetings We also present examples of outcomes like requirement diagrams context diagrams use case diagrams class diagrams interviews and others that contributed to the Game Design Document GDD These were created by each group during the meetings alongside their game prototypes Additionally a discussion on the developed capabilities is included
[ { "created": "Sat, 11 Nov 2023 20:28:51 GMT", "version": "v1" } ]
2023-11-28
[ [ "Lemes", "David de Oliveira", "" ], [ "Santos", "Ezequiel França dos", "" ], [ "Romanek", "Eduardo", "" ], [ "Fujimoto", "Celso", "" ], [ "Valente", "Adriano Felix", "" ] ]
Traditional software engineering education generally emphasizes strict collaboration and technical skills However active teaching strategies where students actively engage with the material transitioning from passive observers to active manipulators of realworld tools have shown effectiveness in software engineering The evolving market demands new skills in the context of digital transformation presenting challenges such as modeling complex business scenarios and navigating the interconnections between people systems and technologies Shifting from conventional software engineering instruction to active methodologies like ProblemBased Learning PBL has proven to bring realworld market challenges and realities into the classroom This article details an experience from the Digital Games Analysis and Modeling course in the Digital Games Masters program at Pontifical Catholic University of Sao Paulo It covers the discussed concepts case study rolebased work method and steps of the meetings We also present examples of outcomes like requirement diagrams context diagrams use case diagrams class diagrams interviews and others that contributed to the Game Design Document GDD These were created by each group during the meetings alongside their game prototypes Additionally a discussion on the developed capabilities is included
2204.11046
Yueqi Xie
Yueqi Xie, Peilin Zhou, Sunghun Kim
Decoupled Side Information Fusion for Sequential Recommendation
Accepted to SIGIR 2022
null
null
null
cs.IR cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Side information fusion for sequential recommendation (SR) aims to effectively leverage various side information to enhance the performance of next-item prediction. Most state-of-the-art methods build on self-attention networks and focus on exploring various solutions to integrate the item embedding and side information embeddings before the attention layer. However, our analysis shows that the early integration of various types of embeddings limits the expressiveness of attention matrices due to a rank bottleneck and constrains the flexibility of gradients. Also, it involves mixed correlations among the different heterogeneous information resources, which brings extra disturbance to attention calculation. Motivated by this, we propose Decoupled Side Information Fusion for Sequential Recommendation (DIF-SR), which moves the side information from the input to the attention layer and decouples the attention calculation of various side information and item representation. We theoretically and empirically show that the proposed solution allows higher-rank attention matrices and flexible gradients to enhance the modeling capacity of side information fusion. Also, auxiliary attribute predictors are proposed to further activate the beneficial interaction between side information and item representation learning. Extensive experiments on four real-world datasets demonstrate that our proposed solution stably outperforms state-of-the-art SR models. Further studies show that our proposed solution can be readily incorporated into current attention-based SR models and significantly boost performance. Our source code is available at https://github.com/AIM-SE/DIF-SR.
[ { "created": "Sat, 23 Apr 2022 10:53:36 GMT", "version": "v1" } ]
2022-04-26
[ [ "Xie", "Yueqi", "" ], [ "Zhou", "Peilin", "" ], [ "Kim", "Sunghun", "" ] ]
Side information fusion for sequential recommendation (SR) aims to effectively leverage various side information to enhance the performance of next-item prediction. Most state-of-the-art methods build on self-attention networks and focus on exploring various solutions to integrate the item embedding and side information embeddings before the attention layer. However, our analysis shows that the early integration of various types of embeddings limits the expressiveness of attention matrices due to a rank bottleneck and constrains the flexibility of gradients. Also, it involves mixed correlations among the different heterogeneous information resources, which brings extra disturbance to attention calculation. Motivated by this, we propose Decoupled Side Information Fusion for Sequential Recommendation (DIF-SR), which moves the side information from the input to the attention layer and decouples the attention calculation of various side information and item representation. We theoretically and empirically show that the proposed solution allows higher-rank attention matrices and flexible gradients to enhance the modeling capacity of side information fusion. Also, auxiliary attribute predictors are proposed to further activate the beneficial interaction between side information and item representation learning. Extensive experiments on four real-world datasets demonstrate that our proposed solution stably outperforms state-of-the-art SR models. Further studies show that our proposed solution can be readily incorporated into current attention-based SR models and significantly boost performance. Our source code is available at https://github.com/AIM-SE/DIF-SR.
1012.5248
Sergey Verlan
Ion Petre, Sergey Verlan
Matrix Insertion-Deletion Systems
null
null
null
null
cs.FL cs.CC cs.CL cs.DM
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In this article, we consider for the first time the operations of insertion and deletion working in a matrix controlled manner. We show that, similarly as in the case of context-free productions, the computational power is strictly increased when using a matrix control: computational completeness can be obtained by systems with insertion or deletion rules involving at most two symbols in a contextual or in a context-free manner and using only binary matrices.
[ { "created": "Thu, 23 Dec 2010 17:00:40 GMT", "version": "v1" } ]
2010-12-24
[ [ "Petre", "Ion", "" ], [ "Verlan", "Sergey", "" ] ]
In this article, we consider for the first time the operations of insertion and deletion working in a matrix controlled manner. We show that, similarly as in the case of context-free productions, the computational power is strictly increased when using a matrix control: computational completeness can be obtained by systems with insertion or deletion rules involving at most two symbols in a contextual or in a context-free manner and using only binary matrices.
2401.09234
Alfredo Go\~ni Sarriguren
Alfredo Go\~ni Sarriguren
SARRIGUREN: a polynomial-time complete algorithm for random $k$-SAT with relatively dense clauses
23 pages, 2 figures, 8 tables, algorithms, results and data in http://bdi.si.ehu.es/bdi/sarriguren
null
null
null
cs.DS cs.CC
http://creativecommons.org/licenses/by-nc-sa/4.0/
SARRIGUREN, a new complete algorithm for SAT based on counting clauses (which is valid also for Unique-SAT and #SAT) is described, analyzed and tested. Although existing complete algorithms for SAT perform slower with clauses with many literals, that is an advantage for SARRIGUREN, because the more literals are in the clauses the bigger is the probability of overlapping among clauses, a property that makes the clause counting process more efficient. Actually, it provides a $O(m^2 \times n/k)$ time complexity for random $k$-SAT instances of $n$ variables and $m$ relatively dense clauses, where that density level is relative to the number of variables $n$, that is, clauses are relatively dense when $k\geq7\sqrt{n}$. Although theoretically there could be worst-cases with exponential complexity, the probability of those cases to happen in random $k$-SAT with relatively dense clauses is practically zero. The algorithm has been empirically tested and that polynomial time complexity maintains also for $k$-SAT instances with less dense clauses ($k\geq5\sqrt{n}$). That density could, for example, be of only 0.049 working with $n=20000$ variables and $k=989$ literals. In addition, they are presented two more complementary algorithms that provide the solutions to $k$-SAT instances and valuable information about number of solutions for each literal. Although this algorithm does not solve the NP=P problem (it is not a polynomial algorithm for 3-SAT), it broads the knowledge about that subject, because $k$-SAT with $k>3$ and dense clauses is not harder than 3-SAT. Moreover, the Python implementation of the algorithms, and all the input datasets and obtained results in the experiments are made available.
[ { "created": "Wed, 17 Jan 2024 14:23:55 GMT", "version": "v1" } ]
2024-01-18
[ [ "Sarriguren", "Alfredo Goñi", "" ] ]
SARRIGUREN, a new complete algorithm for SAT based on counting clauses (which is valid also for Unique-SAT and #SAT) is described, analyzed and tested. Although existing complete algorithms for SAT perform slower with clauses with many literals, that is an advantage for SARRIGUREN, because the more literals are in the clauses the bigger is the probability of overlapping among clauses, a property that makes the clause counting process more efficient. Actually, it provides a $O(m^2 \times n/k)$ time complexity for random $k$-SAT instances of $n$ variables and $m$ relatively dense clauses, where that density level is relative to the number of variables $n$, that is, clauses are relatively dense when $k\geq7\sqrt{n}$. Although theoretically there could be worst-cases with exponential complexity, the probability of those cases to happen in random $k$-SAT with relatively dense clauses is practically zero. The algorithm has been empirically tested and that polynomial time complexity maintains also for $k$-SAT instances with less dense clauses ($k\geq5\sqrt{n}$). That density could, for example, be of only 0.049 working with $n=20000$ variables and $k=989$ literals. In addition, they are presented two more complementary algorithms that provide the solutions to $k$-SAT instances and valuable information about number of solutions for each literal. Although this algorithm does not solve the NP=P problem (it is not a polynomial algorithm for 3-SAT), it broads the knowledge about that subject, because $k$-SAT with $k>3$ and dense clauses is not harder than 3-SAT. Moreover, the Python implementation of the algorithms, and all the input datasets and obtained results in the experiments are made available.
2002.09107
Iretiayo Akinola
Iretiayo Akinola, Jacob Varley and Dmitry Kalashnikov
Learning Precise 3D Manipulation from Multiple Uncalibrated Cameras
Accepted at International Conference on Robotics and Automation (ICRA 2020)
null
null
null
cs.RO cs.CV cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In this work, we present an effective multi-view approach to closed-loop end-to-end learning of precise manipulation tasks that are 3D in nature. Our method learns to accomplish these tasks using multiple statically placed but uncalibrated RGB camera views without building an explicit 3D representation such as a pointcloud or voxel grid. This multi-camera approach achieves superior task performance on difficult stacking and insertion tasks compared to single-view baselines. Single view robotic agents struggle from occlusion and challenges in estimating relative poses between points of interest. While full 3D scene representations (voxels or pointclouds) are obtainable from registered output of multiple depth sensors, several challenges complicate operating off such explicit 3D representations. These challenges include imperfect camera calibration, poor depth maps due to object properties such as reflective surfaces, and slower inference speeds over 3D representations compared to 2D images. Our use of static but uncalibrated cameras does not require camera-robot or camera-camera calibration making the proposed approach easy to setup and our use of \textit{sensor dropout} during training makes it resilient to the loss of camera-views after deployment.
[ { "created": "Fri, 21 Feb 2020 03:28:42 GMT", "version": "v1" }, { "created": "Wed, 31 Mar 2021 18:48:24 GMT", "version": "v2" } ]
2021-04-02
[ [ "Akinola", "Iretiayo", "" ], [ "Varley", "Jacob", "" ], [ "Kalashnikov", "Dmitry", "" ] ]
In this work, we present an effective multi-view approach to closed-loop end-to-end learning of precise manipulation tasks that are 3D in nature. Our method learns to accomplish these tasks using multiple statically placed but uncalibrated RGB camera views without building an explicit 3D representation such as a pointcloud or voxel grid. This multi-camera approach achieves superior task performance on difficult stacking and insertion tasks compared to single-view baselines. Single view robotic agents struggle from occlusion and challenges in estimating relative poses between points of interest. While full 3D scene representations (voxels or pointclouds) are obtainable from registered output of multiple depth sensors, several challenges complicate operating off such explicit 3D representations. These challenges include imperfect camera calibration, poor depth maps due to object properties such as reflective surfaces, and slower inference speeds over 3D representations compared to 2D images. Our use of static but uncalibrated cameras does not require camera-robot or camera-camera calibration making the proposed approach easy to setup and our use of \textit{sensor dropout} during training makes it resilient to the loss of camera-views after deployment.
2011.11545
Xuhong Wang
Xuhong Wang, Ding Lyu, Mengjian Li, Yang Xia, Qi Yang, Xinwen Wang, Xinguang Wang, Ping Cui, Yupu Yang, Bowen Sun, Zhenyu Guo
APAN: Asynchronous Propagation Attention Network for Real-time Temporal Graph Embedding
In Proceedings of the 2021 International Conference on Management of Data (SIGMOD/PODS '21)
null
10.1145/3448016.3457564
null
cs.AI cs.DB cs.SI
http://creativecommons.org/licenses/by-nc-nd/4.0/
Limited by the time complexity of querying k-hop neighbors in a graph database, most graph algorithms cannot be deployed online and execute millisecond-level inference. This problem dramatically limits the potential of applying graph algorithms in certain areas, such as financial fraud detection. Therefore, we propose Asynchronous Propagation Attention Network, an asynchronous continuous time dynamic graph algorithm for real-time temporal graph embedding. Traditional graph models usually execute two serial operations: first graph computation and then model inference. We decouple model inference and graph computation step so that the heavy graph query operations will not damage the speed of model inference. Extensive experiments demonstrate that the proposed method can achieve competitive performance and 8.7 times inference speed improvement in the meantime.
[ { "created": "Mon, 23 Nov 2020 16:58:50 GMT", "version": "v1" }, { "created": "Fri, 27 Nov 2020 03:12:47 GMT", "version": "v2" }, { "created": "Wed, 16 Dec 2020 02:35:09 GMT", "version": "v3" }, { "created": "Fri, 26 Mar 2021 05:42:05 GMT", "version": "v4" } ]
2021-07-06
[ [ "Wang", "Xuhong", "" ], [ "Lyu", "Ding", "" ], [ "Li", "Mengjian", "" ], [ "Xia", "Yang", "" ], [ "Yang", "Qi", "" ], [ "Wang", "Xinwen", "" ], [ "Wang", "Xinguang", "" ], [ "Cui", "Ping", "" ], [ "Yang", "Yupu", "" ], [ "Sun", "Bowen", "" ], [ "Guo", "Zhenyu", "" ] ]
Limited by the time complexity of querying k-hop neighbors in a graph database, most graph algorithms cannot be deployed online and execute millisecond-level inference. This problem dramatically limits the potential of applying graph algorithms in certain areas, such as financial fraud detection. Therefore, we propose Asynchronous Propagation Attention Network, an asynchronous continuous time dynamic graph algorithm for real-time temporal graph embedding. Traditional graph models usually execute two serial operations: first graph computation and then model inference. We decouple model inference and graph computation step so that the heavy graph query operations will not damage the speed of model inference. Extensive experiments demonstrate that the proposed method can achieve competitive performance and 8.7 times inference speed improvement in the meantime.
1908.05905
Omar Sami Oubbati
Omar Sami Oubbati and Noureddine Chaib and Abderrahmane Lakas and Salim Bitam
On-Demand Routing for Urban VANETs using Cooperating UAVs
6 pages, 7 figures, conference
2018 International Conference on Smart Communications in Network Technologies (SaCoNeT)
10.1109/SaCoNeT.2018.8585453
null
cs.NI cs.IT cs.SI math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Vehicular ad hoc networks (VANETs) are characterized by frequent routing path failures due to the high mobility caused by the sudden changes of the direction of vehicles. The routing paths between two different vehicles should be established with this challenge in mind. Stability and connectedness are a mandatory condition to ensure a robust and reliable data delivery. The idea behind this work is to exploit a new reactive routing technique to provide regulated and well-connected routing paths. Unmanned Aerial Vehicles (UAVs) or what are referred to as drones can be both involved in the discovery process and be full members in these discovered paths in order to avoid possible disconnections on the ground when the network is sparsely connected. The different tests of this technique are performed based on NS-2 simulator and the outcomes are compared with those of related on-demand routing protocols dedicated for VANETs. Interesting results are distinguished showing a reduced end-to-end delay and a high delivery ratio, which proving that this heterogeneous communication between vehicles and UAVs is able to extend the network connectivity.
[ { "created": "Fri, 16 Aug 2019 09:21:05 GMT", "version": "v1" } ]
2019-08-19
[ [ "Oubbati", "Omar Sami", "" ], [ "Chaib", "Noureddine", "" ], [ "Lakas", "Abderrahmane", "" ], [ "Bitam", "Salim", "" ] ]
Vehicular ad hoc networks (VANETs) are characterized by frequent routing path failures due to the high mobility caused by the sudden changes of the direction of vehicles. The routing paths between two different vehicles should be established with this challenge in mind. Stability and connectedness are a mandatory condition to ensure a robust and reliable data delivery. The idea behind this work is to exploit a new reactive routing technique to provide regulated and well-connected routing paths. Unmanned Aerial Vehicles (UAVs) or what are referred to as drones can be both involved in the discovery process and be full members in these discovered paths in order to avoid possible disconnections on the ground when the network is sparsely connected. The different tests of this technique are performed based on NS-2 simulator and the outcomes are compared with those of related on-demand routing protocols dedicated for VANETs. Interesting results are distinguished showing a reduced end-to-end delay and a high delivery ratio, which proving that this heterogeneous communication between vehicles and UAVs is able to extend the network connectivity.