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
2306.09925
Daniel Gibert
Daniel Gibert, Jordi Planes, Quan Le, Giulio Zizzo
Query-Free Evasion Attacks Against Machine Learning-Based Malware Detectors with Generative Adversarial Networks
null
2023 IEEE European Symposium on Security and Privacy Workshops
10.1109/EuroSPW59978.2023.00052
null
cs.CR
http://creativecommons.org/licenses/by-nc-sa/4.0/
Malware detectors based on machine learning (ML) have been shown to be susceptible to adversarial malware examples. However, current methods to generate adversarial malware examples still have their limits. They either rely on detailed model information (gradient-based attacks), or on detailed outputs of the model - such as class probabilities (score-based attacks), neither of which are available in real-world scenarios. Alternatively, adversarial examples might be crafted using only the label assigned by the detector (label-based attack) to train a substitute network or an agent using reinforcement learning. Nonetheless, label-based attacks might require querying a black-box system from a small number to thousands of times, depending on the approach, which might not be feasible against malware detectors. This work presents a novel query-free approach to craft adversarial malware examples to evade ML-based malware detectors. To this end, we have devised a GAN-based framework to generate adversarial malware examples that look similar to benign executables in the feature space. To demonstrate the suitability of our approach we have applied the GAN-based attack to three common types of features usually employed by static ML-based malware detectors: (1) Byte histogram features, (2) API-based features, and (3) String-based features. Results show that our model-agnostic approach performs on par with MalGAN, while generating more realistic adversarial malware examples without requiring any query to the malware detectors. Furthermore, we have tested the generated adversarial examples against state-of-the-art multimodal and deep learning malware detectors, showing a decrease in detection performance, as well as a decrease in the average number of detections by the anti-malware engines in VirusTotal.
[ { "created": "Fri, 16 Jun 2023 15:48:40 GMT", "version": "v1" } ]
2023-08-22
[ [ "Gibert", "Daniel", "" ], [ "Planes", "Jordi", "" ], [ "Le", "Quan", "" ], [ "Zizzo", "Giulio", "" ] ]
Malware detectors based on machine learning (ML) have been shown to be susceptible to adversarial malware examples. However, current methods to generate adversarial malware examples still have their limits. They either rely on detailed model information (gradient-based attacks), or on detailed outputs of the model - such as class probabilities (score-based attacks), neither of which are available in real-world scenarios. Alternatively, adversarial examples might be crafted using only the label assigned by the detector (label-based attack) to train a substitute network or an agent using reinforcement learning. Nonetheless, label-based attacks might require querying a black-box system from a small number to thousands of times, depending on the approach, which might not be feasible against malware detectors. This work presents a novel query-free approach to craft adversarial malware examples to evade ML-based malware detectors. To this end, we have devised a GAN-based framework to generate adversarial malware examples that look similar to benign executables in the feature space. To demonstrate the suitability of our approach we have applied the GAN-based attack to three common types of features usually employed by static ML-based malware detectors: (1) Byte histogram features, (2) API-based features, and (3) String-based features. Results show that our model-agnostic approach performs on par with MalGAN, while generating more realistic adversarial malware examples without requiring any query to the malware detectors. Furthermore, we have tested the generated adversarial examples against state-of-the-art multimodal and deep learning malware detectors, showing a decrease in detection performance, as well as a decrease in the average number of detections by the anti-malware engines in VirusTotal.
2403.16178
Manisha Natarajan
Manisha Natarajan, Chunyue Xue, Sanne van Waveren, Karen Feigh, Matthew Gombolay
Mixed-Initiative Human-Robot Teaming under Suboptimality with Online Bayesian Adaptation
8 pages, 4 pages for supplementary
null
null
null
cs.RO cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
For effective human-agent teaming, robots and other artificial intelligence (AI) agents must infer their human partner's abilities and behavioral response patterns and adapt accordingly. Most prior works make the unrealistic assumption that one or more teammates can act near-optimally. In real-world collaboration, humans and autonomous agents can be suboptimal, especially when each only has partial domain knowledge. In this work, we develop computational modeling and optimization techniques for enhancing the performance of suboptimal human-agent teams, where the human and the agent have asymmetric capabilities and act suboptimally due to incomplete environmental knowledge. We adopt an online Bayesian approach that enables a robot to infer people's willingness to comply with its assistance in a sequential decision-making game. Our user studies show that user preferences and team performance indeed vary with robot intervention styles, and our approach for mixed-initiative collaborations enhances objective team performance ($p<.001$) and subjective measures, such as user's trust ($p<.001$) and perceived likeability of the robot ($p<.001$).
[ { "created": "Sun, 24 Mar 2024 14:38:18 GMT", "version": "v1" } ]
2024-03-26
[ [ "Natarajan", "Manisha", "" ], [ "Xue", "Chunyue", "" ], [ "van Waveren", "Sanne", "" ], [ "Feigh", "Karen", "" ], [ "Gombolay", "Matthew", "" ] ]
For effective human-agent teaming, robots and other artificial intelligence (AI) agents must infer their human partner's abilities and behavioral response patterns and adapt accordingly. Most prior works make the unrealistic assumption that one or more teammates can act near-optimally. In real-world collaboration, humans and autonomous agents can be suboptimal, especially when each only has partial domain knowledge. In this work, we develop computational modeling and optimization techniques for enhancing the performance of suboptimal human-agent teams, where the human and the agent have asymmetric capabilities and act suboptimally due to incomplete environmental knowledge. We adopt an online Bayesian approach that enables a robot to infer people's willingness to comply with its assistance in a sequential decision-making game. Our user studies show that user preferences and team performance indeed vary with robot intervention styles, and our approach for mixed-initiative collaborations enhances objective team performance ($p<.001$) and subjective measures, such as user's trust ($p<.001$) and perceived likeability of the robot ($p<.001$).
1912.07744
Siyuan Huang
Siyuan Huang, Yixin Chen, Tao Yuan, Siyuan Qi, Yixin Zhu, Song-Chun Zhu
PerspectiveNet: 3D Object Detection from a Single RGB Image via Perspective Points
NeurIPS 2019
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Detecting 3D objects from a single RGB image is intrinsically ambiguous, thus requiring appropriate prior knowledge and intermediate representations as constraints to reduce the uncertainties and improve the consistencies between the 2D image plane and the 3D world coordinate. To address this challenge, we propose to adopt perspective points as a new intermediate representation for 3D object detection, defined as the 2D projections of local Manhattan 3D keypoints to locate an object; these perspective points satisfy geometric constraints imposed by the perspective projection. We further devise PerspectiveNet, an end-to-end trainable model that simultaneously detects the 2D bounding box, 2D perspective points, and 3D object bounding box for each object from a single RGB image. PerspectiveNet yields three unique advantages: (i) 3D object bounding boxes are estimated based on perspective points, bridging the gap between 2D and 3D bounding boxes without the need of category-specific 3D shape priors. (ii) It predicts the perspective points by a template-based method, and a perspective loss is formulated to maintain the perspective constraints. (iii) It maintains the consistency between the 2D perspective points and 3D bounding boxes via a differentiable projective function. Experiments on SUN RGB-D dataset show that the proposed method significantly outperforms existing RGB-based approaches for 3D object detection.
[ { "created": "Mon, 16 Dec 2019 22:58:53 GMT", "version": "v1" } ]
2019-12-18
[ [ "Huang", "Siyuan", "" ], [ "Chen", "Yixin", "" ], [ "Yuan", "Tao", "" ], [ "Qi", "Siyuan", "" ], [ "Zhu", "Yixin", "" ], [ "Zhu", "Song-Chun", "" ] ]
Detecting 3D objects from a single RGB image is intrinsically ambiguous, thus requiring appropriate prior knowledge and intermediate representations as constraints to reduce the uncertainties and improve the consistencies between the 2D image plane and the 3D world coordinate. To address this challenge, we propose to adopt perspective points as a new intermediate representation for 3D object detection, defined as the 2D projections of local Manhattan 3D keypoints to locate an object; these perspective points satisfy geometric constraints imposed by the perspective projection. We further devise PerspectiveNet, an end-to-end trainable model that simultaneously detects the 2D bounding box, 2D perspective points, and 3D object bounding box for each object from a single RGB image. PerspectiveNet yields three unique advantages: (i) 3D object bounding boxes are estimated based on perspective points, bridging the gap between 2D and 3D bounding boxes without the need of category-specific 3D shape priors. (ii) It predicts the perspective points by a template-based method, and a perspective loss is formulated to maintain the perspective constraints. (iii) It maintains the consistency between the 2D perspective points and 3D bounding boxes via a differentiable projective function. Experiments on SUN RGB-D dataset show that the proposed method significantly outperforms existing RGB-based approaches for 3D object detection.
2405.09309
Bikash Kumar Dey
Abhishek Sarkar and Bikash Kumar Dey
Identification via Permutation Channels
9 pages. Extended and generalized version of submission to ITW 2024
null
null
null
cs.IT math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We study message identification over a $q$-ary uniform permutation channel, where the transmitted vector is permuted by a permutation chosen uniformly at random. For discrete memoryless channels (DMCs), the number of identifiable messages grows doubly exponentially. Identification capacity, the maximum second-order exponent, is known to be the same as the Shannon capacity of the DMC. Permutation channels support reliable communication of only polynomially many messages. A simple achievability result shows that message sizes growing as $2^{c_nn^{q-1}}$ are identifiable for any $c_n\rightarrow 0$. We prove two converse results. A ``soft'' converse shows that for any $R>0$, there is no sequence of identification codes with message size growing as $2^{Rn^{q-1}}$ with a power-law decay ($n^{-\mu}$) of the error probability. We also prove a ``strong" converse showing that for any sequence of identification codes with message size $2^{Rn^{q-1}\log n}$ ($R>0$), the sum of type I and type II error probabilities approaches at least $1$ as $n\rightarrow \infty$. To prove the soft converse, we use a sequence of steps to construct a new identification code with a simpler structure which relates to a set system, and then use a lower bound on the normalized maximum pairwise intersection of a set system. To prove the strong converse, we use results on approximation of distributions.
[ { "created": "Wed, 15 May 2024 13:07:35 GMT", "version": "v1" }, { "created": "Tue, 4 Jun 2024 11:58:02 GMT", "version": "v2" } ]
2024-06-05
[ [ "Sarkar", "Abhishek", "" ], [ "Dey", "Bikash Kumar", "" ] ]
We study message identification over a $q$-ary uniform permutation channel, where the transmitted vector is permuted by a permutation chosen uniformly at random. For discrete memoryless channels (DMCs), the number of identifiable messages grows doubly exponentially. Identification capacity, the maximum second-order exponent, is known to be the same as the Shannon capacity of the DMC. Permutation channels support reliable communication of only polynomially many messages. A simple achievability result shows that message sizes growing as $2^{c_nn^{q-1}}$ are identifiable for any $c_n\rightarrow 0$. We prove two converse results. A ``soft'' converse shows that for any $R>0$, there is no sequence of identification codes with message size growing as $2^{Rn^{q-1}}$ with a power-law decay ($n^{-\mu}$) of the error probability. We also prove a ``strong" converse showing that for any sequence of identification codes with message size $2^{Rn^{q-1}\log n}$ ($R>0$), the sum of type I and type II error probabilities approaches at least $1$ as $n\rightarrow \infty$. To prove the soft converse, we use a sequence of steps to construct a new identification code with a simpler structure which relates to a set system, and then use a lower bound on the normalized maximum pairwise intersection of a set system. To prove the strong converse, we use results on approximation of distributions.
2008.08311
Seokwoo Jung
Seokwoo Jung, Sungha Choi, Mohammad Azam Khan, Jaegul Choo
Towards Lightweight Lane Detection by Optimizing Spatial Embedding
Preprint - work in progress
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
A number of lane detection methods depend on a proposal-free instance segmentation because of its adaptability to flexible object shape, occlusion, and real-time application. This paper addresses the problem that pixel embedding in proposal-free instance segmentation based lane detection is difficult to optimize. A translation invariance of convolution, which is one of the supposed strengths, causes challenges in optimizing pixel embedding. In this work, we propose a lane detection method based on proposal-free instance segmentation, directly optimizing spatial embedding of pixels using image coordinate. Our proposed method allows the post-processing step for center localization and optimizes clustering in an end-to-end manner. The proposed method enables real-time lane detection through the simplicity of post-processing and the adoption of a lightweight backbone. Our proposed method demonstrates competitive performance on public lane detection datasets.
[ { "created": "Wed, 19 Aug 2020 07:37:04 GMT", "version": "v1" }, { "created": "Thu, 27 Aug 2020 06:45:20 GMT", "version": "v2" } ]
2020-08-28
[ [ "Jung", "Seokwoo", "" ], [ "Choi", "Sungha", "" ], [ "Khan", "Mohammad Azam", "" ], [ "Choo", "Jaegul", "" ] ]
A number of lane detection methods depend on a proposal-free instance segmentation because of its adaptability to flexible object shape, occlusion, and real-time application. This paper addresses the problem that pixel embedding in proposal-free instance segmentation based lane detection is difficult to optimize. A translation invariance of convolution, which is one of the supposed strengths, causes challenges in optimizing pixel embedding. In this work, we propose a lane detection method based on proposal-free instance segmentation, directly optimizing spatial embedding of pixels using image coordinate. Our proposed method allows the post-processing step for center localization and optimizes clustering in an end-to-end manner. The proposed method enables real-time lane detection through the simplicity of post-processing and the adoption of a lightweight backbone. Our proposed method demonstrates competitive performance on public lane detection datasets.
1311.1762
Tsvi Kopelowitz
Richard Cole, Tsvi Kopelowitz, Moshe Lewenstein
Suffix Trays and Suffix Trists: Structures for Faster Text Indexing
Results from this paper have appeared as an extended abstract in ICALP 2006
null
null
null
cs.DS
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Suffix trees and suffix arrays are two of the most widely used data structures for text indexing. Each uses linear space and can be constructed in linear time for polynomially sized alphabets. However, when it comes to answering queries with worst-case deterministic time bounds, the prior does so in $O(m\log|\Sigma|)$ time, where $m$ is the query size, $|\Sigma|$ is the alphabet size, and the latter does so in $O(m+\log n)$ time, where $n$ is the text size. If one wants to output all appearances of the query, an additive cost of $O(occ)$ time is sufficient, where $occ$ is the size of the output. We propose a novel way of combining the two into, what we call, a {\em suffix tray}. The space and construction time remain linear and the query time improves to $O(m+\log|\Sigma|)$ for integer alphabets from a linear range, i.e. $\Sigma \subset \{1,\cdots, cn\}$, for an arbitrary constant $c$. The construction and query are deterministic. Here also an additive $O(occ)$ time is sufficient if one desires to output all appearances of the query. We also consider the online version of indexing, where the text arrives online, one character at a time, and indexing queries are answered in tandem. In this variant we create a cross between a suffix tree and a suffix list (a dynamic variant of suffix array) to be called a {\em suffix trist}; it supports queries in $O(m+\log|\Sigma|)$ time. The suffix trist also uses linear space. Furthermore, if there exists an online construction for a linear-space suffix tree such that the cost of adding a character is worst-case deterministic $f(n,|\Sigma|)$ ($n$ is the size of the current text), then one can further update the suffix trist in $O(f(n,|\Sigma|)+\log |\Sigma|)$ time. The best currently known worst-case deterministic bound for $f(n,|\Sigma|)$ is $O(\log n)$ time.
[ { "created": "Thu, 7 Nov 2013 17:44:02 GMT", "version": "v1" } ]
2013-11-08
[ [ "Cole", "Richard", "" ], [ "Kopelowitz", "Tsvi", "" ], [ "Lewenstein", "Moshe", "" ] ]
Suffix trees and suffix arrays are two of the most widely used data structures for text indexing. Each uses linear space and can be constructed in linear time for polynomially sized alphabets. However, when it comes to answering queries with worst-case deterministic time bounds, the prior does so in $O(m\log|\Sigma|)$ time, where $m$ is the query size, $|\Sigma|$ is the alphabet size, and the latter does so in $O(m+\log n)$ time, where $n$ is the text size. If one wants to output all appearances of the query, an additive cost of $O(occ)$ time is sufficient, where $occ$ is the size of the output. We propose a novel way of combining the two into, what we call, a {\em suffix tray}. The space and construction time remain linear and the query time improves to $O(m+\log|\Sigma|)$ for integer alphabets from a linear range, i.e. $\Sigma \subset \{1,\cdots, cn\}$, for an arbitrary constant $c$. The construction and query are deterministic. Here also an additive $O(occ)$ time is sufficient if one desires to output all appearances of the query. We also consider the online version of indexing, where the text arrives online, one character at a time, and indexing queries are answered in tandem. In this variant we create a cross between a suffix tree and a suffix list (a dynamic variant of suffix array) to be called a {\em suffix trist}; it supports queries in $O(m+\log|\Sigma|)$ time. The suffix trist also uses linear space. Furthermore, if there exists an online construction for a linear-space suffix tree such that the cost of adding a character is worst-case deterministic $f(n,|\Sigma|)$ ($n$ is the size of the current text), then one can further update the suffix trist in $O(f(n,|\Sigma|)+\log |\Sigma|)$ time. The best currently known worst-case deterministic bound for $f(n,|\Sigma|)$ is $O(\log n)$ time.
1006.2805
Jenny Blight
Saeed Tavakoli and Amir Banookh
Robust PI Control Design Using Particle Swarm Optimization
Submitted to Journal of Computer Science and Engineering, see http://sites.google.com/site/jcseuk/volume-1-issue-1-may-2010
Journal of Computer Science and Engineering, Volume 1, Issue 1, p36-41, May 2010
null
null
cs.CE
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This paper presents a set of robust PI tuning formulae for a first order plus dead time process using particle swarm optimization. Also, tuning formulae for an integrating process with dead time, which is a special case of a first order plus dead time process, is given. The design problem considers three essential requirements of control problems, namely load disturbance rejection, setpoint regulation and robustness of closed-loop system against model uncertainties. The primary design goal is to optimize load disturbance rejection. Robustness is guaranteed by requiring that the maximum sensitivity is less than or equal to a specified value. In the first step, PI controller parameters are determined such that the IAE criterion to a load disturbance step is minimized and the robustness constraint on maximum sensitivity is satisfied. Using a structure with two degrees of freedom which introduces an extra parameter, the setpoint weight, good setpoint regulation is achieved in the second step. The main advantage of the proposed method is its simplicity. Once the equivalent first order plus dead time model is determined, the PI parameters are explicitly given by a set of tuning formulae. In order to show the performance and effectiveness of the proposed tuning formulae, they are applied to three simulation examples.
[ { "created": "Mon, 14 Jun 2010 19:03:53 GMT", "version": "v1" } ]
2010-06-15
[ [ "Tavakoli", "Saeed", "" ], [ "Banookh", "Amir", "" ] ]
This paper presents a set of robust PI tuning formulae for a first order plus dead time process using particle swarm optimization. Also, tuning formulae for an integrating process with dead time, which is a special case of a first order plus dead time process, is given. The design problem considers three essential requirements of control problems, namely load disturbance rejection, setpoint regulation and robustness of closed-loop system against model uncertainties. The primary design goal is to optimize load disturbance rejection. Robustness is guaranteed by requiring that the maximum sensitivity is less than or equal to a specified value. In the first step, PI controller parameters are determined such that the IAE criterion to a load disturbance step is minimized and the robustness constraint on maximum sensitivity is satisfied. Using a structure with two degrees of freedom which introduces an extra parameter, the setpoint weight, good setpoint regulation is achieved in the second step. The main advantage of the proposed method is its simplicity. Once the equivalent first order plus dead time model is determined, the PI parameters are explicitly given by a set of tuning formulae. In order to show the performance and effectiveness of the proposed tuning formulae, they are applied to three simulation examples.
1509.04674
Anastasios Papazafeiropoulos
Anastasios K. Papazafeiropoulos, Shree Krishna Sharma, and Symeon Chatzinotas
Impact of Transceiver Impairments on the Capacity of Dual-Hop Relay Massive MIMO Systems
6 pages, 4 figures, accepted in IEEE Global Communications Conference (GLOBECOM 2015) - Workshop on Massive MIMO: From theory to practice, 2015
null
10.1109/GLOCOMW.2015.7414137
null
cs.IT math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Despite the deleterious effect of hardware impairments on communication systems, most prior works have not investigated their impact on widely used relay systems. Most importantly, the application of inexpensive transceivers, being prone to hardware impairments, is the most cost-efficient way for the implementation of massive multiple-input multiple-output (MIMO) systems. Consequently, the direction of this paper is towards the investigation of the impact of hardware impairments on MIMO relay networks with large number of antennas. Specifically, we obtain the general expression for the ergodic capacity of dual-hop (DH) amplify-and-forward (AF) relay systems. Next, given the advantages of the free probability (FP) theory with comparison to other known techniques in the area of large random matrix theory, we pursue a large limit analysis in terms of number of antennas and users by shedding light to the behavior of relay systems inflicted by hardware impairments.
[ { "created": "Tue, 15 Sep 2015 18:45:52 GMT", "version": "v1" } ]
2016-11-17
[ [ "Papazafeiropoulos", "Anastasios K.", "" ], [ "Sharma", "Shree Krishna", "" ], [ "Chatzinotas", "Symeon", "" ] ]
Despite the deleterious effect of hardware impairments on communication systems, most prior works have not investigated their impact on widely used relay systems. Most importantly, the application of inexpensive transceivers, being prone to hardware impairments, is the most cost-efficient way for the implementation of massive multiple-input multiple-output (MIMO) systems. Consequently, the direction of this paper is towards the investigation of the impact of hardware impairments on MIMO relay networks with large number of antennas. Specifically, we obtain the general expression for the ergodic capacity of dual-hop (DH) amplify-and-forward (AF) relay systems. Next, given the advantages of the free probability (FP) theory with comparison to other known techniques in the area of large random matrix theory, we pursue a large limit analysis in terms of number of antennas and users by shedding light to the behavior of relay systems inflicted by hardware impairments.
1801.08754
Clare Llewellyn
Clare Llewellyn, Laura Cram, Adrian Favero, Robin L. Hill
For Whom the Bell Trolls: Troll Behaviour in the Twitter Brexit Debate
null
null
null
null
cs.SI cs.CY
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In a review into automated and malicious activity Twitter released a list of accounts that they believed were connected to state sponsored manipulation of the 2016 American Election. This list details 2,752 accounts Twitter believed to be controlled by Russian operatives. In the absence of a similar list of operatives active within the debate on the 2016 UK referendum on membership of the European Union (Brexit) we investigated the behaviour of the same American Election focused accounts in the production of content related to the UK-EU referendum. We found that within our dataset we had Brexit-related content from 419 of these accounts, leading to 3,485 identified tweets gathered between the 29th August 2015 and 3rd October 2017. The behaviour of the accounts altered radically on the day of the referendum, shifting from generalised disruptive tweeting to retweeting each other in order to amplify content produced by other troll accounts. We also demonstrate that, while these accounts are, in general, designed to resemble American citizens, accounts created in 2016 often contained German locations and terms in the user profiles.
[ { "created": "Fri, 26 Jan 2018 11:02:26 GMT", "version": "v1" } ]
2018-01-29
[ [ "Llewellyn", "Clare", "" ], [ "Cram", "Laura", "" ], [ "Favero", "Adrian", "" ], [ "Hill", "Robin L.", "" ] ]
In a review into automated and malicious activity Twitter released a list of accounts that they believed were connected to state sponsored manipulation of the 2016 American Election. This list details 2,752 accounts Twitter believed to be controlled by Russian operatives. In the absence of a similar list of operatives active within the debate on the 2016 UK referendum on membership of the European Union (Brexit) we investigated the behaviour of the same American Election focused accounts in the production of content related to the UK-EU referendum. We found that within our dataset we had Brexit-related content from 419 of these accounts, leading to 3,485 identified tweets gathered between the 29th August 2015 and 3rd October 2017. The behaviour of the accounts altered radically on the day of the referendum, shifting from generalised disruptive tweeting to retweeting each other in order to amplify content produced by other troll accounts. We also demonstrate that, while these accounts are, in general, designed to resemble American citizens, accounts created in 2016 often contained German locations and terms in the user profiles.
2309.15111
Margalit Glasgow
Margalit Glasgow
SGD Finds then Tunes Features in Two-Layer Neural Networks with near-Optimal Sample Complexity: A Case Study in the XOR problem
null
null
null
null
cs.LG stat.ML
http://creativecommons.org/licenses/by/4.0/
In this work, we consider the optimization process of minibatch stochastic gradient descent (SGD) on a 2-layer neural network with data separated by a quadratic ground truth function. We prove that with data drawn from the $d$-dimensional Boolean hypercube labeled by the quadratic ``XOR'' function $y = -x_ix_j$, it is possible to train to a population error $o(1)$ with $d \:\text{polylog}(d)$ samples. Our result considers simultaneously training both layers of the two-layer-neural network with ReLU activations via standard minibatch SGD on the logistic loss. To our knowledge, this work is the first to give a sample complexity of $\tilde{O}(d)$ for efficiently learning the XOR function on isotropic data on a standard neural network with standard training. Our main technique is showing that the network evolves in two phases: a $\textit{signal-finding}$ phase where the network is small and many of the neurons evolve independently to find features, and a $\textit{signal-heavy}$ phase, where SGD maintains and balances the features. We leverage the simultaneous training of the layers to show that it is sufficient for only a small fraction of the neurons to learn features, since those neurons will be amplified by the simultaneous growth of their second layer weights.
[ { "created": "Tue, 26 Sep 2023 17:57:44 GMT", "version": "v1" }, { "created": "Mon, 2 Oct 2023 14:21:45 GMT", "version": "v2" } ]
2023-10-03
[ [ "Glasgow", "Margalit", "" ] ]
In this work, we consider the optimization process of minibatch stochastic gradient descent (SGD) on a 2-layer neural network with data separated by a quadratic ground truth function. We prove that with data drawn from the $d$-dimensional Boolean hypercube labeled by the quadratic ``XOR'' function $y = -x_ix_j$, it is possible to train to a population error $o(1)$ with $d \:\text{polylog}(d)$ samples. Our result considers simultaneously training both layers of the two-layer-neural network with ReLU activations via standard minibatch SGD on the logistic loss. To our knowledge, this work is the first to give a sample complexity of $\tilde{O}(d)$ for efficiently learning the XOR function on isotropic data on a standard neural network with standard training. Our main technique is showing that the network evolves in two phases: a $\textit{signal-finding}$ phase where the network is small and many of the neurons evolve independently to find features, and a $\textit{signal-heavy}$ phase, where SGD maintains and balances the features. We leverage the simultaneous training of the layers to show that it is sufficient for only a small fraction of the neurons to learn features, since those neurons will be amplified by the simultaneous growth of their second layer weights.
2004.02753
Joshua Knights Mr
Joshua Knights, Ben Harwood, Daniel Ward, Anthony Vanderkop, Olivia Mackenzie-Ross, Peyman Moghadam
Temporally Coherent Embeddings for Self-Supervised Video Representation Learning
Accepted at ICPR 2020. Project page: https://csiro-robotics.github.io/TCE-Webpage/
null
null
null
cs.CV cs.LG eess.IV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This paper presents TCE: Temporally Coherent Embeddings for self-supervised video representation learning. The proposed method exploits inherent structure of unlabeled video data to explicitly enforce temporal coherency in the embedding space, rather than indirectly learning it through ranking or predictive proxy tasks. In the same way that high-level visual information in the world changes smoothly, we believe that nearby frames in learned representations will benefit from demonstrating similar properties. Using this assumption, we train our TCE model to encode videos such that adjacent frames exist close to each other and videos are separated from one another. Using TCE we learn robust representations from large quantities of unlabeled video data. We thoroughly analyse and evaluate our self-supervised learned TCE models on a downstream task of video action recognition using multiple challenging benchmarks (Kinetics400, UCF101, HMDB51). With a simple but effective 2D-CNN backbone and only RGB stream inputs, TCE pre-trained representations outperform all previous selfsupervised 2D-CNN and 3D-CNN pre-trained on UCF101. The code and pre-trained models for this paper can be downloaded at: https://github.com/csiro-robotics/TCE
[ { "created": "Sat, 21 Mar 2020 12:25:50 GMT", "version": "v1" }, { "created": "Fri, 1 May 2020 00:24:07 GMT", "version": "v2" }, { "created": "Fri, 24 Jul 2020 09:03:55 GMT", "version": "v3" }, { "created": "Tue, 11 Aug 2020 05:48:04 GMT", "version": "v4" }, { "created": "Tue, 17 Nov 2020 04:21:35 GMT", "version": "v5" } ]
2020-11-18
[ [ "Knights", "Joshua", "" ], [ "Harwood", "Ben", "" ], [ "Ward", "Daniel", "" ], [ "Vanderkop", "Anthony", "" ], [ "Mackenzie-Ross", "Olivia", "" ], [ "Moghadam", "Peyman", "" ] ]
This paper presents TCE: Temporally Coherent Embeddings for self-supervised video representation learning. The proposed method exploits inherent structure of unlabeled video data to explicitly enforce temporal coherency in the embedding space, rather than indirectly learning it through ranking or predictive proxy tasks. In the same way that high-level visual information in the world changes smoothly, we believe that nearby frames in learned representations will benefit from demonstrating similar properties. Using this assumption, we train our TCE model to encode videos such that adjacent frames exist close to each other and videos are separated from one another. Using TCE we learn robust representations from large quantities of unlabeled video data. We thoroughly analyse and evaluate our self-supervised learned TCE models on a downstream task of video action recognition using multiple challenging benchmarks (Kinetics400, UCF101, HMDB51). With a simple but effective 2D-CNN backbone and only RGB stream inputs, TCE pre-trained representations outperform all previous selfsupervised 2D-CNN and 3D-CNN pre-trained on UCF101. The code and pre-trained models for this paper can be downloaded at: https://github.com/csiro-robotics/TCE
2007.00959
Mingyuan Jiu
Mingyuan Jiu, Nelly Pustelnik
A deep primal-dual proximal network for image restoration
null
null
null
null
cs.CV
http://creativecommons.org/licenses/by-nc-nd/4.0/
Image restoration remains a challenging task in image processing. Numerous methods tackle this problem, often solved by minimizing a non-smooth penalized co-log-likelihood function. Although the solution is easily interpretable with theoretic guarantees, its estimation relies on an optimization process that can take time. Considering the research effort in deep learning for image classification and segmentation, this class of methods offers a serious alternative to perform image restoration but stays challenging to solve inverse problems. In this work, we design a deep network, named DeepPDNet, built from primal-dual proximal iterations associated with the minimization of a standard penalized likelihood with an analysis prior, allowing us to take advantage of both worlds. We reformulate a specific instance of the Condat-Vu primal-dual hybrid gradient (PDHG) algorithm as a deep network with fixed layers. The learned parameters are both the PDHG algorithm step-sizes and the analysis linear operator involved in the penalization (including the regularization parameter). These parameters are allowed to vary from a layer to another one. Two different learning strategies: "Full learning" and "Partial learning" are proposed, the first one is the most efficient numerically while the second one relies on standard constraints ensuring convergence in the standard PDHG iterations. Moreover, global and local sparse analysis prior are studied to seek a better feature representation. We apply the proposed methods to image restoration on the MNIST and BSD68 datasets and to single image super-resolution on the BSD100 and SET14 datasets. Extensive results show that the proposed DeepPDNet demonstrates excellent performance on the MNIST and the more complex BSD68, BSD100, and SET14 datasets for image restoration and single image super-resolution task.
[ { "created": "Thu, 2 Jul 2020 08:29:52 GMT", "version": "v1" }, { "created": "Tue, 9 Feb 2021 06:10:23 GMT", "version": "v2" }, { "created": "Mon, 20 Dec 2021 14:12:15 GMT", "version": "v3" } ]
2021-12-21
[ [ "Jiu", "Mingyuan", "" ], [ "Pustelnik", "Nelly", "" ] ]
Image restoration remains a challenging task in image processing. Numerous methods tackle this problem, often solved by minimizing a non-smooth penalized co-log-likelihood function. Although the solution is easily interpretable with theoretic guarantees, its estimation relies on an optimization process that can take time. Considering the research effort in deep learning for image classification and segmentation, this class of methods offers a serious alternative to perform image restoration but stays challenging to solve inverse problems. In this work, we design a deep network, named DeepPDNet, built from primal-dual proximal iterations associated with the minimization of a standard penalized likelihood with an analysis prior, allowing us to take advantage of both worlds. We reformulate a specific instance of the Condat-Vu primal-dual hybrid gradient (PDHG) algorithm as a deep network with fixed layers. The learned parameters are both the PDHG algorithm step-sizes and the analysis linear operator involved in the penalization (including the regularization parameter). These parameters are allowed to vary from a layer to another one. Two different learning strategies: "Full learning" and "Partial learning" are proposed, the first one is the most efficient numerically while the second one relies on standard constraints ensuring convergence in the standard PDHG iterations. Moreover, global and local sparse analysis prior are studied to seek a better feature representation. We apply the proposed methods to image restoration on the MNIST and BSD68 datasets and to single image super-resolution on the BSD100 and SET14 datasets. Extensive results show that the proposed DeepPDNet demonstrates excellent performance on the MNIST and the more complex BSD68, BSD100, and SET14 datasets for image restoration and single image super-resolution task.
1810.09729
Mohammed Ali Al-Garadi Dr
Reza Shakeri, Mohammed Ali Al-Garadi, Ahmed Badawy, Amr Mohamed, Tamer Khattab, Abdulla Al-Ali, Khaled A. Harras, Mohsen Guizani
Design Challenges of Multi-UAV Systems in Cyber-Physical Applications: A Comprehensive Survey, and Future Directions
null
null
null
null
cs.RO cs.AI cs.SY
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Unmanned Aerial Vehicles (UAVs) have recently rapidly grown to facilitate a wide range of innovative applications that can fundamentally change the way cyber-physical systems (CPSs) are designed. CPSs are a modern generation of systems with synergic cooperation between computational and physical potentials that can interact with humans through several new mechanisms. The main advantages of using UAVs in CPS application is their exceptional features, including their mobility, dynamism, effortless deployment, adaptive altitude, agility, adjustability, and effective appraisal of real-world functions anytime and anywhere. Furthermore, from the technology perspective, UAVs are predicted to be a vital element of the development of advanced CPSs. Therefore, in this survey, we aim to pinpoint the most fundamental and important design challenges of multi-UAV systems for CPS applications. We highlight key and versatile aspects that span the coverage and tracking of targets and infrastructure objects, energy-efficient navigation, and image analysis using machine learning for fine-grained CPS applications. Key prototypes and testbeds are also investigated to show how these practical technologies can facilitate CPS applications. We present and propose state-of-the-art algorithms to address design challenges with both quantitative and qualitative methods and map these challenges with important CPS applications to draw insightful conclusions on the challenges of each application. Finally, we summarize potential new directions and ideas that could shape future research in these areas.
[ { "created": "Tue, 23 Oct 2018 08:51:54 GMT", "version": "v1" } ]
2018-10-24
[ [ "Shakeri", "Reza", "" ], [ "Al-Garadi", "Mohammed Ali", "" ], [ "Badawy", "Ahmed", "" ], [ "Mohamed", "Amr", "" ], [ "Khattab", "Tamer", "" ], [ "Al-Ali", "Abdulla", "" ], [ "Harras", "Khaled A.", "" ], [ "Guizani", "Mohsen", "" ] ]
Unmanned Aerial Vehicles (UAVs) have recently rapidly grown to facilitate a wide range of innovative applications that can fundamentally change the way cyber-physical systems (CPSs) are designed. CPSs are a modern generation of systems with synergic cooperation between computational and physical potentials that can interact with humans through several new mechanisms. The main advantages of using UAVs in CPS application is their exceptional features, including their mobility, dynamism, effortless deployment, adaptive altitude, agility, adjustability, and effective appraisal of real-world functions anytime and anywhere. Furthermore, from the technology perspective, UAVs are predicted to be a vital element of the development of advanced CPSs. Therefore, in this survey, we aim to pinpoint the most fundamental and important design challenges of multi-UAV systems for CPS applications. We highlight key and versatile aspects that span the coverage and tracking of targets and infrastructure objects, energy-efficient navigation, and image analysis using machine learning for fine-grained CPS applications. Key prototypes and testbeds are also investigated to show how these practical technologies can facilitate CPS applications. We present and propose state-of-the-art algorithms to address design challenges with both quantitative and qualitative methods and map these challenges with important CPS applications to draw insightful conclusions on the challenges of each application. Finally, we summarize potential new directions and ideas that could shape future research in these areas.
2205.01324
Alon Berliner
Alon Berliner, Guy Rotman, Yossi Adi, Roi Reichart, Tamir Hazan
Learning Discrete Structured Variational Auto-Encoder using Natural Evolution Strategies
Published as a conference paper at ICLR 2022
null
null
null
cs.LG cs.NE stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Discrete variational auto-encoders (VAEs) are able to represent semantic latent spaces in generative learning. In many real-life settings, the discrete latent space consists of high-dimensional structures, and propagating gradients through the relevant structures often requires enumerating over an exponentially large latent space. Recently, various approaches were devised to propagate approximated gradients without enumerating over the space of possible structures. In this work, we use Natural Evolution Strategies (NES), a class of gradient-free black-box optimization algorithms, to learn discrete structured VAEs. The NES algorithms are computationally appealing as they estimate gradients with forward pass evaluations only, thus they do not require to propagate gradients through their discrete structures. We demonstrate empirically that optimizing discrete structured VAEs using NES is as effective as gradient-based approximations. Lastly, we prove NES converges for non-Lipschitz functions as appear in discrete structured VAEs.
[ { "created": "Tue, 3 May 2022 06:21:40 GMT", "version": "v1" } ]
2022-05-04
[ [ "Berliner", "Alon", "" ], [ "Rotman", "Guy", "" ], [ "Adi", "Yossi", "" ], [ "Reichart", "Roi", "" ], [ "Hazan", "Tamir", "" ] ]
Discrete variational auto-encoders (VAEs) are able to represent semantic latent spaces in generative learning. In many real-life settings, the discrete latent space consists of high-dimensional structures, and propagating gradients through the relevant structures often requires enumerating over an exponentially large latent space. Recently, various approaches were devised to propagate approximated gradients without enumerating over the space of possible structures. In this work, we use Natural Evolution Strategies (NES), a class of gradient-free black-box optimization algorithms, to learn discrete structured VAEs. The NES algorithms are computationally appealing as they estimate gradients with forward pass evaluations only, thus they do not require to propagate gradients through their discrete structures. We demonstrate empirically that optimizing discrete structured VAEs using NES is as effective as gradient-based approximations. Lastly, we prove NES converges for non-Lipschitz functions as appear in discrete structured VAEs.
2006.00577
Alessandro Ecclesie Agazzi
Alessandro Ecclesie Agazzi
Phishing and Spear Phishing: examples in Cyber Espionage and techniques to protect against them
null
null
null
null
cs.CR cs.CY
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Phishing attacks have become the most used technique in the online scams, initiating more than 91% of cyberattacks, from 2012 onwards. This study reviews how Phishing and Spear Phishing attacks are carried out by the phishers, through 5 steps which magnify the outcome, increasing the chance of success. The focus will be also given on four different layers of protection against these social engineering attacks, showing their strengths and weaknesses; the first and second layers consist of automated tools and decision-aid tools. the third one is users' knowledge and expertise to deal with potential threats. The last layer, defined as "external", will underline the importance of having a Multi-factor authentication, an effective way to provide an enhanced security, creating a further layer of protection against Phishing and Spear Phishing.
[ { "created": "Sun, 31 May 2020 18:10:09 GMT", "version": "v1" } ]
2020-06-02
[ [ "Agazzi", "Alessandro Ecclesie", "" ] ]
Phishing attacks have become the most used technique in the online scams, initiating more than 91% of cyberattacks, from 2012 onwards. This study reviews how Phishing and Spear Phishing attacks are carried out by the phishers, through 5 steps which magnify the outcome, increasing the chance of success. The focus will be also given on four different layers of protection against these social engineering attacks, showing their strengths and weaknesses; the first and second layers consist of automated tools and decision-aid tools. the third one is users' knowledge and expertise to deal with potential threats. The last layer, defined as "external", will underline the importance of having a Multi-factor authentication, an effective way to provide an enhanced security, creating a further layer of protection against Phishing and Spear Phishing.
1003.3689
Murat Manguoglu
Murat Manguoglu
A Highly Efficient Parallel Algorithm for Computing the Fiedler Vector
This paper has been withdrawn by the author because it is under revision
null
null
null
cs.NA cs.MS
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This paper has been withdrawn by the author.
[ { "created": "Thu, 18 Mar 2010 22:56:57 GMT", "version": "v1" }, { "created": "Tue, 12 Feb 2013 19:44:27 GMT", "version": "v2" } ]
2015-03-13
[ [ "Manguoglu", "Murat", "" ] ]
This paper has been withdrawn by the author.
1404.2943
Thomas Bl\"asius
Thomas Bl\"asius, Sebastian Lehmann, Ignaz Rutter
Orthogonal Graph Drawing with Inflexible Edges
23 pages, 5 figures
null
null
null
cs.DS cs.DM
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We consider the problem of creating plane orthogonal drawings of 4-planar graphs (planar graphs with maximum degree 4) with constraints on the number of bends per edge. More precisely, we have a flexibility function assigning to each edge $e$ a natural number $\mathrm{flex}(e)$, its flexibility. The problem FlexDraw asks whether there exists an orthogonal drawing such that each edge $e$ has at most $\mathrm{flex}(e)$ bends. It is known that FlexDraw is NP-hard if $\mathrm{flex}(e) = 0$ for every edge $e$. On the other hand, FlexDraw can be solved efficiently if $\mathrm{flex}(e) \ge 1$ and is trivial if $\mathrm{flex}(e) \ge 2$ for every edge $e$. To close the gap between the NP-hardness for $\mathrm{flex}(e) = 0$ and the efficient algorithm for $\mathrm{flex}(e) \ge 1$, we investigate the computational complexity of FlexDraw in case only few edges are inflexible (i.e., have flexibility~$0$). We show that for any $\varepsilon > 0$ FlexDraw is NP-complete for instances with $O(n^\varepsilon)$ inflexible edges with pairwise distance $\Omega(n^{1-\varepsilon})$ (including the case where they induce a matching). On the other hand, we give an FPT-algorithm with running time $O(2^k\cdot n \cdot T_{\mathrm{flow}}(n))$, where $T_{\mathrm{flow}}(n)$ is the time necessary to compute a maximum flow in a planar flow network with multiple sources and sinks, and $k$ is the number of inflexible edges having at least one endpoint of degree 4.
[ { "created": "Thu, 10 Apr 2014 20:24:06 GMT", "version": "v1" }, { "created": "Wed, 7 Jan 2015 16:03:13 GMT", "version": "v2" } ]
2015-01-08
[ [ "Bläsius", "Thomas", "" ], [ "Lehmann", "Sebastian", "" ], [ "Rutter", "Ignaz", "" ] ]
We consider the problem of creating plane orthogonal drawings of 4-planar graphs (planar graphs with maximum degree 4) with constraints on the number of bends per edge. More precisely, we have a flexibility function assigning to each edge $e$ a natural number $\mathrm{flex}(e)$, its flexibility. The problem FlexDraw asks whether there exists an orthogonal drawing such that each edge $e$ has at most $\mathrm{flex}(e)$ bends. It is known that FlexDraw is NP-hard if $\mathrm{flex}(e) = 0$ for every edge $e$. On the other hand, FlexDraw can be solved efficiently if $\mathrm{flex}(e) \ge 1$ and is trivial if $\mathrm{flex}(e) \ge 2$ for every edge $e$. To close the gap between the NP-hardness for $\mathrm{flex}(e) = 0$ and the efficient algorithm for $\mathrm{flex}(e) \ge 1$, we investigate the computational complexity of FlexDraw in case only few edges are inflexible (i.e., have flexibility~$0$). We show that for any $\varepsilon > 0$ FlexDraw is NP-complete for instances with $O(n^\varepsilon)$ inflexible edges with pairwise distance $\Omega(n^{1-\varepsilon})$ (including the case where they induce a matching). On the other hand, we give an FPT-algorithm with running time $O(2^k\cdot n \cdot T_{\mathrm{flow}}(n))$, where $T_{\mathrm{flow}}(n)$ is the time necessary to compute a maximum flow in a planar flow network with multiple sources and sinks, and $k$ is the number of inflexible edges having at least one endpoint of degree 4.
1901.10645
Sara Rouhani
Sara Rouhani, Luke Butterworth, Adam D. Simmons, Darryl G. Humphery, and Ralph Deters
MediChainTM: A Secure Decentralized Medical Data Asset Management System
2018 IEEE Confs on Internet of Things, Green Computing and Communications, Cyber, Physical and Social Computing, Smart Data, Blockchain, Computer and Information Technology, Congress on Cybermatics
null
10.1109/Cybermatics_2018.2018.00258
null
cs.DC
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The set of distributed ledger architectures known as blockchain is best known for cryptocurrency applications such as Bitcoin and Ethereum. These permissionless block chains are showing the potential to be disruptive to the financial services industry. Their broader adoption is likely to be limited by the maximum block size, the cost of the Proof of Work consensus mechanism, and the increasing size of any given chain overwhelming most of the participating nodes. These factors have led to many cryptocurrency blockchains to become centralized in the nodes with enough computing power and storage to be a dominant miner and validator. Permissioned chains operate in trusted environments and can, therefore, avoid the computationally expensive consensus mechanisms. Permissioned chains are still susceptible to asset storage demands and non-standard user interfaces that will impede their adoption. This paper describes an approach to addressing these limitations: permissioned blockchain that uses off-chain storage of the data assets and this is accessed through a standard browser and mobile app. The implementation in the Hyperledger framework is described as is an example use of patient-centered health data management.
[ { "created": "Wed, 30 Jan 2019 02:22:07 GMT", "version": "v1" } ]
2019-01-31
[ [ "Rouhani", "Sara", "" ], [ "Butterworth", "Luke", "" ], [ "Simmons", "Adam D.", "" ], [ "Humphery", "Darryl G.", "" ], [ "Deters", "Ralph", "" ] ]
The set of distributed ledger architectures known as blockchain is best known for cryptocurrency applications such as Bitcoin and Ethereum. These permissionless block chains are showing the potential to be disruptive to the financial services industry. Their broader adoption is likely to be limited by the maximum block size, the cost of the Proof of Work consensus mechanism, and the increasing size of any given chain overwhelming most of the participating nodes. These factors have led to many cryptocurrency blockchains to become centralized in the nodes with enough computing power and storage to be a dominant miner and validator. Permissioned chains operate in trusted environments and can, therefore, avoid the computationally expensive consensus mechanisms. Permissioned chains are still susceptible to asset storage demands and non-standard user interfaces that will impede their adoption. This paper describes an approach to addressing these limitations: permissioned blockchain that uses off-chain storage of the data assets and this is accessed through a standard browser and mobile app. The implementation in the Hyperledger framework is described as is an example use of patient-centered health data management.
0811.1335
Mugurel Ionut Andreica
Mugurel Ionut Andreica
Algorithmic Techniques for Several Optimization Problems Regarding Distributed Systems with Tree Topologies
The 16th International Conference on Applied and Industrial Mathematics, Oradea, Romania, 9-11 October, 2008. ROMAI Journal, vol. 4, 2008. (ISSN: 841-5512). In Press
ROMAI Journal, vol. 4, no. 1, pp. 1-25, 2008 (ISSN: 1841-5512) ; http://www.romai.ro
null
null
cs.DS cs.DM cs.NI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
As the development of distributed systems progresses, more and more challenges arise and the need for developing optimized systems and for optimizing existing systems from multiple perspectives becomes more stringent. In this paper I present novel algorithmic techniques for solving several optimization problems regarding distributed systems with tree topologies. I address topics like: reliability improvement, partitioning, coloring, content delivery, optimal matchings, as well as some tree counting aspects. Some of the presented techniques are only of theoretical interest, while others can be used in practical settings.
[ { "created": "Sun, 9 Nov 2008 12:59:45 GMT", "version": "v1" } ]
2009-03-21
[ [ "Andreica", "Mugurel Ionut", "" ] ]
As the development of distributed systems progresses, more and more challenges arise and the need for developing optimized systems and for optimizing existing systems from multiple perspectives becomes more stringent. In this paper I present novel algorithmic techniques for solving several optimization problems regarding distributed systems with tree topologies. I address topics like: reliability improvement, partitioning, coloring, content delivery, optimal matchings, as well as some tree counting aspects. Some of the presented techniques are only of theoretical interest, while others can be used in practical settings.
2405.07621
Satheesh Kumar Perepu Dr
Kaushik Dey, Satheesh K. Perepu, Abir Das, Pallab Dasgupta
Towards Adaptive IMFs -- Generalization of utility functions in Multi-Agent Frameworks
Accepted in Netsoft-2024 conference
null
null
null
cs.LG cs.AI
http://creativecommons.org/licenses/by/4.0/
Intent Management Function (IMF) is an integral part of future-generation networks. In recent years, there has been some work on AI-based IMFs that can handle conflicting intents and prioritize the global objective based on apriori definition of the utility function and accorded priorities for competing intents. Some of the earlier works use Multi-Agent Reinforcement Learning (MARL) techniques with AdHoc Teaming (AHT) approaches for efficient conflict handling in IMF. However, the success of such frameworks in real-life scenarios requires them to be flexible to business situations. The intent priorities can change and the utility function, which measures the extent of intent fulfilment, may also vary in definition. This paper proposes a novel mechanism whereby the IMF can generalize to different forms of utility functions and change of intent priorities at run-time without additional training. Such generalization ability, without additional training requirements, would help to deploy IMF in live networks where customer intents and priorities change frequently. Results on the network emulator demonstrate the efficacy of the approach, scalability for new intents, outperforming existing techniques that require additional training to achieve the same degree of flexibility thereby saving cost, and increasing efficiency and adaptability.
[ { "created": "Mon, 13 May 2024 10:27:11 GMT", "version": "v1" }, { "created": "Tue, 14 May 2024 06:29:36 GMT", "version": "v2" } ]
2024-05-15
[ [ "Dey", "Kaushik", "" ], [ "Perepu", "Satheesh K.", "" ], [ "Das", "Abir", "" ], [ "Dasgupta", "Pallab", "" ] ]
Intent Management Function (IMF) is an integral part of future-generation networks. In recent years, there has been some work on AI-based IMFs that can handle conflicting intents and prioritize the global objective based on apriori definition of the utility function and accorded priorities for competing intents. Some of the earlier works use Multi-Agent Reinforcement Learning (MARL) techniques with AdHoc Teaming (AHT) approaches for efficient conflict handling in IMF. However, the success of such frameworks in real-life scenarios requires them to be flexible to business situations. The intent priorities can change and the utility function, which measures the extent of intent fulfilment, may also vary in definition. This paper proposes a novel mechanism whereby the IMF can generalize to different forms of utility functions and change of intent priorities at run-time without additional training. Such generalization ability, without additional training requirements, would help to deploy IMF in live networks where customer intents and priorities change frequently. Results on the network emulator demonstrate the efficacy of the approach, scalability for new intents, outperforming existing techniques that require additional training to achieve the same degree of flexibility thereby saving cost, and increasing efficiency and adaptability.
1807.02800
Pascal Mettes
Pascal Mettes and Cees G. M. Snoek
Spatio-Temporal Instance Learning: Action Tubes from Class Supervision
null
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The goal of this work is spatio-temporal action localization in videos, using only the supervision from video-level class labels. The state-of-the-art casts this weakly-supervised action localization regime as a Multiple Instance Learning problem, where instances are a priori computed spatio-temporal proposals. Rather than disconnecting the spatio-temporal learning from the training, we propose Spatio-Temporal Instance Learning, which enables action localization directly from box proposals in video frames. We outline the assumptions of our model and propose a max-margin objective and optimization with latent variables that enable spatio-temporal learning of actions from video labels. We also provide an efficient linking algorithm and two reranking strategies to facilitate and further improve the action localization. Experimental evaluation on four action datasets demonstrate the effectiveness of our approach for localization from weak supervision. Moreover, we show how to incorporate other supervision levels and mixtures, as a step towards determining optimal supervision strategies for action localization.
[ { "created": "Sun, 8 Jul 2018 11:12:51 GMT", "version": "v1" }, { "created": "Wed, 21 Nov 2018 21:13:28 GMT", "version": "v2" } ]
2018-11-26
[ [ "Mettes", "Pascal", "" ], [ "Snoek", "Cees G. M.", "" ] ]
The goal of this work is spatio-temporal action localization in videos, using only the supervision from video-level class labels. The state-of-the-art casts this weakly-supervised action localization regime as a Multiple Instance Learning problem, where instances are a priori computed spatio-temporal proposals. Rather than disconnecting the spatio-temporal learning from the training, we propose Spatio-Temporal Instance Learning, which enables action localization directly from box proposals in video frames. We outline the assumptions of our model and propose a max-margin objective and optimization with latent variables that enable spatio-temporal learning of actions from video labels. We also provide an efficient linking algorithm and two reranking strategies to facilitate and further improve the action localization. Experimental evaluation on four action datasets demonstrate the effectiveness of our approach for localization from weak supervision. Moreover, we show how to incorporate other supervision levels and mixtures, as a step towards determining optimal supervision strategies for action localization.
2402.01126
Douglas Poland
Douglas Poland and Amar Saini
Seeing Objects in a Cluttered World: Computational Objectness from Motion in Video
10 pages, 11 figures, plus 18 pages of Supplemental Information
null
null
LLNL-JRNL-859920
cs.CV
http://creativecommons.org/licenses/by-sa/4.0/
Perception of the visually disjoint surfaces of our cluttered world as whole objects, physically distinct from those overlapping them, is a cognitive phenomenon called objectness that forms the basis of our visual perception. Shared by all vertebrates and present at birth in humans, it enables object-centric representation and reasoning about the visual world. We present a computational approach to objectness that leverages motion cues and spatio-temporal attention using a pair of supervised spatio-temporal R(2+1)U-Nets. The first network detects motion boundaries and classifies the pixels at those boundaries in terms of their local foreground-background sense. This motion boundary sense (MBS) information is passed, along with a spatio-temporal object attention cue, to an attentional surface perception (ASP) module which infers the form of the attended object over a sequence of frames and classifies its 'pixels' as visible or obscured. The spatial form of the attention cue is flexible, but it must loosely track the attended object which need not be visible. We demonstrate the ability of this simple but novel approach to infer objectness from phenomenology without object models, and show that it delivers robust perception of individual attended objects in cluttered scenes, even with blur and camera shake. We show that our data diversity and augmentation minimizes bias and facilitates transfer to real video. Finally, we describe how this computational objectness capability can grow in sophistication and anchor a robust modular video object perception framework.
[ { "created": "Fri, 2 Feb 2024 03:57:11 GMT", "version": "v1" } ]
2024-02-05
[ [ "Poland", "Douglas", "" ], [ "Saini", "Amar", "" ] ]
Perception of the visually disjoint surfaces of our cluttered world as whole objects, physically distinct from those overlapping them, is a cognitive phenomenon called objectness that forms the basis of our visual perception. Shared by all vertebrates and present at birth in humans, it enables object-centric representation and reasoning about the visual world. We present a computational approach to objectness that leverages motion cues and spatio-temporal attention using a pair of supervised spatio-temporal R(2+1)U-Nets. The first network detects motion boundaries and classifies the pixels at those boundaries in terms of their local foreground-background sense. This motion boundary sense (MBS) information is passed, along with a spatio-temporal object attention cue, to an attentional surface perception (ASP) module which infers the form of the attended object over a sequence of frames and classifies its 'pixels' as visible or obscured. The spatial form of the attention cue is flexible, but it must loosely track the attended object which need not be visible. We demonstrate the ability of this simple but novel approach to infer objectness from phenomenology without object models, and show that it delivers robust perception of individual attended objects in cluttered scenes, even with blur and camera shake. We show that our data diversity and augmentation minimizes bias and facilitates transfer to real video. Finally, we describe how this computational objectness capability can grow in sophistication and anchor a robust modular video object perception framework.
2110.12661
Jiawei Zhao
Jiawei Zhao, Florian Sch\"afer, Anima Anandkumar
ZerO Initialization: Initializing Neural Networks with only Zeros and Ones
null
null
null
null
cs.LG cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Deep neural networks are usually initialized with random weights, with adequately selected initial variance to ensure stable signal propagation during training. However, selecting the appropriate variance becomes challenging especially as the number of layers grows. In this work, we replace random weight initialization with a fully deterministic initialization scheme, viz., ZerO, which initializes the weights of networks with only zeros and ones (up to a normalization factor), based on identity and Hadamard transforms. Through both theoretical and empirical studies, we demonstrate that ZerO is able to train networks without damaging their expressivity. Applying ZerO on ResNet achieves state-of-the-art performance on various datasets, including ImageNet, which suggests random weights may be unnecessary for network initialization. In addition, ZerO has many benefits, such as training ultra deep networks (without batch-normalization), exhibiting low-rank learning trajectories that result in low-rank and sparse solutions, and improving training reproducibility.
[ { "created": "Mon, 25 Oct 2021 06:17:33 GMT", "version": "v1" }, { "created": "Tue, 23 Aug 2022 03:00:36 GMT", "version": "v2" }, { "created": "Fri, 4 Nov 2022 17:17:26 GMT", "version": "v3" } ]
2022-11-07
[ [ "Zhao", "Jiawei", "" ], [ "Schäfer", "Florian", "" ], [ "Anandkumar", "Anima", "" ] ]
Deep neural networks are usually initialized with random weights, with adequately selected initial variance to ensure stable signal propagation during training. However, selecting the appropriate variance becomes challenging especially as the number of layers grows. In this work, we replace random weight initialization with a fully deterministic initialization scheme, viz., ZerO, which initializes the weights of networks with only zeros and ones (up to a normalization factor), based on identity and Hadamard transforms. Through both theoretical and empirical studies, we demonstrate that ZerO is able to train networks without damaging their expressivity. Applying ZerO on ResNet achieves state-of-the-art performance on various datasets, including ImageNet, which suggests random weights may be unnecessary for network initialization. In addition, ZerO has many benefits, such as training ultra deep networks (without batch-normalization), exhibiting low-rank learning trajectories that result in low-rank and sparse solutions, and improving training reproducibility.
2312.10624
Jie JW Wu PhD
Jie JW Wu
AutoOffAB: Toward Automated Offline A/B Testing for Data-Driven Requirement Engineering
5 pages, 2 figures. Accepted at FSE 2024 (32nd ACM International Conference on the Foundations of Software Engineering)
32nd ACM International Conference on the Foundations of Software Engineering (FSE 2024)
10.1145/3663529.3663780
null
cs.SE
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Software companies have widely used online A/B testing to evaluate the impact of a new technology by offering it to groups of users and comparing it against the unmodified product. However, running online A/B testing needs not only efforts in design, implementation, and stakeholders' approval to be served in production but also several weeks to collect the data in iterations. To address these issues, a recently emerging topic, called "Offline A/B Testing", is getting increasing attention, intending to conduct the offline evaluation of new technologies by estimating historical logged data. Although this approach is promising due to lower implementation effort, faster turnaround time, and no potential user harm, for it to be effectively prioritized as requirements in practice, several limitations need to be addressed, including its discrepancy with online A/B test results, and lack of systematic updates on varying data and parameters. In response, in this vision paper, I introduce AutoOffAB, an idea to automatically run variants of offline A/B testing against recent logging and update the offline evaluation results, which are used to make decisions on requirements more reliably and systematically.
[ { "created": "Sun, 17 Dec 2023 06:49:14 GMT", "version": "v1" }, { "created": "Fri, 9 Aug 2024 08:17:37 GMT", "version": "v2" } ]
2024-08-12
[ [ "Wu", "Jie JW", "" ] ]
Software companies have widely used online A/B testing to evaluate the impact of a new technology by offering it to groups of users and comparing it against the unmodified product. However, running online A/B testing needs not only efforts in design, implementation, and stakeholders' approval to be served in production but also several weeks to collect the data in iterations. To address these issues, a recently emerging topic, called "Offline A/B Testing", is getting increasing attention, intending to conduct the offline evaluation of new technologies by estimating historical logged data. Although this approach is promising due to lower implementation effort, faster turnaround time, and no potential user harm, for it to be effectively prioritized as requirements in practice, several limitations need to be addressed, including its discrepancy with online A/B test results, and lack of systematic updates on varying data and parameters. In response, in this vision paper, I introduce AutoOffAB, an idea to automatically run variants of offline A/B testing against recent logging and update the offline evaluation results, which are used to make decisions on requirements more reliably and systematically.
1609.03773
Li Cheng
Chi Xu, Lakshmi Narasimhan Govindarajan, Yu Zhang, Li Cheng
Lie-X: Depth Image Based Articulated Object Pose Estimation, Tracking, and Action Recognition on Lie Groups
null
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Pose estimation, tracking, and action recognition of articulated objects from depth images are important and challenging problems, which are normally considered separately. In this paper, a unified paradigm based on Lie group theory is proposed, which enables us to collectively address these related problems. Our approach is also applicable to a wide range of articulated objects. Empirically it is evaluated on lab animals including mouse and fish, as well as on human hand. On these applications, it is shown to deliver competitive results compared to the state-of-the-arts, and non-trivial baselines including convolutional neural networks and regression forest methods.
[ { "created": "Tue, 13 Sep 2016 11:36:26 GMT", "version": "v1" } ]
2016-09-14
[ [ "Xu", "Chi", "" ], [ "Govindarajan", "Lakshmi Narasimhan", "" ], [ "Zhang", "Yu", "" ], [ "Cheng", "Li", "" ] ]
Pose estimation, tracking, and action recognition of articulated objects from depth images are important and challenging problems, which are normally considered separately. In this paper, a unified paradigm based on Lie group theory is proposed, which enables us to collectively address these related problems. Our approach is also applicable to a wide range of articulated objects. Empirically it is evaluated on lab animals including mouse and fish, as well as on human hand. On these applications, it is shown to deliver competitive results compared to the state-of-the-arts, and non-trivial baselines including convolutional neural networks and regression forest methods.
cs/0012014
Gyongyi Szilagyi
Gyongyi Szilagyi, Tibor Gyimothy and Jan Maluszynski
Slicing of Constraint Logic Programs
In M. Ducasse (ed), proceedings of the Fourth International Workshop on Automated Debugging (AADEBUG 2000), August 2000, Munich. cs.SE/0010035
null
null
null
cs.SE
null
Slicing is a program analysis technique originally developed for imperative languages. It facilitates understanding of data flow and debugging. This paper discusses slicing of Constraint Logic Programs. Constraint Logic Programming (CLP) is an emerging software technology with a growing number of applications. Data flow in constraint programs is not explicit, and for this reason the concepts of slice and the slicing techniques of imperative languages are not directly applicable. This paper formulates declarative notions of slice suitable for CLP. They provide a basis for defining slicing techniques (both dynamic and static) based on variable sharing. The techniques are further extended by using groundness information. A prototype dynamic slicer of CLP programs implementing the presented ideas is briefly described together with the results of some slicing experiments.
[ { "created": "Mon, 18 Dec 2000 11:59:31 GMT", "version": "v1" } ]
2007-05-23
[ [ "Szilagyi", "Gyongyi", "" ], [ "Gyimothy", "Tibor", "" ], [ "Maluszynski", "Jan", "" ] ]
Slicing is a program analysis technique originally developed for imperative languages. It facilitates understanding of data flow and debugging. This paper discusses slicing of Constraint Logic Programs. Constraint Logic Programming (CLP) is an emerging software technology with a growing number of applications. Data flow in constraint programs is not explicit, and for this reason the concepts of slice and the slicing techniques of imperative languages are not directly applicable. This paper formulates declarative notions of slice suitable for CLP. They provide a basis for defining slicing techniques (both dynamic and static) based on variable sharing. The techniques are further extended by using groundness information. A prototype dynamic slicer of CLP programs implementing the presented ideas is briefly described together with the results of some slicing experiments.
2403.05493
Agnes Luhtaru
Agnes Luhtaru, Taido Purason, Martin Vainikko, Maksym Del, Mark Fishel
To Err Is Human, but Llamas Can Learn It Too
null
null
null
null
cs.CL
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This study explores enhancing grammatical error correction (GEC) through artificial error generation (AEG) using language models (LMs). Specifically, we fine-tune Llama 2-based LMs for error generation and find that this approach yields synthetic errors akin to human errors. Next, we train GEC Llama models with the help of these artificial errors and outperform previous state-of-the-art error correction models, with gains ranging between 0.8 and 6 F0.5 points across all tested languages (German, Ukrainian, and Estonian). Moreover, we demonstrate that generating errors by fine-tuning smaller sequence-to-sequence models and prompting large commercial LMs (GPT-3.5 and GPT-4) also results in synthetic errors beneficially affecting error generation models.
[ { "created": "Fri, 8 Mar 2024 18:04:03 GMT", "version": "v1" } ]
2024-03-11
[ [ "Luhtaru", "Agnes", "" ], [ "Purason", "Taido", "" ], [ "Vainikko", "Martin", "" ], [ "Del", "Maksym", "" ], [ "Fishel", "Mark", "" ] ]
This study explores enhancing grammatical error correction (GEC) through artificial error generation (AEG) using language models (LMs). Specifically, we fine-tune Llama 2-based LMs for error generation and find that this approach yields synthetic errors akin to human errors. Next, we train GEC Llama models with the help of these artificial errors and outperform previous state-of-the-art error correction models, with gains ranging between 0.8 and 6 F0.5 points across all tested languages (German, Ukrainian, and Estonian). Moreover, we demonstrate that generating errors by fine-tuning smaller sequence-to-sequence models and prompting large commercial LMs (GPT-3.5 and GPT-4) also results in synthetic errors beneficially affecting error generation models.
1606.07056
Abhay Prakash
Abhay Prakash, Chris Brockett, Puneet Agrawal
Emulating Human Conversations using Convolutional Neural Network-based IR
5 pages, Neu-IR'16 SIGIR Workshop on Neural Information Retrieval, July 21, 2016, Pisa, Italy
null
null
null
cs.AI cs.CL cs.IR
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Conversational agents ("bots") are beginning to be widely used in conversational interfaces. To design a system that is capable of emulating human-like interactions, a conversational layer that can serve as a fabric for chat-like interaction with the agent is needed. In this paper, we introduce a model that employs Information Retrieval by utilizing convolutional deep structured semantic neural network-based features in the ranker to present human-like responses in ongoing conversation with a user. In conversations, accounting for context is critical to the retrieval model; we show that our context-sensitive approach using a Convolutional Deep Structured Semantic Model (cDSSM) with character trigrams significantly outperforms several conventional baselines in terms of the relevance of responses retrieved.
[ { "created": "Wed, 22 Jun 2016 19:55:24 GMT", "version": "v1" } ]
2016-06-23
[ [ "Prakash", "Abhay", "" ], [ "Brockett", "Chris", "" ], [ "Agrawal", "Puneet", "" ] ]
Conversational agents ("bots") are beginning to be widely used in conversational interfaces. To design a system that is capable of emulating human-like interactions, a conversational layer that can serve as a fabric for chat-like interaction with the agent is needed. In this paper, we introduce a model that employs Information Retrieval by utilizing convolutional deep structured semantic neural network-based features in the ranker to present human-like responses in ongoing conversation with a user. In conversations, accounting for context is critical to the retrieval model; we show that our context-sensitive approach using a Convolutional Deep Structured Semantic Model (cDSSM) with character trigrams significantly outperforms several conventional baselines in terms of the relevance of responses retrieved.
2406.17659
Xiaohan Zhang
Xiaohan Zhang, Zainab Altaweel, Yohei Hayamizu, Yan Ding, Saeid Amiri, Hao Yang, Andy Kaminski, Chad Esselink, Shiqi Zhang
DKPROMPT: Domain Knowledge Prompting Vision-Language Models for Open-World Planning
null
null
null
null
cs.AI cs.RO
http://creativecommons.org/licenses/by/4.0/
Vision-language models (VLMs) have been applied to robot task planning problems, where the robot receives a task in natural language and generates plans based on visual inputs. While current VLMs have demonstrated strong vision-language understanding capabilities, their performance is still far from being satisfactory in planning tasks. At the same time, although classical task planners, such as PDDL-based, are strong in planning for long-horizon tasks, they do not work well in open worlds where unforeseen situations are common. In this paper, we propose a novel task planning and execution framework, called DKPROMPT, which automates VLM prompting using domain knowledge in PDDL for classical planning in open worlds. Results from quantitative experiments show that DKPROMPT outperforms classical planning, pure VLM-based and a few other competitive baselines in task completion rate.
[ { "created": "Tue, 25 Jun 2024 15:49:47 GMT", "version": "v1" } ]
2024-06-26
[ [ "Zhang", "Xiaohan", "" ], [ "Altaweel", "Zainab", "" ], [ "Hayamizu", "Yohei", "" ], [ "Ding", "Yan", "" ], [ "Amiri", "Saeid", "" ], [ "Yang", "Hao", "" ], [ "Kaminski", "Andy", "" ], [ "Esselink", "Chad", "" ], [ "Zhang", "Shiqi", "" ] ]
Vision-language models (VLMs) have been applied to robot task planning problems, where the robot receives a task in natural language and generates plans based on visual inputs. While current VLMs have demonstrated strong vision-language understanding capabilities, their performance is still far from being satisfactory in planning tasks. At the same time, although classical task planners, such as PDDL-based, are strong in planning for long-horizon tasks, they do not work well in open worlds where unforeseen situations are common. In this paper, we propose a novel task planning and execution framework, called DKPROMPT, which automates VLM prompting using domain knowledge in PDDL for classical planning in open worlds. Results from quantitative experiments show that DKPROMPT outperforms classical planning, pure VLM-based and a few other competitive baselines in task completion rate.
1112.2774
Tina Eliassi-Rad
Mangesh Gupte and Tina Eliassi-Rad
Measuring Tie Strength in Implicit Social Networks
10 pages
null
null
null
cs.SI physics.soc-ph
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Given a set of people and a set of events they attend, we address the problem of measuring connectedness or tie strength between each pair of persons given that attendance at mutual events gives an implicit social network between people. We take an axiomatic approach to this problem. Starting from a list of axioms that a measure of tie strength must satisfy, we characterize functions that satisfy all the axioms and show that there is a range of measures that satisfy this characterization. A measure of tie strength induces a ranking on the edges (and on the set of neighbors for every person). We show that for applications where the ranking, and not the absolute value of the tie strength, is the important thing about the measure, the axioms are equivalent to a natural partial order. Also, to settle on a particular measure, we must make a non-obvious decision about extending this partial order to a total order, and that this decision is best left to particular applications. We classify measures found in prior literature according to the axioms that they satisfy. In our experiments, we measure tie strength and the coverage of our axioms in several datasets. Also, for each dataset, we bound the maximum Kendall's Tau divergence (which measures the number of pairwise disagreements between two lists) between all measures that satisfy the axioms using the partial order. This informs us if particular datasets are well behaved where we do not have to worry about which measure to choose, or we have to be careful about the exact choice of measure we make.
[ { "created": "Tue, 13 Dec 2011 02:30:22 GMT", "version": "v1" } ]
2011-12-14
[ [ "Gupte", "Mangesh", "" ], [ "Eliassi-Rad", "Tina", "" ] ]
Given a set of people and a set of events they attend, we address the problem of measuring connectedness or tie strength between each pair of persons given that attendance at mutual events gives an implicit social network between people. We take an axiomatic approach to this problem. Starting from a list of axioms that a measure of tie strength must satisfy, we characterize functions that satisfy all the axioms and show that there is a range of measures that satisfy this characterization. A measure of tie strength induces a ranking on the edges (and on the set of neighbors for every person). We show that for applications where the ranking, and not the absolute value of the tie strength, is the important thing about the measure, the axioms are equivalent to a natural partial order. Also, to settle on a particular measure, we must make a non-obvious decision about extending this partial order to a total order, and that this decision is best left to particular applications. We classify measures found in prior literature according to the axioms that they satisfy. In our experiments, we measure tie strength and the coverage of our axioms in several datasets. Also, for each dataset, we bound the maximum Kendall's Tau divergence (which measures the number of pairwise disagreements between two lists) between all measures that satisfy the axioms using the partial order. This informs us if particular datasets are well behaved where we do not have to worry about which measure to choose, or we have to be careful about the exact choice of measure we make.
2107.12048
Wei Liu
Wei Liu, Li Chen, and Wenyi Zhang
Decentralized Federated Learning: Balancing Communication and Computing Costs
null
null
null
null
cs.LG cs.DC
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Decentralized stochastic gradient descent (SGD) is a driving engine for decentralized federated learning (DFL). The performance of decentralized SGD is jointly influenced by inter-node communications and local updates. In this paper, we propose a general DFL framework, which implements both multiple local updates and multiple inter-node communications periodically, to strike a balance between communication efficiency and model consensus. It can provide a general decentralized SGD analytical framework. We establish strong convergence guarantees for the proposed DFL algorithm without the assumption of convex objectives. The convergence rate of DFL can be optimized to achieve the balance of communication and computing costs under constrained resources. For improving communication efficiency of DFL, compressed communication is further introduced to the proposed DFL as a new scheme, named DFL with compressed communication (C-DFL). The proposed C-DFL exhibits linear convergence for strongly convex objectives. Experiment results based on MNIST and CIFAR-10 datasets illustrate the superiority of DFL over traditional decentralized SGD methods and show that C-DFL further enhances communication efficiency.
[ { "created": "Mon, 26 Jul 2021 09:09:45 GMT", "version": "v1" }, { "created": "Sun, 8 Aug 2021 03:57:17 GMT", "version": "v2" }, { "created": "Wed, 19 Jan 2022 05:10:09 GMT", "version": "v3" }, { "created": "Fri, 11 Feb 2022 04:15:35 GMT", "version": "v4" } ]
2022-02-14
[ [ "Liu", "Wei", "" ], [ "Chen", "Li", "" ], [ "Zhang", "Wenyi", "" ] ]
Decentralized stochastic gradient descent (SGD) is a driving engine for decentralized federated learning (DFL). The performance of decentralized SGD is jointly influenced by inter-node communications and local updates. In this paper, we propose a general DFL framework, which implements both multiple local updates and multiple inter-node communications periodically, to strike a balance between communication efficiency and model consensus. It can provide a general decentralized SGD analytical framework. We establish strong convergence guarantees for the proposed DFL algorithm without the assumption of convex objectives. The convergence rate of DFL can be optimized to achieve the balance of communication and computing costs under constrained resources. For improving communication efficiency of DFL, compressed communication is further introduced to the proposed DFL as a new scheme, named DFL with compressed communication (C-DFL). The proposed C-DFL exhibits linear convergence for strongly convex objectives. Experiment results based on MNIST and CIFAR-10 datasets illustrate the superiority of DFL over traditional decentralized SGD methods and show that C-DFL further enhances communication efficiency.
2205.13804
Moritz Reuss
Moritz Reuss, Niels van Duijkeren, Robert Krug, Philipp Becker, Vaisakh Shaj and Gerhard Neumann
End-to-End Learning of Hybrid Inverse Dynamics Models for Precise and Compliant Impedance Control
Accepted for publication at Robotics: Science and System XVIII (RSS), year 2022. Paper length is 13 pages (i.e. 9 pages of technical content, 1 page of the Bibliography/References and 3 pages of Appendix)
null
null
null
cs.RO cs.LG
http://creativecommons.org/licenses/by-nc-nd/4.0/
It is well-known that inverse dynamics models can improve tracking performance in robot control. These models need to precisely capture the robot dynamics, which consist of well-understood components, e.g., rigid body dynamics, and effects that remain challenging to capture, e.g., stick-slip friction and mechanical flexibilities. Such effects exhibit hysteresis and partial observability, rendering them, particularly challenging to model. Hence, hybrid models, which combine a physical prior with data-driven approaches are especially well-suited in this setting. We present a novel hybrid model formulation that enables us to identify fully physically consistent inertial parameters of a rigid body dynamics model which is paired with a recurrent neural network architecture, allowing us to capture unmodeled partially observable effects using the network memory. We compare our approach against state-of-the-art inverse dynamics models on a 7 degree of freedom manipulator. Using data sets obtained through an optimal experiment design approach, we study the accuracy of offline torque prediction and generalization capabilities of joint learning methods. In control experiments on the real system, we evaluate the model as a feed-forward term for impedance control and show the feedback gains can be drastically reduced to achieve a given tracking accuracy.
[ { "created": "Fri, 27 May 2022 07:39:28 GMT", "version": "v1" } ]
2022-05-30
[ [ "Reuss", "Moritz", "" ], [ "van Duijkeren", "Niels", "" ], [ "Krug", "Robert", "" ], [ "Becker", "Philipp", "" ], [ "Shaj", "Vaisakh", "" ], [ "Neumann", "Gerhard", "" ] ]
It is well-known that inverse dynamics models can improve tracking performance in robot control. These models need to precisely capture the robot dynamics, which consist of well-understood components, e.g., rigid body dynamics, and effects that remain challenging to capture, e.g., stick-slip friction and mechanical flexibilities. Such effects exhibit hysteresis and partial observability, rendering them, particularly challenging to model. Hence, hybrid models, which combine a physical prior with data-driven approaches are especially well-suited in this setting. We present a novel hybrid model formulation that enables us to identify fully physically consistent inertial parameters of a rigid body dynamics model which is paired with a recurrent neural network architecture, allowing us to capture unmodeled partially observable effects using the network memory. We compare our approach against state-of-the-art inverse dynamics models on a 7 degree of freedom manipulator. Using data sets obtained through an optimal experiment design approach, we study the accuracy of offline torque prediction and generalization capabilities of joint learning methods. In control experiments on the real system, we evaluate the model as a feed-forward term for impedance control and show the feedback gains can be drastically reduced to achieve a given tracking accuracy.
2406.11555
Lukas Vierling
Lukas Vierling, Jie Fu, Kai Chen
Input Conditioned Graph Generation for Language Agents
null
null
null
null
cs.CL cs.AI
http://creativecommons.org/licenses/by-nc-sa/4.0/
Recent progress in Large Language Models (LLMs) and language agents has demonstrated significant promise for various future applications across multiple disciplines. While traditional approaches to language agents often rely on fixed, handcrafted designs, our research aims to develop both learnable and dynamic agents. Our method uses an existing framework that abstracts language agents as graphs. Within this graph framework, we aim to learn a model that can generate edges for every given input to the language agent. This allows us to generate edges that represent the flow of communication within the graph based on the given input, thereby adjusting the internal communication of a language agent. We learn to generate these edges using a pretrained LLM that is fine-tuned with reinforcement learning. This LLM can be fine-tuned on several datasets simultaneously, and we hypothesize that the model learns to adapt to these different domains during training, achieving good overall performance when encountering data from different domains during deployment. We demonstrate that our approach surpasses the previous static approach by nearly 6% accuracy on a combined dataset of MMLU and CMMLU, and by more than 10% when trained with a sparsity-inducing loss. It also performs superior in additional experiments conducted with the MMLU and Mini Crossword Puzzles datasets. The code is available at https://github.com/lukasVierling/DynamicGPTSwarm.
[ { "created": "Mon, 17 Jun 2024 13:53:15 GMT", "version": "v1" } ]
2024-06-18
[ [ "Vierling", "Lukas", "" ], [ "Fu", "Jie", "" ], [ "Chen", "Kai", "" ] ]
Recent progress in Large Language Models (LLMs) and language agents has demonstrated significant promise for various future applications across multiple disciplines. While traditional approaches to language agents often rely on fixed, handcrafted designs, our research aims to develop both learnable and dynamic agents. Our method uses an existing framework that abstracts language agents as graphs. Within this graph framework, we aim to learn a model that can generate edges for every given input to the language agent. This allows us to generate edges that represent the flow of communication within the graph based on the given input, thereby adjusting the internal communication of a language agent. We learn to generate these edges using a pretrained LLM that is fine-tuned with reinforcement learning. This LLM can be fine-tuned on several datasets simultaneously, and we hypothesize that the model learns to adapt to these different domains during training, achieving good overall performance when encountering data from different domains during deployment. We demonstrate that our approach surpasses the previous static approach by nearly 6% accuracy on a combined dataset of MMLU and CMMLU, and by more than 10% when trained with a sparsity-inducing loss. It also performs superior in additional experiments conducted with the MMLU and Mini Crossword Puzzles datasets. The code is available at https://github.com/lukasVierling/DynamicGPTSwarm.
2403.01238
Kaituo Feng
Kaituo Feng, Changsheng Li, Dongchun Ren, Ye Yuan, Guoren Wang
On the Road to Portability: Compressing End-to-End Motion Planner for Autonomous Driving
Accepted by CVPR 2024
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
End-to-end motion planning models equipped with deep neural networks have shown great potential for enabling full autonomous driving. However, the oversized neural networks render them impractical for deployment on resource-constrained systems, which unavoidably requires more computational time and resources during reference.To handle this, knowledge distillation offers a promising approach that compresses models by enabling a smaller student model to learn from a larger teacher model. Nevertheless, how to apply knowledge distillation to compress motion planners has not been explored so far. In this paper, we propose PlanKD, the first knowledge distillation framework tailored for compressing end-to-end motion planners. First, considering that driving scenes are inherently complex, often containing planning-irrelevant or even noisy information, transferring such information is not beneficial for the student planner. Thus, we design an information bottleneck based strategy to only distill planning-relevant information, rather than transfer all information indiscriminately. Second, different waypoints in an output planned trajectory may hold varying degrees of importance for motion planning, where a slight deviation in certain crucial waypoints might lead to a collision. Therefore, we devise a safety-aware waypoint-attentive distillation module that assigns adaptive weights to different waypoints based on the importance, to encourage the student to accurately mimic more crucial waypoints, thereby improving overall safety. Experiments demonstrate that our PlanKD can boost the performance of smaller planners by a large margin, and significantly reduce their reference time.
[ { "created": "Sat, 2 Mar 2024 15:47:42 GMT", "version": "v1" }, { "created": "Mon, 15 Apr 2024 07:12:20 GMT", "version": "v2" } ]
2024-04-16
[ [ "Feng", "Kaituo", "" ], [ "Li", "Changsheng", "" ], [ "Ren", "Dongchun", "" ], [ "Yuan", "Ye", "" ], [ "Wang", "Guoren", "" ] ]
End-to-end motion planning models equipped with deep neural networks have shown great potential for enabling full autonomous driving. However, the oversized neural networks render them impractical for deployment on resource-constrained systems, which unavoidably requires more computational time and resources during reference.To handle this, knowledge distillation offers a promising approach that compresses models by enabling a smaller student model to learn from a larger teacher model. Nevertheless, how to apply knowledge distillation to compress motion planners has not been explored so far. In this paper, we propose PlanKD, the first knowledge distillation framework tailored for compressing end-to-end motion planners. First, considering that driving scenes are inherently complex, often containing planning-irrelevant or even noisy information, transferring such information is not beneficial for the student planner. Thus, we design an information bottleneck based strategy to only distill planning-relevant information, rather than transfer all information indiscriminately. Second, different waypoints in an output planned trajectory may hold varying degrees of importance for motion planning, where a slight deviation in certain crucial waypoints might lead to a collision. Therefore, we devise a safety-aware waypoint-attentive distillation module that assigns adaptive weights to different waypoints based on the importance, to encourage the student to accurately mimic more crucial waypoints, thereby improving overall safety. Experiments demonstrate that our PlanKD can boost the performance of smaller planners by a large margin, and significantly reduce their reference time.
1908.03999
Jason Teutsch
Jason Teutsch, Michael Straka, Dan Boneh
Retrofitting a two-way peg between blockchains
null
null
null
null
cs.CR cs.LO econ.TH
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In December 2015, a bounty emerged to establish both reliable communication and secure transfer of value between the Dogecoin and Ethereum blockchains. This prized "Dogethereum bridge" would allow parties to "lock" a DOGE coin on Dogecoin and in exchange receive a newly minted WOW token in Ethereum. Any subsequent owner of the WOW token could burn it and, in exchange, earn the right to "unlock" a DOGE on Dogecoin. We describe an efficient, trustless, and retrofitting Dogethereum construction which requires no fork but rather employs economic collateral to achieve a "lock" operation in Dogecoin. The protocol relies on bulletproofs, Truebit, and parametrized tokens to efficiently and trustlessly relay events from the "true" Dogecoin blockchain into Ethereum. The present construction not only enables cross-platform exchange but also allows Ethereum smart contracts to trustlessly access Dogecoin. A similar technique adds Ethereum-based smart contracts to Bitcoin and Bitcoin data to Ethereum smart contracts.
[ { "created": "Mon, 12 Aug 2019 04:41:13 GMT", "version": "v1" } ]
2019-08-13
[ [ "Teutsch", "Jason", "" ], [ "Straka", "Michael", "" ], [ "Boneh", "Dan", "" ] ]
In December 2015, a bounty emerged to establish both reliable communication and secure transfer of value between the Dogecoin and Ethereum blockchains. This prized "Dogethereum bridge" would allow parties to "lock" a DOGE coin on Dogecoin and in exchange receive a newly minted WOW token in Ethereum. Any subsequent owner of the WOW token could burn it and, in exchange, earn the right to "unlock" a DOGE on Dogecoin. We describe an efficient, trustless, and retrofitting Dogethereum construction which requires no fork but rather employs economic collateral to achieve a "lock" operation in Dogecoin. The protocol relies on bulletproofs, Truebit, and parametrized tokens to efficiently and trustlessly relay events from the "true" Dogecoin blockchain into Ethereum. The present construction not only enables cross-platform exchange but also allows Ethereum smart contracts to trustlessly access Dogecoin. A similar technique adds Ethereum-based smart contracts to Bitcoin and Bitcoin data to Ethereum smart contracts.
1702.06028
Andrea Cerone
Andrea Cerone, Alexey Gotsman, Hongseok Yang
Algebraic Laws for Weak Consistency (Extended Version)
Extended Version of the CONCUR'17 paper
null
10.4230/LIPIcs.CONCUR.2017.22
null
cs.LO
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Modern distributed systems often rely on so called weakly-consistent databases, which achieve scalability by sacrificing the consistency guarantee of distributed transaction processing. Such databases have been formalised in two different styles, one based on abstract executions and the other based on dependency graphs. The choice between these styles has been made according to intended applications: the former has been used to specify and verify the implementation of these databases, and the latter to prove properties of programs running on top of the databases. In this paper, we present a set of novel algebraic laws (i.e. inequations) that connect these two styles of specifications; the laws relate binary relations used in a specification based on abstract executions, to those used in a specification based on dependency graphs. We then show that this algebraic connection gives rise to so called robustness criteria, conditions which ensures that a program running on top of a weakly-consistent database does not exhibit anomalous behaviours due to this weak consistency. These criteria make it easy to reason about programs running on top of these databases, and may become a basis for dynamic or static program analyses. For a certain class of consistency models specifications, we prove a full abstraction result that connects the two styles of specifications.
[ { "created": "Mon, 20 Feb 2017 15:55:20 GMT", "version": "v1" }, { "created": "Fri, 28 Apr 2017 00:11:58 GMT", "version": "v2" }, { "created": "Thu, 4 May 2017 18:36:07 GMT", "version": "v3" }, { "created": "Tue, 1 Aug 2017 15:47:01 GMT", "version": "v4" } ]
2017-08-02
[ [ "Cerone", "Andrea", "" ], [ "Gotsman", "Alexey", "" ], [ "Yang", "Hongseok", "" ] ]
Modern distributed systems often rely on so called weakly-consistent databases, which achieve scalability by sacrificing the consistency guarantee of distributed transaction processing. Such databases have been formalised in two different styles, one based on abstract executions and the other based on dependency graphs. The choice between these styles has been made according to intended applications: the former has been used to specify and verify the implementation of these databases, and the latter to prove properties of programs running on top of the databases. In this paper, we present a set of novel algebraic laws (i.e. inequations) that connect these two styles of specifications; the laws relate binary relations used in a specification based on abstract executions, to those used in a specification based on dependency graphs. We then show that this algebraic connection gives rise to so called robustness criteria, conditions which ensures that a program running on top of a weakly-consistent database does not exhibit anomalous behaviours due to this weak consistency. These criteria make it easy to reason about programs running on top of these databases, and may become a basis for dynamic or static program analyses. For a certain class of consistency models specifications, we prove a full abstraction result that connects the two styles of specifications.
1207.2860
Shafqat Shad Mr
Shafqat Ali Shad, Enhong Chen, Faisal Malik Faisal Azeem
Enterprise Resource Planning - Real blessing or a Blessing in Disguise : An Exploration of the Contextual Factors in Public Sector
null
Interdisciplinary Journal of Contemporary Research in Business, vol. 2, no. 10, pp. 294-307, 2011
null
null
cs.CY
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Information systems have always been in a prime focus in organizations in both local (Pakistani) and global environment. Now the race of being the best through Information Systems has created its importance in public sector organizations to meet the global challenges. Public sector organizations have been facing problems in different segments of technology adoption especially in ERP projects. ERP adoption/implementation projects in public sector organizations still encounter major setbacks in terms of partly/completely success/failure. Cultural and other social barriers have been resistant in technology adoption in Pakistan. Now in the case of big ERP adoptions the contextual factors must be identified and addressed. The paper investigates the reasons of success or failure by addressing nature of complexities regarding different contextual factors. The study includes a sample of Pakistan s four public sector organizations. The sample of this four organizations includes two organizations (Type-A) i.e. Oil & Gas Development Company Limited (OGDCL) and National Database Registration Authority (NADRA) where ERP has been successfully implemented and other two (Type-B) i.e. Pakistan Telecommunication Corporation Limited (PTCL), Higher Education Commission (HEC) where ERP implementation is in progress. The findings address the contextual factors i.e. cultural, environmental & political changes which have a variable impact on ERP systems adoption/implementation in addition to Business Process Re-engineering (BPR). Paper also briefly includes analysis of gaps between pre & post ERP implementation scenarios.
[ { "created": "Thu, 12 Jul 2012 07:24:45 GMT", "version": "v1" }, { "created": "Fri, 21 Sep 2012 02:06:21 GMT", "version": "v2" } ]
2012-09-24
[ [ "Shad", "Shafqat Ali", "" ], [ "Chen", "Enhong", "" ], [ "Azeem", "Faisal Malik Faisal", "" ] ]
Information systems have always been in a prime focus in organizations in both local (Pakistani) and global environment. Now the race of being the best through Information Systems has created its importance in public sector organizations to meet the global challenges. Public sector organizations have been facing problems in different segments of technology adoption especially in ERP projects. ERP adoption/implementation projects in public sector organizations still encounter major setbacks in terms of partly/completely success/failure. Cultural and other social barriers have been resistant in technology adoption in Pakistan. Now in the case of big ERP adoptions the contextual factors must be identified and addressed. The paper investigates the reasons of success or failure by addressing nature of complexities regarding different contextual factors. The study includes a sample of Pakistan s four public sector organizations. The sample of this four organizations includes two organizations (Type-A) i.e. Oil & Gas Development Company Limited (OGDCL) and National Database Registration Authority (NADRA) where ERP has been successfully implemented and other two (Type-B) i.e. Pakistan Telecommunication Corporation Limited (PTCL), Higher Education Commission (HEC) where ERP implementation is in progress. The findings address the contextual factors i.e. cultural, environmental & political changes which have a variable impact on ERP systems adoption/implementation in addition to Business Process Re-engineering (BPR). Paper also briefly includes analysis of gaps between pre & post ERP implementation scenarios.
1508.02086
Hassan Kingravi
Hassan A. Kingravi, Harshal Maske, Girish Chowdhary
Kernel Controllers: A Systems-Theoretic Approach for Data-Driven Modeling and Control of Spatiotemporally Evolving Processes
null
null
null
null
cs.SY
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We consider the problem of modeling, estimating, and controlling the latent state of a spatiotemporally evolving continuous function using very few sensor measurements and actuator locations. Our solution to the problem consists of two parts: a predictive model of functional evolution, and feedback based estimator and controllers that can robustly recover the state of the model and drive it to a desired function. We show that layering a dynamical systems prior over temporal evolution of weights of a kernel model is a valid approach to spatiotemporal modeling that leads to systems theoretic, control-usable, predictive models. We provide sufficient conditions on the number of sensors and actuators required to guarantee observability and controllability. The approach is validated on a large real dataset, and in simulation for the control of spatiotemporally evolving function.
[ { "created": "Sun, 9 Aug 2015 21:26:55 GMT", "version": "v1" } ]
2015-08-11
[ [ "Kingravi", "Hassan A.", "" ], [ "Maske", "Harshal", "" ], [ "Chowdhary", "Girish", "" ] ]
We consider the problem of modeling, estimating, and controlling the latent state of a spatiotemporally evolving continuous function using very few sensor measurements and actuator locations. Our solution to the problem consists of two parts: a predictive model of functional evolution, and feedback based estimator and controllers that can robustly recover the state of the model and drive it to a desired function. We show that layering a dynamical systems prior over temporal evolution of weights of a kernel model is a valid approach to spatiotemporal modeling that leads to systems theoretic, control-usable, predictive models. We provide sufficient conditions on the number of sensors and actuators required to guarantee observability and controllability. The approach is validated on a large real dataset, and in simulation for the control of spatiotemporally evolving function.
1804.07379
Garegin Grigoryan
Garegin Grigoryan, Yaoqing Liu
Toward a Programmable FIB Caching Architecture
null
Network Protocols (ICNP), 2017 IEEE 25th International Conference on, 1-2
null
null
cs.NI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The current Internet routing ecosystem is neither sustainable nor economical. More than 711K IPv4 routes and more than 41K IPv6 routes exist in current global Forwarding Information Base (FIBs) with growth rates increasing. This rapid growth has serious consequences, such as creating the need for costly FIB memory upgrades and increased potential for Internet service outages. And while FIB memories are power-hungry and prohibitively expensive, more than 70\% of the routes in FIBs carry no traffic for long time periods, a wasteful use of these expensive resources. Taking advantage of the emerging concept of programmable data plane, we design a programmable FIB caching architecture to address the existing concerns. Our preliminary evaluation results show that the architecture can significantly mitigate the global routing scalability and poor FIB utilization issues.
[ { "created": "Thu, 19 Apr 2018 21:10:17 GMT", "version": "v1" } ]
2018-04-23
[ [ "Grigoryan", "Garegin", "" ], [ "Liu", "Yaoqing", "" ] ]
The current Internet routing ecosystem is neither sustainable nor economical. More than 711K IPv4 routes and more than 41K IPv6 routes exist in current global Forwarding Information Base (FIBs) with growth rates increasing. This rapid growth has serious consequences, such as creating the need for costly FIB memory upgrades and increased potential for Internet service outages. And while FIB memories are power-hungry and prohibitively expensive, more than 70\% of the routes in FIBs carry no traffic for long time periods, a wasteful use of these expensive resources. Taking advantage of the emerging concept of programmable data plane, we design a programmable FIB caching architecture to address the existing concerns. Our preliminary evaluation results show that the architecture can significantly mitigate the global routing scalability and poor FIB utilization issues.
1206.5247
Daniel Eaton
Daniel Eaton, Kevin Murphy
Bayesian structure learning using dynamic programming and MCMC
Appears in Proceedings of the Twenty-Third Conference on Uncertainty in Artificial Intelligence (UAI2007)
null
null
UAI-P-2007-PG-101-108
cs.LG stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
MCMC methods for sampling from the space of DAGs can mix poorly due to the local nature of the proposals that are commonly used. It has been shown that sampling from the space of node orders yields better results [FK03, EW06]. Recently, Koivisto and Sood showed how one can analytically marginalize over orders using dynamic programming (DP) [KS04, Koi06]. Their method computes the exact marginal posterior edge probabilities, thus avoiding the need for MCMC. Unfortunately, there are four drawbacks to the DP technique: it can only use modular priors, it can only compute posteriors over modular features, it is difficult to compute a predictive density, and it takes exponential time and space. We show how to overcome the first three of these problems by using the DP algorithm as a proposal distribution for MCMC in DAG space. We show that this hybrid technique converges to the posterior faster than other methods, resulting in more accurate structure learning and higher predictive likelihoods on test data.
[ { "created": "Wed, 20 Jun 2012 14:54:43 GMT", "version": "v1" } ]
2012-06-26
[ [ "Eaton", "Daniel", "" ], [ "Murphy", "Kevin", "" ] ]
MCMC methods for sampling from the space of DAGs can mix poorly due to the local nature of the proposals that are commonly used. It has been shown that sampling from the space of node orders yields better results [FK03, EW06]. Recently, Koivisto and Sood showed how one can analytically marginalize over orders using dynamic programming (DP) [KS04, Koi06]. Their method computes the exact marginal posterior edge probabilities, thus avoiding the need for MCMC. Unfortunately, there are four drawbacks to the DP technique: it can only use modular priors, it can only compute posteriors over modular features, it is difficult to compute a predictive density, and it takes exponential time and space. We show how to overcome the first three of these problems by using the DP algorithm as a proposal distribution for MCMC in DAG space. We show that this hybrid technique converges to the posterior faster than other methods, resulting in more accurate structure learning and higher predictive likelihoods on test data.
1004.3580
Loet Leydesdorff
Loet Leydesdorff, Tobias Opthof
Scopus's Source Normalized Impact per Paper (SNIP) versus a Journal Impact Factor based on Fractional Counting of Citations
null
null
null
null
cs.DL physics.soc-ph
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Impact factors (and similar measures such as the Scimago Journal Rankings) suffer from two problems: (i) citation behavior varies among fields of science and therefore leads to systematic differences, and (ii) there are no statistics to inform us whether differences are significant. The recently introduced SNIP indicator of Scopus tries to remedy the first of these two problems, but a number of normalization decisions are involved which makes it impossible to test for significance. Using fractional counting of citations-based on the assumption that impact is proportionate to the number of references in the citing documents-citations can be contextualized at the paper level and aggregated impacts of sets can be tested for their significance. It can be shown that the weighted impact of Annals of Mathematics (0.247) is not so much lower than that of Molecular Cell (0.386) despite a five-fold difference between their impact factors (2.793 and 13.156, respectively).
[ { "created": "Tue, 20 Apr 2010 21:17:52 GMT", "version": "v1" }, { "created": "Sun, 25 Apr 2010 07:08:46 GMT", "version": "v2" } ]
2010-04-27
[ [ "Leydesdorff", "Loet", "" ], [ "Opthof", "Tobias", "" ] ]
Impact factors (and similar measures such as the Scimago Journal Rankings) suffer from two problems: (i) citation behavior varies among fields of science and therefore leads to systematic differences, and (ii) there are no statistics to inform us whether differences are significant. The recently introduced SNIP indicator of Scopus tries to remedy the first of these two problems, but a number of normalization decisions are involved which makes it impossible to test for significance. Using fractional counting of citations-based on the assumption that impact is proportionate to the number of references in the citing documents-citations can be contextualized at the paper level and aggregated impacts of sets can be tested for their significance. It can be shown that the weighted impact of Annals of Mathematics (0.247) is not so much lower than that of Molecular Cell (0.386) despite a five-fold difference between their impact factors (2.793 and 13.156, respectively).
1710.09177
Stefan M. Moser
Stefan M. Moser, Ligong Wang, Mich\`ele Wigger
Capacity Results on Multiple-Input Single-Output Wireless Optical Channels
Submitted to IEEE Transactions on Information Theory
null
null
null
cs.IT math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This paper derives upper and lower bounds on the capacity of the multiple-input single-output free-space optical intensity channel with signal-independent additive Gaussian noise subject to both an average-intensity and a peak-intensity constraint. In the limit where the signal-to-noise ratio (SNR) tends to infinity, the asymptotic capacity is specified, while in the limit where the SNR tends to zero, the exact slope of the capacity is also given.
[ { "created": "Wed, 25 Oct 2017 11:36:08 GMT", "version": "v1" } ]
2017-10-26
[ [ "Moser", "Stefan M.", "" ], [ "Wang", "Ligong", "" ], [ "Wigger", "Michèle", "" ] ]
This paper derives upper and lower bounds on the capacity of the multiple-input single-output free-space optical intensity channel with signal-independent additive Gaussian noise subject to both an average-intensity and a peak-intensity constraint. In the limit where the signal-to-noise ratio (SNR) tends to infinity, the asymptotic capacity is specified, while in the limit where the SNR tends to zero, the exact slope of the capacity is also given.
2404.18708
Andy L\"ucking
Andy L\"ucking, Alexander Henlein, Alexander Mehler
Iconic Gesture Semantics
39 pages, 28 figures, under revision
null
null
null
cs.CL
http://creativecommons.org/licenses/by-sa/4.0/
The "meaning" of an iconic gesture is conditioned on its informational evaluation. Only informational evaluation lifts a gesture to a quasi-linguistic level that can interact with verbal content. Interaction is either vacuous or regimented by usual lexicon-driven inferences. Informational evaluation is spelled out as extended exemplification (extemplification) in terms of perceptual classification of a gesture's visual iconic model. The iconic model is derived from Frege/Montague-like truth-functional evaluation of a gesture's form within spatially extended domains. We further argue that the perceptual classification of instances of visual communication requires a notion of meaning different from Frege/Montague frameworks. Therefore, a heuristic for gesture interpretation is provided that can guide the working semanticist. In sum, an iconic gesture semantics is introduced which covers the full range from kinematic gesture representations over model-theoretic evaluation to inferential interpretation in dynamic semantic frameworks.
[ { "created": "Mon, 29 Apr 2024 13:58:03 GMT", "version": "v1" } ]
2024-04-30
[ [ "Lücking", "Andy", "" ], [ "Henlein", "Alexander", "" ], [ "Mehler", "Alexander", "" ] ]
The "meaning" of an iconic gesture is conditioned on its informational evaluation. Only informational evaluation lifts a gesture to a quasi-linguistic level that can interact with verbal content. Interaction is either vacuous or regimented by usual lexicon-driven inferences. Informational evaluation is spelled out as extended exemplification (extemplification) in terms of perceptual classification of a gesture's visual iconic model. The iconic model is derived from Frege/Montague-like truth-functional evaluation of a gesture's form within spatially extended domains. We further argue that the perceptual classification of instances of visual communication requires a notion of meaning different from Frege/Montague frameworks. Therefore, a heuristic for gesture interpretation is provided that can guide the working semanticist. In sum, an iconic gesture semantics is introduced which covers the full range from kinematic gesture representations over model-theoretic evaluation to inferential interpretation in dynamic semantic frameworks.
0905.0283
Kevin Wortman
David Eppstein and Kevin A. Wortman
Optimal Embedding Into Star Metrics
12 pages, 3 figures
null
null
null
cs.DS
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We present an O(n^3 log^2 n)-time algorithm for the following problem: given a finite metric space X, create a star-topology network with the points of X as its leaves, such that the distances in the star are at least as large as in X, with minimum dilation. As part of our algorithm, we solve in the same time bound the parametric negative cycle detection problem: given a directed graph with edge weights that are increasing linear functions of a parameter lambda, find the smallest value of lambda such that the graph contains no negative-weight cycles.
[ { "created": "Sun, 3 May 2009 19:21:52 GMT", "version": "v1" } ]
2009-05-05
[ [ "Eppstein", "David", "" ], [ "Wortman", "Kevin A.", "" ] ]
We present an O(n^3 log^2 n)-time algorithm for the following problem: given a finite metric space X, create a star-topology network with the points of X as its leaves, such that the distances in the star are at least as large as in X, with minimum dilation. As part of our algorithm, we solve in the same time bound the parametric negative cycle detection problem: given a directed graph with edge weights that are increasing linear functions of a parameter lambda, find the smallest value of lambda such that the graph contains no negative-weight cycles.
1203.3923
Muhammad Anshari Mr
Mohammad Nabil Almunawar and Muhammad Anshari
Health Information Systems (HIS): Concept and Technology
International Conference Informatics Development, 2011
null
null
null
cs.OH
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
A health information system (HIS) is the intersection of between healthcare's business process, and information systems to deliver better healthcare services. The nature of healthcare industry, which is highly influenced by economic, social, politic, and technological factors, has changed over time. This paper will address some important concepts of healthcare and related terminologies to provide a holistic view for HIS. Related technological milestones and major events are briefly summarized. The trends and rapid development of health information technologies are also discussed.
[ { "created": "Sun, 18 Mar 2012 06:59:22 GMT", "version": "v1" } ]
2012-03-20
[ [ "Almunawar", "Mohammad Nabil", "" ], [ "Anshari", "Muhammad", "" ] ]
A health information system (HIS) is the intersection of between healthcare's business process, and information systems to deliver better healthcare services. The nature of healthcare industry, which is highly influenced by economic, social, politic, and technological factors, has changed over time. This paper will address some important concepts of healthcare and related terminologies to provide a holistic view for HIS. Related technological milestones and major events are briefly summarized. The trends and rapid development of health information technologies are also discussed.
1401.3449
Vincent Conitzer
Vincent Conitzer
Eliciting Single-Peaked Preferences Using Comparison Queries
null
Journal Of Artificial Intelligence Research, Volume 35, pages 161-191, 2009
10.1613/jair.2606
null
cs.GT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Voting is a general method for aggregating the preferences of multiple agents. Each agent ranks all the possible alternatives, and based on this, an aggregate ranking of the alternatives (or at least a winning alternative) is produced. However, when there are many alternatives, it is impractical to simply ask agents to report their complete preferences. Rather, the agents preferences, or at least the relevant parts thereof, need to be elicited. This is done by asking the agents a (hopefully small) number of simple queries about their preferences, such as comparison queries, which ask an agent to compare two of the alternatives. Prior work on preference elicitation in voting has focused on the case of unrestricted preferences. It has been shown that in this setting, it is sometimes necessary to ask each agent (almost) as many queries as would be required to determine an arbitrary ranking of the alternatives. In contrast, in this paper, we focus on single-peaked preferences. We show that such preferences can be elicited using only a linear number of comparison queries, if either the order with respect to which preferences are single-peaked is known, or at least one other agents complete preferences are known. We show that using a sublinear number of queries does not suffice. We also consider the case of cardinally single-peaked preferences. For this case, we show that if the alternatives cardinal positions are known, then an agents preferences can be elicited using only a logarithmic number of queries; however, we also show that if the cardinal positions are not known, then a sublinear number of queries does not suffice. We present experimental results for all elicitation algorithms. We also consider the problem of only eliciting enough information to determine the aggregate ranking, and show that even for this more modest objective, a sublinear number of queries per agent does not suffice for known ordinal or unknown cardinal positions. Finally, we discuss whether and how these techniques can be applied when preferences are almost single-peaked.
[ { "created": "Wed, 15 Jan 2014 05:10:11 GMT", "version": "v1" } ]
2014-01-16
[ [ "Conitzer", "Vincent", "" ] ]
Voting is a general method for aggregating the preferences of multiple agents. Each agent ranks all the possible alternatives, and based on this, an aggregate ranking of the alternatives (or at least a winning alternative) is produced. However, when there are many alternatives, it is impractical to simply ask agents to report their complete preferences. Rather, the agents preferences, or at least the relevant parts thereof, need to be elicited. This is done by asking the agents a (hopefully small) number of simple queries about their preferences, such as comparison queries, which ask an agent to compare two of the alternatives. Prior work on preference elicitation in voting has focused on the case of unrestricted preferences. It has been shown that in this setting, it is sometimes necessary to ask each agent (almost) as many queries as would be required to determine an arbitrary ranking of the alternatives. In contrast, in this paper, we focus on single-peaked preferences. We show that such preferences can be elicited using only a linear number of comparison queries, if either the order with respect to which preferences are single-peaked is known, or at least one other agents complete preferences are known. We show that using a sublinear number of queries does not suffice. We also consider the case of cardinally single-peaked preferences. For this case, we show that if the alternatives cardinal positions are known, then an agents preferences can be elicited using only a logarithmic number of queries; however, we also show that if the cardinal positions are not known, then a sublinear number of queries does not suffice. We present experimental results for all elicitation algorithms. We also consider the problem of only eliciting enough information to determine the aggregate ranking, and show that even for this more modest objective, a sublinear number of queries per agent does not suffice for known ordinal or unknown cardinal positions. Finally, we discuss whether and how these techniques can be applied when preferences are almost single-peaked.
2211.11962
Hai Wu
Hai Wu and Chenglu Wen and Wei Li and Xin Li and Ruigang Yang and Cheng Wang
Transformation-Equivariant 3D Object Detection for Autonomous Driving
Accepted by AAAI 2023
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
3D object detection received increasing attention in autonomous driving recently. Objects in 3D scenes are distributed with diverse orientations. Ordinary detectors do not explicitly model the variations of rotation and reflection transformations. Consequently, large networks and extensive data augmentation are required for robust detection. Recent equivariant networks explicitly model the transformation variations by applying shared networks on multiple transformed point clouds, showing great potential in object geometry modeling. However, it is difficult to apply such networks to 3D object detection in autonomous driving due to its large computation cost and slow reasoning speed. In this work, we present TED, an efficient Transformation-Equivariant 3D Detector to overcome the computation cost and speed issues. TED first applies a sparse convolution backbone to extract multi-channel transformation-equivariant voxel features; and then aligns and aggregates these equivariant features into lightweight and compact representations for high-performance 3D object detection. On the highly competitive KITTI 3D car detection leaderboard, TED ranked 1st among all submissions with competitive efficiency.
[ { "created": "Tue, 22 Nov 2022 02:51:56 GMT", "version": "v1" }, { "created": "Wed, 23 Nov 2022 01:51:39 GMT", "version": "v2" }, { "created": "Thu, 1 Dec 2022 08:00:16 GMT", "version": "v3" } ]
2022-12-02
[ [ "Wu", "Hai", "" ], [ "Wen", "Chenglu", "" ], [ "Li", "Wei", "" ], [ "Li", "Xin", "" ], [ "Yang", "Ruigang", "" ], [ "Wang", "Cheng", "" ] ]
3D object detection received increasing attention in autonomous driving recently. Objects in 3D scenes are distributed with diverse orientations. Ordinary detectors do not explicitly model the variations of rotation and reflection transformations. Consequently, large networks and extensive data augmentation are required for robust detection. Recent equivariant networks explicitly model the transformation variations by applying shared networks on multiple transformed point clouds, showing great potential in object geometry modeling. However, it is difficult to apply such networks to 3D object detection in autonomous driving due to its large computation cost and slow reasoning speed. In this work, we present TED, an efficient Transformation-Equivariant 3D Detector to overcome the computation cost and speed issues. TED first applies a sparse convolution backbone to extract multi-channel transformation-equivariant voxel features; and then aligns and aggregates these equivariant features into lightweight and compact representations for high-performance 3D object detection. On the highly competitive KITTI 3D car detection leaderboard, TED ranked 1st among all submissions with competitive efficiency.
2209.09543
Peter Belcak
Peter Belc\'ak, Ard Kastrati, Flavio Schenker, Roger Wattenhofer
FACT: Learning Governing Abstractions Behind Integer Sequences
Accepted to the 36th Conference on Neural Information Processing Systems (NeurIPS 2022) Track on Datasets and Benchmarks. 37 pages
null
null
null
cs.LG cs.AI cs.SC
http://creativecommons.org/licenses/by/4.0/
Integer sequences are of central importance to the modeling of concepts admitting complete finitary descriptions. We introduce a novel view on the learning of such concepts and lay down a set of benchmarking tasks aimed at conceptual understanding by machine learning models. These tasks indirectly assess model ability to abstract, and challenge them to reason both interpolatively and extrapolatively from the knowledge gained by observing representative examples. To further aid research in knowledge representation and reasoning, we present FACT, the Finitary Abstraction Comprehension Toolkit. The toolkit surrounds a large dataset of integer sequences comprising both organic and synthetic entries, a library for data pre-processing and generation, a set of model performance evaluation tools, and a collection of baseline model implementations, enabling the making of the future advancements with ease.
[ { "created": "Tue, 20 Sep 2022 08:20:03 GMT", "version": "v1" } ]
2022-09-21
[ [ "Belcák", "Peter", "" ], [ "Kastrati", "Ard", "" ], [ "Schenker", "Flavio", "" ], [ "Wattenhofer", "Roger", "" ] ]
Integer sequences are of central importance to the modeling of concepts admitting complete finitary descriptions. We introduce a novel view on the learning of such concepts and lay down a set of benchmarking tasks aimed at conceptual understanding by machine learning models. These tasks indirectly assess model ability to abstract, and challenge them to reason both interpolatively and extrapolatively from the knowledge gained by observing representative examples. To further aid research in knowledge representation and reasoning, we present FACT, the Finitary Abstraction Comprehension Toolkit. The toolkit surrounds a large dataset of integer sequences comprising both organic and synthetic entries, a library for data pre-processing and generation, a set of model performance evaluation tools, and a collection of baseline model implementations, enabling the making of the future advancements with ease.
2312.12676
Morteza Haghir Chehreghani
Jack Sandberg, Niklas {\AA}kerblom, Morteza Haghir Chehreghani
Combinatorial Gaussian Process Bandits in Bayesian Settings: Theory and Application for Energy-Efficient Navigation
39 pages, 10 figures
null
null
null
cs.LG cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We consider a combinatorial Gaussian process semi-bandit problem with time-varying arm availability. Each round, an agent is provided a set of available base arms and must select a subset of them to maximize the long-term cumulative reward. Assuming the expected rewards are sampled from a Gaussian process (GP) over the arm space, the agent can efficiently learn. We study the Bayesian setting and provide novel Bayesian regret bounds for three GP-based algorithms: GP-UCB, Bayes-GP-UCB and GP-TS. Our bounds extend previous results for GP-UCB and GP-TS to a combinatorial setting with varying arm availability and to the best of our knowledge, we provide the first Bayesian regret bound for Bayes-GP-UCB. Time-varying arm availability encompasses other widely considered bandit problems such as contextual bandits. We formulate the online energy-efficient navigation problem as a combinatorial and contextual bandit and provide a comprehensive experimental study on synthetic and real-world road networks with detailed simulations. The contextual GP model obtains lower regret and is less dependent on the informativeness of the prior compared to the non-contextual Bayesian inference model. In addition, Thompson sampling obtains lower regret than Bayes-UCB for both the contextual and non-contextual model.
[ { "created": "Wed, 20 Dec 2023 00:31:43 GMT", "version": "v1" } ]
2023-12-21
[ [ "Sandberg", "Jack", "" ], [ "Åkerblom", "Niklas", "" ], [ "Chehreghani", "Morteza Haghir", "" ] ]
We consider a combinatorial Gaussian process semi-bandit problem with time-varying arm availability. Each round, an agent is provided a set of available base arms and must select a subset of them to maximize the long-term cumulative reward. Assuming the expected rewards are sampled from a Gaussian process (GP) over the arm space, the agent can efficiently learn. We study the Bayesian setting and provide novel Bayesian regret bounds for three GP-based algorithms: GP-UCB, Bayes-GP-UCB and GP-TS. Our bounds extend previous results for GP-UCB and GP-TS to a combinatorial setting with varying arm availability and to the best of our knowledge, we provide the first Bayesian regret bound for Bayes-GP-UCB. Time-varying arm availability encompasses other widely considered bandit problems such as contextual bandits. We formulate the online energy-efficient navigation problem as a combinatorial and contextual bandit and provide a comprehensive experimental study on synthetic and real-world road networks with detailed simulations. The contextual GP model obtains lower regret and is less dependent on the informativeness of the prior compared to the non-contextual Bayesian inference model. In addition, Thompson sampling obtains lower regret than Bayes-UCB for both the contextual and non-contextual model.
1007.5139
Secretary Ijaia
Anuradha Banerjee (1) and Paramartha Dutta (2) ((1) Kalyani Govt. Engg. College, India and (2) Visva-Bharati University, India)
Reputation-Based Attack-Resistant Cooperation Stimulation (RACS) For Mobile Ad hoc Networks
20 pages, 4 figures
International Journal of Artificial Intelligence & Applications 1.3 (2010) 71-90
10.5121/ijaia.2010.1306
null
cs.NI
http://creativecommons.org/licenses/by-nc-sa/3.0/
In mobile ad hoc networks (MANET), nodes usually belong to different authorities and pursue different goals. In order to maximize their own performance, nodes in such networks tend to be selfish and are not willing to forward packets for benefit of others. Meanwhile, some nodes may behave maliciously and try to disrupt the network through wasting other nodes resources in a very large scale. In this article, we present a reputation-based attack resistant cooperation stimulation (RACS) system which ensures that damage caused by malicious nodes can be bounded and cooperation among the selfish nodes can be enforced. Mathematical analyses of the system as well as the simulation results have confirmed effectiveness of our proposed system. RACS is completely self-organizing and distributed. It does not require any tamper-proof hardware or central management policy.
[ { "created": "Thu, 29 Jul 2010 07:54:51 GMT", "version": "v1" } ]
2010-07-30
[ [ "Banerjee", "Anuradha", "" ], [ "Dutta", "Paramartha", "" ] ]
In mobile ad hoc networks (MANET), nodes usually belong to different authorities and pursue different goals. In order to maximize their own performance, nodes in such networks tend to be selfish and are not willing to forward packets for benefit of others. Meanwhile, some nodes may behave maliciously and try to disrupt the network through wasting other nodes resources in a very large scale. In this article, we present a reputation-based attack resistant cooperation stimulation (RACS) system which ensures that damage caused by malicious nodes can be bounded and cooperation among the selfish nodes can be enforced. Mathematical analyses of the system as well as the simulation results have confirmed effectiveness of our proposed system. RACS is completely self-organizing and distributed. It does not require any tamper-proof hardware or central management policy.
2407.08564
Hengshu Zhu
Meng Hua, Yuan Cheng, Hengshu Zhu
The Career Interests of Large Language Models
null
null
null
null
cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Recent advancements in Large Language Models (LLMs) have significantly extended their capabilities, evolving from basic text generation to complex, human-like interactions. In light of the possibilities that LLMs could assume significant workplace responsibilities, it becomes imminently necessary to explore LLMs' capacities as professional assistants. This study focuses on the aspect of career interests by applying the Occupation Network's Interest Profiler short form to LLMs as if they were human participants and investigates their hypothetical career interests and competence, examining how these vary with language changes and model advancements. We analyzed the answers using a general linear mixed model approach and found distinct career interest inclinations among LLMs, particularly towards the social and artistic domains. Interestingly, these preferences did not align with the occupations where LLMs exhibited higher competence. This novel approach of using psychometric instruments and sophisticated statistical tools on LLMs unveils fresh perspectives on their integration into professional environments, highlighting human-like tendencies and promoting a reevaluation of LLMs' self-perception and competency alignment in the workforce.
[ { "created": "Thu, 11 Jul 2024 14:54:46 GMT", "version": "v1" } ]
2024-07-12
[ [ "Hua", "Meng", "" ], [ "Cheng", "Yuan", "" ], [ "Zhu", "Hengshu", "" ] ]
Recent advancements in Large Language Models (LLMs) have significantly extended their capabilities, evolving from basic text generation to complex, human-like interactions. In light of the possibilities that LLMs could assume significant workplace responsibilities, it becomes imminently necessary to explore LLMs' capacities as professional assistants. This study focuses on the aspect of career interests by applying the Occupation Network's Interest Profiler short form to LLMs as if they were human participants and investigates their hypothetical career interests and competence, examining how these vary with language changes and model advancements. We analyzed the answers using a general linear mixed model approach and found distinct career interest inclinations among LLMs, particularly towards the social and artistic domains. Interestingly, these preferences did not align with the occupations where LLMs exhibited higher competence. This novel approach of using psychometric instruments and sophisticated statistical tools on LLMs unveils fresh perspectives on their integration into professional environments, highlighting human-like tendencies and promoting a reevaluation of LLMs' self-perception and competency alignment in the workforce.
2204.11343
Saeed Banaeian Far
Saeed Banaeian Far, Azadeh Imani Rad
Applying Digital Twins in Metaverse: User Interface, Security and Privacy Challenges
This article has been accepted in "Journal of Metaverse". You can cite as (APA): Banaeian Far, S. & Imani Rad, A. (2022). Applying Digital Twins in Metaverse: User Interface, Security and Privacy Challenges. Journal of Metaverse, 2 (1), 8-16. Retrieved from https://dergipark.org.tr/en/pub/jmv/issue/67967/1072189
2022
null
null
cs.NI
http://creativecommons.org/licenses/by/4.0/
Digital Twins (DTs) are a conventional and well-known concept, proposed in 70s, that are popular in a broad spectrum of sciences, industry innovations, and consortium alliances. However, in the last few years, the growth of digital assets and online communications has attracted attention to DTs as highly accurate twins of physical objects. Metaverse, as a digital world, is a concept proposed in 1992 and has also become a popular paradigm and hot topic in public where DTs can play critical roles. This study first presents definitions, applications, and general challenges of DT and Metaverse. It then offers a three-layer architecture linking the physical world to the Metaverse through a user interface. Further, it investigates the security and privacy challenges of using DTs in Metaverse. Finally, a conclusion, including possible solutions for mentioned challenges and future works, will be provided.
[ { "created": "Sun, 24 Apr 2022 19:41:05 GMT", "version": "v1" } ]
2022-04-26
[ [ "Far", "Saeed Banaeian", "" ], [ "Rad", "Azadeh Imani", "" ] ]
Digital Twins (DTs) are a conventional and well-known concept, proposed in 70s, that are popular in a broad spectrum of sciences, industry innovations, and consortium alliances. However, in the last few years, the growth of digital assets and online communications has attracted attention to DTs as highly accurate twins of physical objects. Metaverse, as a digital world, is a concept proposed in 1992 and has also become a popular paradigm and hot topic in public where DTs can play critical roles. This study first presents definitions, applications, and general challenges of DT and Metaverse. It then offers a three-layer architecture linking the physical world to the Metaverse through a user interface. Further, it investigates the security and privacy challenges of using DTs in Metaverse. Finally, a conclusion, including possible solutions for mentioned challenges and future works, will be provided.
1706.00977
Vashist Avadhanula
Shipra Agrawal, Vashist Avadhanula, Vineet Goyal, Assaf Zeevi
Thompson Sampling for the MNL-Bandit
Accepted for presentation at Conference on Learning Theory (COLT) 2017
null
null
null
cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We consider a sequential subset selection problem under parameter uncertainty, where at each time step, the decision maker selects a subset of cardinality $K$ from $N$ possible items (arms), and observes a (bandit) feedback in the form of the index of one of the items in said subset, or none. Each item in the index set is ascribed a certain value (reward), and the feedback is governed by a Multinomial Logit (MNL) choice model whose parameters are a priori unknown. The objective of the decision maker is to maximize the expected cumulative rewards over a finite horizon $T$, or alternatively, minimize the regret relative to an oracle that knows the MNL parameters. We refer to this as the MNL-Bandit problem. This problem is representative of a larger family of exploration-exploitation problems that involve a combinatorial objective, and arise in several important application domains. We present an approach to adapt Thompson Sampling to this problem and show that it achieves near-optimal regret as well as attractive numerical performance.
[ { "created": "Sat, 3 Jun 2017 16:48:34 GMT", "version": "v1" }, { "created": "Tue, 13 Jun 2017 09:47:40 GMT", "version": "v2" }, { "created": "Sat, 1 Jul 2017 17:36:16 GMT", "version": "v3" }, { "created": "Sat, 27 Oct 2018 09:53:17 GMT", "version": "v4" }, { "created": "Wed, 31 Oct 2018 06:57:46 GMT", "version": "v5" }, { "created": "Wed, 19 Dec 2018 23:14:39 GMT", "version": "v6" }, { "created": "Thu, 3 Jan 2019 19:45:01 GMT", "version": "v7" } ]
2019-01-07
[ [ "Agrawal", "Shipra", "" ], [ "Avadhanula", "Vashist", "" ], [ "Goyal", "Vineet", "" ], [ "Zeevi", "Assaf", "" ] ]
We consider a sequential subset selection problem under parameter uncertainty, where at each time step, the decision maker selects a subset of cardinality $K$ from $N$ possible items (arms), and observes a (bandit) feedback in the form of the index of one of the items in said subset, or none. Each item in the index set is ascribed a certain value (reward), and the feedback is governed by a Multinomial Logit (MNL) choice model whose parameters are a priori unknown. The objective of the decision maker is to maximize the expected cumulative rewards over a finite horizon $T$, or alternatively, minimize the regret relative to an oracle that knows the MNL parameters. We refer to this as the MNL-Bandit problem. This problem is representative of a larger family of exploration-exploitation problems that involve a combinatorial objective, and arise in several important application domains. We present an approach to adapt Thompson Sampling to this problem and show that it achieves near-optimal regret as well as attractive numerical performance.
2007.09834
Ioannis Korkontzelos
Isa Inuwa-Dutse, Mark Liptrott and Ioannis Korkontzelos
Migration and Refugee Crisis: a Critical Analysis of Online Public Perception
15 pages, 8 figures
null
null
null
cs.SI cs.CY
http://creativecommons.org/licenses/by/4.0/
The migration rate and the level of resentments towards migrants are an important issue in modern civilisation. The infamous EU refugee crisis caught many countries unprepared, leading to sporadic and rudimentary containment measures that, in turn, led to significant public discourse. Decades of offline data collected via traditional survey methods have been utilised earlier to understand public opinion to foster peaceful coexistence. Capturing and understanding online public opinion via social media is crucial towards a joint strategic regulation spanning safety, rights of migrants and cordial integration for economic prosperity. We present a analysis of opinions on migrants and refugees expressed by the users of a very popular social platform, Twitter. We analyse sentiment and the associated context of expressions in a vast collection of tweets related to the EU refugee crisis. Our study reveals a marginally higher proportion of negative sentiments vis-a-vis migrants and a large proportion of the negative sentiments is more reflected among the ordinary users. Users with many followers and non-governmental organisations (NGO) tend to tweet favourably about the topic, offsetting the distribution of negative sentiment. We opine that they can be encouraged to be more proactive in neutralising negative attitudes that may arise concerning similar incidences.
[ { "created": "Mon, 20 Jul 2020 02:04:01 GMT", "version": "v1" } ]
2020-07-21
[ [ "Inuwa-Dutse", "Isa", "" ], [ "Liptrott", "Mark", "" ], [ "Korkontzelos", "Ioannis", "" ] ]
The migration rate and the level of resentments towards migrants are an important issue in modern civilisation. The infamous EU refugee crisis caught many countries unprepared, leading to sporadic and rudimentary containment measures that, in turn, led to significant public discourse. Decades of offline data collected via traditional survey methods have been utilised earlier to understand public opinion to foster peaceful coexistence. Capturing and understanding online public opinion via social media is crucial towards a joint strategic regulation spanning safety, rights of migrants and cordial integration for economic prosperity. We present a analysis of opinions on migrants and refugees expressed by the users of a very popular social platform, Twitter. We analyse sentiment and the associated context of expressions in a vast collection of tweets related to the EU refugee crisis. Our study reveals a marginally higher proportion of negative sentiments vis-a-vis migrants and a large proportion of the negative sentiments is more reflected among the ordinary users. Users with many followers and non-governmental organisations (NGO) tend to tweet favourably about the topic, offsetting the distribution of negative sentiment. We opine that they can be encouraged to be more proactive in neutralising negative attitudes that may arise concerning similar incidences.
2209.04171
Anastasios Papazafeiropoulos
Anastasios Papazafeiropoulos, Ioannis Krikidis, Pandelis Kourtessis
Impact of Channel Aging on Reconfigurable Intelligent Surface Aided Massive MIMO Systems with Statistical CSI
accepted in IEEE TVT
null
null
null
cs.IT math.IT
http://creativecommons.org/licenses/by/4.0/
The incorporation of reconfigurable intelligent surface (RIS) into massive multiple-input-multiple-output (mMIMO) systems can unleash the potential of next-generation networks by improving the performance of user equipments (UEs) in service dead zones. However, their requirement for accurate channel state information (CSI) is critical, and especially, applications with UE mobility that induce channel aging make challenging the achievement of adequate quality of service. Hence, in this work, we investigate the impact of channel aging on the performance of RIS-assisted mMIMO systems under both spatial correlation and imperfect CSI conditions. Specifically, by accounting for channel aging during both uplink training and downlink data transmission phases, we first perform minimum mean square error (MMSE) channel estimation to obtain the UE effective channels with low overhead similar to conventional systems without RIS. Next, we derive the downlink achievable sum spectral efficiency (SE) with regularized zero-forcing (RZF) precoding in closed-form being dependent only on large-scale statistics by using the deterministic equivalent (DE) analysis. Subsequently, we present the attractive optimization of the achievable sum SE with respect to the phase shifts and the total transmit power that can be performed every several coherence intervals due to the slow variation of the large-scale statistics. Numerical results validate the analytical expressions and demonstrate the performance while allowing the extraction of insightful design conclusions for common scenarios including UE mobility. In particular, channel aging degrades the performance but its impact can be controlled by choosing appropriately the frame duration or by increasing the number of RIS elements.
[ { "created": "Fri, 9 Sep 2022 08:10:23 GMT", "version": "v1" } ]
2022-09-12
[ [ "Papazafeiropoulos", "Anastasios", "" ], [ "Krikidis", "Ioannis", "" ], [ "Kourtessis", "Pandelis", "" ] ]
The incorporation of reconfigurable intelligent surface (RIS) into massive multiple-input-multiple-output (mMIMO) systems can unleash the potential of next-generation networks by improving the performance of user equipments (UEs) in service dead zones. However, their requirement for accurate channel state information (CSI) is critical, and especially, applications with UE mobility that induce channel aging make challenging the achievement of adequate quality of service. Hence, in this work, we investigate the impact of channel aging on the performance of RIS-assisted mMIMO systems under both spatial correlation and imperfect CSI conditions. Specifically, by accounting for channel aging during both uplink training and downlink data transmission phases, we first perform minimum mean square error (MMSE) channel estimation to obtain the UE effective channels with low overhead similar to conventional systems without RIS. Next, we derive the downlink achievable sum spectral efficiency (SE) with regularized zero-forcing (RZF) precoding in closed-form being dependent only on large-scale statistics by using the deterministic equivalent (DE) analysis. Subsequently, we present the attractive optimization of the achievable sum SE with respect to the phase shifts and the total transmit power that can be performed every several coherence intervals due to the slow variation of the large-scale statistics. Numerical results validate the analytical expressions and demonstrate the performance while allowing the extraction of insightful design conclusions for common scenarios including UE mobility. In particular, channel aging degrades the performance but its impact can be controlled by choosing appropriately the frame duration or by increasing the number of RIS elements.
2211.07875
Ye Tao
Ye Tao, Yuze Jiang, Pengfei Lin, Manabu Tsukada and Hiroshi Esaki
zk-PoT: Zero-Knowledge Proof of Traffic for Privacy Enabled Cooperative Perception
IEEE Consumer Communications & Networking Conference (CCNC) 2023
null
null
null
cs.NI cs.CR
http://creativecommons.org/licenses/by-nc-nd/4.0/
Cooperative perception is an essential and widely discussed application of connected automated vehicles. However, the authenticity of perception data is not ensured, because the vehicles cannot independently verify the event they did not see. Many methods, including trust-based (i.e., statistical) approaches and plausibility-based methods, have been proposed to determine data authenticity. However, these methods cannot verify data without a priori knowledge. In this study, a novel approach of constructing a self-proving data from the number plate of target vehicles was proposed. By regarding the pseudonym and number plate as a shared secret and letting multiple vehicles prove they know it independently, the data authenticity problem can be transformed to a cryptography problem that can be solved without trust or plausibility evaluations. Our work can be adapted to the existing works including ETSI/ISO ITS standards while maintaining backward compatibility. Analyses of common attacks and attacks specific to the proposed method reveal that most attacks can be prevented, whereas preventing some other attacks, such as collusion attacks, can be mitigated. Experiments based on realistic data set show that the rate of successful verification can achieve 70\% to 80\% at rush hours.
[ { "created": "Tue, 15 Nov 2022 03:50:08 GMT", "version": "v1" } ]
2022-11-16
[ [ "Tao", "Ye", "" ], [ "Jiang", "Yuze", "" ], [ "Lin", "Pengfei", "" ], [ "Tsukada", "Manabu", "" ], [ "Esaki", "Hiroshi", "" ] ]
Cooperative perception is an essential and widely discussed application of connected automated vehicles. However, the authenticity of perception data is not ensured, because the vehicles cannot independently verify the event they did not see. Many methods, including trust-based (i.e., statistical) approaches and plausibility-based methods, have been proposed to determine data authenticity. However, these methods cannot verify data without a priori knowledge. In this study, a novel approach of constructing a self-proving data from the number plate of target vehicles was proposed. By regarding the pseudonym and number plate as a shared secret and letting multiple vehicles prove they know it independently, the data authenticity problem can be transformed to a cryptography problem that can be solved without trust or plausibility evaluations. Our work can be adapted to the existing works including ETSI/ISO ITS standards while maintaining backward compatibility. Analyses of common attacks and attacks specific to the proposed method reveal that most attacks can be prevented, whereas preventing some other attacks, such as collusion attacks, can be mitigated. Experiments based on realistic data set show that the rate of successful verification can achieve 70\% to 80\% at rush hours.
2105.07855
A Mallikarjuna Reddy dr
Swarajya lakshmi v papineni, A.Mallikarjuna Reddy, Sudeepti yarlagadda, Snigdha Yarlagadda, Haritha Akkinen
An Extensive Analytical Approach on Human Resources using Random Forest Algorithm
null
null
10.14445/22315381/IJETT-V69I5P217
null
cs.CY cs.AI
http://creativecommons.org/licenses/by/4.0/
The current job survey shows that most software employees are planning to change their job role due to high pay for recent jobs such as data scientists, business analysts and artificial intelligence fields. The survey also indicated that work life imbalances, low pay, uneven shifts and many other factors also make employees think about changing their work life. In this paper, for an efficient organisation of the company in terms of human resources, the proposed system designed a model with the help of a random forest algorithm by considering different employee parameters. This helps the HR department retain the employee by identifying gaps and helping the organisation to run smoothly with a good employee retention ratio. This combination of HR and data science can help the productivity, collaboration and well-being of employees of the organisation. It also helps to develop strategies that have an impact on the performance of employees in terms of external and social factors.
[ { "created": "Fri, 7 May 2021 07:35:23 GMT", "version": "v1" } ]
2021-05-18
[ [ "papineni", "Swarajya lakshmi v", "" ], [ "Reddy", "A. Mallikarjuna", "" ], [ "yarlagadda", "Sudeepti", "" ], [ "Yarlagadda", "Snigdha", "" ], [ "Akkinen", "Haritha", "" ] ]
The current job survey shows that most software employees are planning to change their job role due to high pay for recent jobs such as data scientists, business analysts and artificial intelligence fields. The survey also indicated that work life imbalances, low pay, uneven shifts and many other factors also make employees think about changing their work life. In this paper, for an efficient organisation of the company in terms of human resources, the proposed system designed a model with the help of a random forest algorithm by considering different employee parameters. This helps the HR department retain the employee by identifying gaps and helping the organisation to run smoothly with a good employee retention ratio. This combination of HR and data science can help the productivity, collaboration and well-being of employees of the organisation. It also helps to develop strategies that have an impact on the performance of employees in terms of external and social factors.
1811.11700
Richard Spence
Faryad Darabi Sahneh, Alon Efrat, Stephen Kobourov, Spencer Krieger, Richard Spence
Approximation algorithms for the vertex-weighted grade-of-service Steiner tree problem
null
null
null
null
cs.DS
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Given a graph $G = (V,E)$ and a subset $T \subseteq V$ of terminals, a \emph{Steiner tree} of $G$ is a tree that spans $T$. In the vertex-weighted Steiner tree (VST) problem, each vertex is assigned a non-negative weight, and the goal is to compute a minimum weight Steiner tree of $G$. We study a natural generalization of the VST problem motivated by multi-level graph construction, the \emph{vertex-weighted grade-of-service Steiner tree problem} (V-GSST), which can be stated as follows: given a graph $G$ and terminals $T$, where each terminal $v \in T$ requires a facility of a minimum grade of service $R(v)\in \{1,2,\ldots\ell\}$, compute a Steiner tree $G'$ by installing facilities on a subset of vertices, such that any two vertices requiring a certain grade of service are connected by a path in $G'$ with the minimum grade of service or better. Facilities of higher grade are more costly than facilities of lower grade. Multi-level variants such as this one can be useful in network design problems where vertices may require facilities of varying priority. While similar problems have been studied in the edge-weighted case, they have not been studied as well in the more general vertex-weighted case. We first describe a simple heuristic for the V-GSST problem whose approximation ratio depends on $\ell$, the number of grades of service. We then generalize the greedy algorithm of [Klein \& Ravi, 1995] to show that the V-GSST problem admits a $(2 \ln |T|)$-approximation, where $T$ is the set of terminals requiring some facility. This result is surprising, as it shows that the (seemingly harder) multi-grade problem can be approximated as well as the VST problem, and that the approximation ratio does not depend on the number of grades of service.
[ { "created": "Wed, 28 Nov 2018 17:37:13 GMT", "version": "v1" }, { "created": "Fri, 3 May 2019 23:02:41 GMT", "version": "v2" } ]
2019-05-07
[ [ "Sahneh", "Faryad Darabi", "" ], [ "Efrat", "Alon", "" ], [ "Kobourov", "Stephen", "" ], [ "Krieger", "Spencer", "" ], [ "Spence", "Richard", "" ] ]
Given a graph $G = (V,E)$ and a subset $T \subseteq V$ of terminals, a \emph{Steiner tree} of $G$ is a tree that spans $T$. In the vertex-weighted Steiner tree (VST) problem, each vertex is assigned a non-negative weight, and the goal is to compute a minimum weight Steiner tree of $G$. We study a natural generalization of the VST problem motivated by multi-level graph construction, the \emph{vertex-weighted grade-of-service Steiner tree problem} (V-GSST), which can be stated as follows: given a graph $G$ and terminals $T$, where each terminal $v \in T$ requires a facility of a minimum grade of service $R(v)\in \{1,2,\ldots\ell\}$, compute a Steiner tree $G'$ by installing facilities on a subset of vertices, such that any two vertices requiring a certain grade of service are connected by a path in $G'$ with the minimum grade of service or better. Facilities of higher grade are more costly than facilities of lower grade. Multi-level variants such as this one can be useful in network design problems where vertices may require facilities of varying priority. While similar problems have been studied in the edge-weighted case, they have not been studied as well in the more general vertex-weighted case. We first describe a simple heuristic for the V-GSST problem whose approximation ratio depends on $\ell$, the number of grades of service. We then generalize the greedy algorithm of [Klein \& Ravi, 1995] to show that the V-GSST problem admits a $(2 \ln |T|)$-approximation, where $T$ is the set of terminals requiring some facility. This result is surprising, as it shows that the (seemingly harder) multi-grade problem can be approximated as well as the VST problem, and that the approximation ratio does not depend on the number of grades of service.
2011.13118
Xiaoxiao Long
Xiaoxiao Long, Lingjie Liu, Wei Li, Christian Theobalt, Wenping Wang
Multi-view Depth Estimation using Epipolar Spatio-Temporal Networks
null
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We present a novel method for multi-view depth estimation from a single video, which is a critical task in various applications, such as perception, reconstruction and robot navigation. Although previous learning-based methods have demonstrated compelling results, most works estimate depth maps of individual video frames independently, without taking into consideration the strong geometric and temporal coherence among the frames. Moreover, current state-of-the-art (SOTA) models mostly adopt a fully 3D convolution network for cost regularization and therefore require high computational cost, thus limiting their deployment in real-world applications. Our method achieves temporally coherent depth estimation results by using a novel Epipolar Spatio-Temporal (EST) transformer to explicitly associate geometric and temporal correlation with multiple estimated depth maps. Furthermore, to reduce the computational cost, inspired by recent Mixture-of-Experts models, we design a compact hybrid network consisting of a 2D context-aware network and a 3D matching network which learn 2D context information and 3D disparity cues separately. Extensive experiments demonstrate that our method achieves higher accuracy in depth estimation and significant speedup than the SOTA methods.
[ { "created": "Thu, 26 Nov 2020 04:04:21 GMT", "version": "v1" }, { "created": "Tue, 1 Dec 2020 02:55:11 GMT", "version": "v2" }, { "created": "Mon, 12 Jul 2021 16:02:54 GMT", "version": "v3" } ]
2021-07-13
[ [ "Long", "Xiaoxiao", "" ], [ "Liu", "Lingjie", "" ], [ "Li", "Wei", "" ], [ "Theobalt", "Christian", "" ], [ "Wang", "Wenping", "" ] ]
We present a novel method for multi-view depth estimation from a single video, which is a critical task in various applications, such as perception, reconstruction and robot navigation. Although previous learning-based methods have demonstrated compelling results, most works estimate depth maps of individual video frames independently, without taking into consideration the strong geometric and temporal coherence among the frames. Moreover, current state-of-the-art (SOTA) models mostly adopt a fully 3D convolution network for cost regularization and therefore require high computational cost, thus limiting their deployment in real-world applications. Our method achieves temporally coherent depth estimation results by using a novel Epipolar Spatio-Temporal (EST) transformer to explicitly associate geometric and temporal correlation with multiple estimated depth maps. Furthermore, to reduce the computational cost, inspired by recent Mixture-of-Experts models, we design a compact hybrid network consisting of a 2D context-aware network and a 3D matching network which learn 2D context information and 3D disparity cues separately. Extensive experiments demonstrate that our method achieves higher accuracy in depth estimation and significant speedup than the SOTA methods.
2006.02471
Julio C. S. Reis
Julio C. S. Reis, Philipe de Freitas Melo, Kiran Garimella, Fabr\'icio Benevenuto
Can WhatsApp Benefit from Debunked Fact-Checked Stories to Reduce Misinformation?
This is a preprint version of an accepted manuscript on The Harvard Kennedy School (HKS) Misinformation Review. Please, consider to cite it instead of this one
null
null
null
cs.CY cs.CR
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
WhatsApp was alleged to be widely used to spread misinformation and propaganda during elections in Brazil and India. Due to the private encrypted nature of the messages on WhatsApp, it is hard to track the dissemination of misinformation at scale. In this work, using public WhatsApp data, we observe that misinformation has been largely shared on WhatsApp public groups even after they were already fact-checked by popular fact-checking agencies. This represents a significant portion of misinformation spread in both Brazil and India in the groups analyzed. We posit that such misinformation content could be prevented if WhatsApp had a means to flag already fact-checked content. To this end, we propose an architecture that could be implemented by WhatsApp to counter such misinformation. Our proposal respects the current end-to-end encryption architecture on WhatsApp, thus protecting users' privacy while providing an approach to detect the misinformation that benefits from fact-checking efforts.
[ { "created": "Wed, 3 Jun 2020 18:28:57 GMT", "version": "v1" }, { "created": "Thu, 6 Aug 2020 03:11:38 GMT", "version": "v2" } ]
2020-08-07
[ [ "Reis", "Julio C. S.", "" ], [ "Melo", "Philipe de Freitas", "" ], [ "Garimella", "Kiran", "" ], [ "Benevenuto", "Fabrício", "" ] ]
WhatsApp was alleged to be widely used to spread misinformation and propaganda during elections in Brazil and India. Due to the private encrypted nature of the messages on WhatsApp, it is hard to track the dissemination of misinformation at scale. In this work, using public WhatsApp data, we observe that misinformation has been largely shared on WhatsApp public groups even after they were already fact-checked by popular fact-checking agencies. This represents a significant portion of misinformation spread in both Brazil and India in the groups analyzed. We posit that such misinformation content could be prevented if WhatsApp had a means to flag already fact-checked content. To this end, we propose an architecture that could be implemented by WhatsApp to counter such misinformation. Our proposal respects the current end-to-end encryption architecture on WhatsApp, thus protecting users' privacy while providing an approach to detect the misinformation that benefits from fact-checking efforts.
1711.10639
EPTCS
Hadi Ravanbakhsh (1), Sriram Sankaranarayanan (1) ((1) University of Colorado, Boulder)
A Class of Control Certificates to Ensure Reach-While-Stay for Switched Systems
In Proceedings SYNT 2017, arXiv:1711.10224
EPTCS 260, 2017, pp. 44-61
10.4204/EPTCS.260.6
null
cs.SY cs.LO
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In this article, we consider the problem of synthesizing switching controllers for temporal properties through the composition of simple primitive reach-while-stay (RWS) properties. Reach-while-stay properties specify that the system states starting from an initial set I, must reach a goal (target) set G in finite time, while remaining inside a safe set S. Our approach synthesizes switched controllers that select between finitely many modes to satisfy the given RWS specification. To do so, we consider control certificates, which are Lyapunov-like functions that represent control strategies to achieve the desired specification. However, for RWS problems, a control Lyapunov-like function is often hard to synthesize in a simple polynomial form. Therefore, we combine control barrier and Lyapunov functions with an additional compatibility condition between them. Using this approach, the controller synthesis problem reduces to one of solving quantified nonlinear constrained problems that are handled using a combination of SMT solvers. The synthesis of controllers is demonstrated through a set of interesting numerical examples drawn from the related work, and compared with the state-of-the-art tool SCOTS. Our evaluation suggests that our approach is computationally feasible, and adds to the growing body of formal approaches to controller synthesis.
[ { "created": "Wed, 29 Nov 2017 01:26:09 GMT", "version": "v1" } ]
2017-11-30
[ [ "Ravanbakhsh", "Hadi", "" ], [ "Sankaranarayanan", "Sriram", "" ] ]
In this article, we consider the problem of synthesizing switching controllers for temporal properties through the composition of simple primitive reach-while-stay (RWS) properties. Reach-while-stay properties specify that the system states starting from an initial set I, must reach a goal (target) set G in finite time, while remaining inside a safe set S. Our approach synthesizes switched controllers that select between finitely many modes to satisfy the given RWS specification. To do so, we consider control certificates, which are Lyapunov-like functions that represent control strategies to achieve the desired specification. However, for RWS problems, a control Lyapunov-like function is often hard to synthesize in a simple polynomial form. Therefore, we combine control barrier and Lyapunov functions with an additional compatibility condition between them. Using this approach, the controller synthesis problem reduces to one of solving quantified nonlinear constrained problems that are handled using a combination of SMT solvers. The synthesis of controllers is demonstrated through a set of interesting numerical examples drawn from the related work, and compared with the state-of-the-art tool SCOTS. Our evaluation suggests that our approach is computationally feasible, and adds to the growing body of formal approaches to controller synthesis.
2108.13892
Liesbeth Allein
Liesbeth Allein, Marie-Francine Moens and Domenico Perrotta
Like Article, Like Audience: Enforcing Multimodal Correlations for Disinformation Detection
null
null
null
null
cs.CL cs.IR
http://creativecommons.org/licenses/by/4.0/
User-generated content (e.g., tweets and profile descriptions) and shared content between users (e.g., news articles) reflect a user's online identity. This paper investigates whether correlations between user-generated and user-shared content can be leveraged for detecting disinformation in online news articles. We develop a multimodal learning algorithm for disinformation detection. The latent representations of news articles and user-generated content allow that during training the model is guided by the profile of users who prefer content similar to the news article that is evaluated, and this effect is reinforced if that content is shared among different users. By only leveraging user information during model optimization, the model does not rely on user profiling when predicting an article's veracity. The algorithm is successfully applied to three widely used neural classifiers, and results are obtained on different datasets. Visualization techniques show that the proposed model learns feature representations of unseen news articles that better discriminate between fake and real news texts.
[ { "created": "Tue, 31 Aug 2021 14:50:16 GMT", "version": "v1" } ]
2021-09-01
[ [ "Allein", "Liesbeth", "" ], [ "Moens", "Marie-Francine", "" ], [ "Perrotta", "Domenico", "" ] ]
User-generated content (e.g., tweets and profile descriptions) and shared content between users (e.g., news articles) reflect a user's online identity. This paper investigates whether correlations between user-generated and user-shared content can be leveraged for detecting disinformation in online news articles. We develop a multimodal learning algorithm for disinformation detection. The latent representations of news articles and user-generated content allow that during training the model is guided by the profile of users who prefer content similar to the news article that is evaluated, and this effect is reinforced if that content is shared among different users. By only leveraging user information during model optimization, the model does not rely on user profiling when predicting an article's veracity. The algorithm is successfully applied to three widely used neural classifiers, and results are obtained on different datasets. Visualization techniques show that the proposed model learns feature representations of unseen news articles that better discriminate between fake and real news texts.
2009.12724
Shixian Wen
Shixian Wen, Amanda Rios, Laurent Itti
Beneficial Perturbations Network for Defending Adversarial Examples
The paper is under consideration at Pattern Recognition Letters
null
null
null
cs.LG cs.CR stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Deep neural networks can be fooled by adversarial attacks: adding carefully computed small adversarial perturbations to clean inputs can cause misclassification on state-of-the-art machine learning models. The reason is that neural networks fail to accommodate the distribution drift of the input data caused by adversarial perturbations. Here, we present a new solution - Beneficial Perturbation Network (BPN) - to defend against adversarial attacks by fixing the distribution drift. During training, BPN generates and leverages beneficial perturbations (somewhat opposite to well-known adversarial perturbations) by adding new, out-of-network biasing units. Biasing units influence the parameter space of the network, to preempt and neutralize future adversarial perturbations on input data samples. To achieve this, BPN creates reverse adversarial attacks during training, with very little cost, by recycling the training gradients already computed. Reverse attacks are captured by the biasing units, and the biases can in turn effectively defend against future adversarial examples. Reverse attacks are a shortcut, i.e., they affect the network's parameters without requiring instantiation of adversarial examples that could assist training. We provide comprehensive empirical evidence showing that 1) BPN is robust to adversarial examples and is much more running memory and computationally efficient compared to classical adversarial training. 2) BPN can defend against adversarial examples with negligible additional computation and parameter costs compared to training only on clean examples; 3) BPN hurts the accuracy on clean examples much less than classic adversarial training; 4) BPN can improve the generalization of the network 5) BPN trained only with Fast Gradient Sign Attack can generalize to defend PGD attacks.
[ { "created": "Sun, 27 Sep 2020 02:05:26 GMT", "version": "v1" }, { "created": "Wed, 17 Mar 2021 07:25:51 GMT", "version": "v2" }, { "created": "Mon, 13 Sep 2021 13:05:55 GMT", "version": "v3" } ]
2021-09-14
[ [ "Wen", "Shixian", "" ], [ "Rios", "Amanda", "" ], [ "Itti", "Laurent", "" ] ]
Deep neural networks can be fooled by adversarial attacks: adding carefully computed small adversarial perturbations to clean inputs can cause misclassification on state-of-the-art machine learning models. The reason is that neural networks fail to accommodate the distribution drift of the input data caused by adversarial perturbations. Here, we present a new solution - Beneficial Perturbation Network (BPN) - to defend against adversarial attacks by fixing the distribution drift. During training, BPN generates and leverages beneficial perturbations (somewhat opposite to well-known adversarial perturbations) by adding new, out-of-network biasing units. Biasing units influence the parameter space of the network, to preempt and neutralize future adversarial perturbations on input data samples. To achieve this, BPN creates reverse adversarial attacks during training, with very little cost, by recycling the training gradients already computed. Reverse attacks are captured by the biasing units, and the biases can in turn effectively defend against future adversarial examples. Reverse attacks are a shortcut, i.e., they affect the network's parameters without requiring instantiation of adversarial examples that could assist training. We provide comprehensive empirical evidence showing that 1) BPN is robust to adversarial examples and is much more running memory and computationally efficient compared to classical adversarial training. 2) BPN can defend against adversarial examples with negligible additional computation and parameter costs compared to training only on clean examples; 3) BPN hurts the accuracy on clean examples much less than classic adversarial training; 4) BPN can improve the generalization of the network 5) BPN trained only with Fast Gradient Sign Attack can generalize to defend PGD attacks.
2101.03438
Junde Li
Junde Li, Rasit Topaloglu, Swaroop Ghosh
Quantum Generative Models for Small Molecule Drug Discovery
null
null
null
null
cs.ET cs.LG quant-ph
http://creativecommons.org/licenses/by/4.0/
Existing drug discovery pipelines take 5-10 years and cost billions of dollars. Computational approaches aim to sample from regions of the whole molecular and solid-state compounds called chemical space which could be on the order of 1060 . Deep generative models can model the underlying probability distribution of both the physical structures and property of drugs and relate them nonlinearly. By exploiting patterns in massive datasets, these models can distill salient features that characterize the molecules. Generative Adversarial Networks (GANs) discover drug candidates by generating molecular structures that obey chemical and physical properties and show affinity towards binding with the receptor for a target disease. However, classical GANs cannot explore certain regions of the chemical space and suffer from curse-of-dimensionality. A full quantum GAN may require more than 90 qubits even to generate QM9-like small molecules. We propose a qubit-efficient quantum GAN with a hybrid generator (QGAN-HG) to learn richer representation of molecules via searching exponentially large chemical space with few qubits more efficiently than classical GAN. The QGANHG model is composed of a hybrid quantum generator that supports various number of qubits and quantum circuit layers, and, a classical discriminator. QGAN-HG with only 14.93% retained parameters can learn molecular distribution as efficiently as classical counterpart. The QGAN-HG variation with patched circuits considerably accelerates our standard QGANHG training process and avoids potential gradient vanishing issue of deep neural networks. Code is available on GitHub https://github.com/jundeli/quantum-gan.
[ { "created": "Sat, 9 Jan 2021 22:33:16 GMT", "version": "v1" } ]
2021-01-12
[ [ "Li", "Junde", "" ], [ "Topaloglu", "Rasit", "" ], [ "Ghosh", "Swaroop", "" ] ]
Existing drug discovery pipelines take 5-10 years and cost billions of dollars. Computational approaches aim to sample from regions of the whole molecular and solid-state compounds called chemical space which could be on the order of 1060 . Deep generative models can model the underlying probability distribution of both the physical structures and property of drugs and relate them nonlinearly. By exploiting patterns in massive datasets, these models can distill salient features that characterize the molecules. Generative Adversarial Networks (GANs) discover drug candidates by generating molecular structures that obey chemical and physical properties and show affinity towards binding with the receptor for a target disease. However, classical GANs cannot explore certain regions of the chemical space and suffer from curse-of-dimensionality. A full quantum GAN may require more than 90 qubits even to generate QM9-like small molecules. We propose a qubit-efficient quantum GAN with a hybrid generator (QGAN-HG) to learn richer representation of molecules via searching exponentially large chemical space with few qubits more efficiently than classical GAN. The QGANHG model is composed of a hybrid quantum generator that supports various number of qubits and quantum circuit layers, and, a classical discriminator. QGAN-HG with only 14.93% retained parameters can learn molecular distribution as efficiently as classical counterpart. The QGAN-HG variation with patched circuits considerably accelerates our standard QGANHG training process and avoids potential gradient vanishing issue of deep neural networks. Code is available on GitHub https://github.com/jundeli/quantum-gan.
1602.01168
Zhuolin Jiang
Zhuolin Jiang, Yaming Wang, Larry Davis, Walt Andrews, Viktor Rozgic
Learning Discriminative Features via Label Consistent Neural Network
null
null
null
null
cs.CV cs.LG cs.MM cs.NE stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Deep Convolutional Neural Networks (CNN) enforces supervised information only at the output layer, and hidden layers are trained by back propagating the prediction error from the output layer without explicit supervision. We propose a supervised feature learning approach, Label Consistent Neural Network, which enforces direct supervision in late hidden layers. We associate each neuron in a hidden layer with a particular class label and encourage it to be activated for input signals from the same class. More specifically, we introduce a label consistency regularization called "discriminative representation error" loss for late hidden layers and combine it with classification error loss to build our overall objective function. This label consistency constraint alleviates the common problem of gradient vanishing and tends to faster convergence; it also makes the features derived from late hidden layers discriminative enough for classification even using a simple $k$-NN classifier, since input signals from the same class will have very similar representations. Experimental results demonstrate that our approach achieves state-of-the-art performances on several public benchmarks for action and object category recognition.
[ { "created": "Wed, 3 Feb 2016 02:41:33 GMT", "version": "v1" }, { "created": "Sun, 5 Jun 2016 02:45:35 GMT", "version": "v2" } ]
2016-06-07
[ [ "Jiang", "Zhuolin", "" ], [ "Wang", "Yaming", "" ], [ "Davis", "Larry", "" ], [ "Andrews", "Walt", "" ], [ "Rozgic", "Viktor", "" ] ]
Deep Convolutional Neural Networks (CNN) enforces supervised information only at the output layer, and hidden layers are trained by back propagating the prediction error from the output layer without explicit supervision. We propose a supervised feature learning approach, Label Consistent Neural Network, which enforces direct supervision in late hidden layers. We associate each neuron in a hidden layer with a particular class label and encourage it to be activated for input signals from the same class. More specifically, we introduce a label consistency regularization called "discriminative representation error" loss for late hidden layers and combine it with classification error loss to build our overall objective function. This label consistency constraint alleviates the common problem of gradient vanishing and tends to faster convergence; it also makes the features derived from late hidden layers discriminative enough for classification even using a simple $k$-NN classifier, since input signals from the same class will have very similar representations. Experimental results demonstrate that our approach achieves state-of-the-art performances on several public benchmarks for action and object category recognition.
1403.0338
Sanjaya Kumar Panda
Jitendra Kumar Rout, Sourav Kumar Bhoi, Sanjaya Kumar Panda
SFTP : A Secure and Fault-Tolerant Paradigm against Blackhole Attack in MANET
6 pages, 9 figures
International Journal of Computer Applications 2013
10.5120/10623-5343
pxc3885343
cs.NI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Security issues in MANET are a challenging task nowadays. MANETs are vulnerable to passive attacks and active attacks because of a limited number of resources and lack of centralized authority. Blackhole attack is an attack in network layer which degrade the network performance by dropping the packets. In this paper, we have proposed a Secure Fault-Tolerant Paradigm (SFTP) which checks the Blackhole attack in the network. The three phases used in SFTP algorithm are designing of coverage area to find the area of coverage, Network Connection algorithm to design a fault-tolerant model and Route Discovery algorithm to discover the route and data delivery from source to destination. SFTP gives better network performance by making the network fault free.
[ { "created": "Mon, 3 Mar 2014 08:36:22 GMT", "version": "v1" } ]
2014-03-04
[ [ "Rout", "Jitendra Kumar", "" ], [ "Bhoi", "Sourav Kumar", "" ], [ "Panda", "Sanjaya Kumar", "" ] ]
Security issues in MANET are a challenging task nowadays. MANETs are vulnerable to passive attacks and active attacks because of a limited number of resources and lack of centralized authority. Blackhole attack is an attack in network layer which degrade the network performance by dropping the packets. In this paper, we have proposed a Secure Fault-Tolerant Paradigm (SFTP) which checks the Blackhole attack in the network. The three phases used in SFTP algorithm are designing of coverage area to find the area of coverage, Network Connection algorithm to design a fault-tolerant model and Route Discovery algorithm to discover the route and data delivery from source to destination. SFTP gives better network performance by making the network fault free.
2104.00538
Timur \.Inan
Inan Timur, Baba Ahmet Fevzi
Prediction of Wind Speed Using Artificial Neural Networks and ANFIS Methods (Observation Buoy Example)
5 pages, in Turkish language
null
null
null
cs.LG cs.SY eess.SY
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Estimation of the wind speed plays an important role in many issues such as route determination of ships, efficient use of wind roses, and correct planning of agricultural activities. In this study, wind velocity estimation is calculated using artificial neural networks (ANN) and adaptive artificial neural fuzzy inference system (ANFIS) methods. The data required for estimation was obtained from the float named E1M3A, which is a float inside the POSEIDON float system. The proposed ANN is a Nonlinear Auto Regressive with External Input (NARX) type of artificial neural network with 3 layers, 50 neurons, 6 inputs and 1 output. The ANFIS system introduced is a fuzzy inference system with 6 inputs, 1 output, and 3 membership functions (MF) per input. The proposed systems were trained to make wind speed estimates after 3 hours and the data obtained were obtained and the successes of the systems were revealed by comparing the obtained values with real measurements. Mean Squarred Error (MSE) and the regression between the predictions and expected values (R) were used to evaluate the success of the estimation values obtained from the systems. According to estimation results, ANN achieved 2.19 MSE and 0.897 R values in training, 2.88 MSE and 0.866 R values in validation, and 2.93 MSE and 0.857 R values in testing. ANFIS method has obtained 0.31634 MSE and 0.99 R values
[ { "created": "Mon, 29 Mar 2021 19:01:43 GMT", "version": "v1" } ]
2021-04-02
[ [ "Timur", "Inan", "" ], [ "Fevzi", "Baba Ahmet", "" ] ]
Estimation of the wind speed plays an important role in many issues such as route determination of ships, efficient use of wind roses, and correct planning of agricultural activities. In this study, wind velocity estimation is calculated using artificial neural networks (ANN) and adaptive artificial neural fuzzy inference system (ANFIS) methods. The data required for estimation was obtained from the float named E1M3A, which is a float inside the POSEIDON float system. The proposed ANN is a Nonlinear Auto Regressive with External Input (NARX) type of artificial neural network with 3 layers, 50 neurons, 6 inputs and 1 output. The ANFIS system introduced is a fuzzy inference system with 6 inputs, 1 output, and 3 membership functions (MF) per input. The proposed systems were trained to make wind speed estimates after 3 hours and the data obtained were obtained and the successes of the systems were revealed by comparing the obtained values with real measurements. Mean Squarred Error (MSE) and the regression between the predictions and expected values (R) were used to evaluate the success of the estimation values obtained from the systems. According to estimation results, ANN achieved 2.19 MSE and 0.897 R values in training, 2.88 MSE and 0.866 R values in validation, and 2.93 MSE and 0.857 R values in testing. ANFIS method has obtained 0.31634 MSE and 0.99 R values
2002.06761
Yongming Li
Yongming Li, Yan Lei, Pin Wang, Yuchuan Liu
Hybrid Embedded Deep Stacked Sparse Autoencoder with w_LPPD SVM Ensemble
null
null
null
null
cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Deep learning is a kind of feature learning method with strong nonliear feature transformation and becomes more and more important in many fields of artificial intelligence. Deep autoencoder is one representative method of the deep learning methods, and can effectively extract abstract the information of datasets. However, it does not consider the complementarity between the deep features and original features during deep feature transformation. Besides, it suffers from small sample problem. In order to solve these problems, a novel deep autoencoder - hybrid feature embedded stacked sparse autoencoder(HESSAE) has been proposed in this paper. HFESAE is capable to learn discriminant deep features with the help of embedding original features to filter weak hidden-layer outputs during training. For the issue that class representation ability of abstract information is limited by small sample problem, a feature fusion strategy has been designed aiming to combining abstract information learned by HFESAE with original feature and obtain hybrid features for feature reduction. The strategy is hybrid feature selection strategy based on L1 regularization followed by an support vector machine(SVM) ensemble model, in which weighted local discriminant preservation projection (w_LPPD), is designed and employed on each base classifier. At the end of this paper, several representative public datasets are used to verify the effectiveness of the proposed algorithm. The experimental results demonstrated that, the proposed feature learning method yields superior performance compared to other existing and state of art feature learning algorithms including some representative deep autoencoder methods.
[ { "created": "Mon, 17 Feb 2020 04:06:05 GMT", "version": "v1" } ]
2020-02-18
[ [ "Li", "Yongming", "" ], [ "Lei", "Yan", "" ], [ "Wang", "Pin", "" ], [ "Liu", "Yuchuan", "" ] ]
Deep learning is a kind of feature learning method with strong nonliear feature transformation and becomes more and more important in many fields of artificial intelligence. Deep autoencoder is one representative method of the deep learning methods, and can effectively extract abstract the information of datasets. However, it does not consider the complementarity between the deep features and original features during deep feature transformation. Besides, it suffers from small sample problem. In order to solve these problems, a novel deep autoencoder - hybrid feature embedded stacked sparse autoencoder(HESSAE) has been proposed in this paper. HFESAE is capable to learn discriminant deep features with the help of embedding original features to filter weak hidden-layer outputs during training. For the issue that class representation ability of abstract information is limited by small sample problem, a feature fusion strategy has been designed aiming to combining abstract information learned by HFESAE with original feature and obtain hybrid features for feature reduction. The strategy is hybrid feature selection strategy based on L1 regularization followed by an support vector machine(SVM) ensemble model, in which weighted local discriminant preservation projection (w_LPPD), is designed and employed on each base classifier. At the end of this paper, several representative public datasets are used to verify the effectiveness of the proposed algorithm. The experimental results demonstrated that, the proposed feature learning method yields superior performance compared to other existing and state of art feature learning algorithms including some representative deep autoencoder methods.
2209.14795
Salman Manzoor
Salman Manzoor and Antonios Gouglidis and Matthew Bradbury and Neeraj Suri
ThreatPro: Multi-Layer Threat Analysis in the Cloud
32 pages, 14 figures
null
null
null
cs.CR
http://creativecommons.org/licenses/by/4.0/
Many effective Threat Analysis (TA) techniques exist that focus on analyzing threats to targeted assets (e.g., components, services). These techniques consider static interconnections among the assets. However, in dynamic environments, such as the Cloud, resources can instantiate, migrate across physical hosts, or decommission to provide rapid resource elasticity to the users. It is evident that existing TA techniques cannot address all these requirements. In addition, there is an increasing number of complex multi-layer/multi-asset attacks on Cloud systems, such as the Equifax data breach. Hence, there is a need for threat analysis approaches that are designed to analyze threats in complex, dynamic, and multi-layer Cloud environments. In this paper, we propose ThreatPro that addresses the analysis of multi-layer attacks and supports dynamic interconnections in the Cloud. ThreatPro facilitates threat analysis by developing a technology-agnostic information flow model, which represents the Cloud's functionality through a set of conditional transitions. The model establishes the basis to capture the multi-layer and dynamic interconnections during the life-cycle of a Virtual Machine (VM). Specifically, ThreatPro contributes in (a) enabling the exploration of a threat's behavior and its propagation across the Cloud, and (b) assessing the security of the Cloud by analyzing the impact of multiple threats across various operational layers/assets. Using public information on threats from the National Vulnerability Database (NVD), we validate ThreatPro's capabilities, i.e., (a) identify and trace actual Cloud attacks and (b) speculatively postulate alternate potential attack paths.
[ { "created": "Thu, 29 Sep 2022 14:00:55 GMT", "version": "v1" } ]
2022-09-30
[ [ "Manzoor", "Salman", "" ], [ "Gouglidis", "Antonios", "" ], [ "Bradbury", "Matthew", "" ], [ "Suri", "Neeraj", "" ] ]
Many effective Threat Analysis (TA) techniques exist that focus on analyzing threats to targeted assets (e.g., components, services). These techniques consider static interconnections among the assets. However, in dynamic environments, such as the Cloud, resources can instantiate, migrate across physical hosts, or decommission to provide rapid resource elasticity to the users. It is evident that existing TA techniques cannot address all these requirements. In addition, there is an increasing number of complex multi-layer/multi-asset attacks on Cloud systems, such as the Equifax data breach. Hence, there is a need for threat analysis approaches that are designed to analyze threats in complex, dynamic, and multi-layer Cloud environments. In this paper, we propose ThreatPro that addresses the analysis of multi-layer attacks and supports dynamic interconnections in the Cloud. ThreatPro facilitates threat analysis by developing a technology-agnostic information flow model, which represents the Cloud's functionality through a set of conditional transitions. The model establishes the basis to capture the multi-layer and dynamic interconnections during the life-cycle of a Virtual Machine (VM). Specifically, ThreatPro contributes in (a) enabling the exploration of a threat's behavior and its propagation across the Cloud, and (b) assessing the security of the Cloud by analyzing the impact of multiple threats across various operational layers/assets. Using public information on threats from the National Vulnerability Database (NVD), we validate ThreatPro's capabilities, i.e., (a) identify and trace actual Cloud attacks and (b) speculatively postulate alternate potential attack paths.
2206.05282
Ian Covert
Ian Covert, Chanwoo Kim, Su-In Lee
Learning to Estimate Shapley Values with Vision Transformers
ICLR 2023 camera-ready
null
null
null
cs.CV cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Transformers have become a default architecture in computer vision, but understanding what drives their predictions remains a challenging problem. Current explanation approaches rely on attention values or input gradients, but these provide a limited view of a model's dependencies. Shapley values offer a theoretically sound alternative, but their computational cost makes them impractical for large, high-dimensional models. In this work, we aim to make Shapley values practical for vision transformers (ViTs). To do so, we first leverage an attention masking approach to evaluate ViTs with partial information, and we then develop a procedure to generate Shapley value explanations via a separate, learned explainer model. Our experiments compare Shapley values to many baseline methods (e.g., attention rollout, GradCAM, LRP), and we find that our approach provides more accurate explanations than existing methods for ViTs.
[ { "created": "Fri, 10 Jun 2022 07:09:28 GMT", "version": "v1" }, { "created": "Fri, 30 Sep 2022 08:49:55 GMT", "version": "v2" }, { "created": "Wed, 1 Mar 2023 20:24:58 GMT", "version": "v3" } ]
2023-03-03
[ [ "Covert", "Ian", "" ], [ "Kim", "Chanwoo", "" ], [ "Lee", "Su-In", "" ] ]
Transformers have become a default architecture in computer vision, but understanding what drives their predictions remains a challenging problem. Current explanation approaches rely on attention values or input gradients, but these provide a limited view of a model's dependencies. Shapley values offer a theoretically sound alternative, but their computational cost makes them impractical for large, high-dimensional models. In this work, we aim to make Shapley values practical for vision transformers (ViTs). To do so, we first leverage an attention masking approach to evaluate ViTs with partial information, and we then develop a procedure to generate Shapley value explanations via a separate, learned explainer model. Our experiments compare Shapley values to many baseline methods (e.g., attention rollout, GradCAM, LRP), and we find that our approach provides more accurate explanations than existing methods for ViTs.
2204.06598
Sheng He
Sheng He, Yanfang Feng, P. Ellen Grant, Yangming Ou
Deep Relation Learning for Regression and Its Application to Brain Age Estimation
null
IEEE Transactions on Medical Imaging. 2022
null
null
cs.CV
http://creativecommons.org/licenses/by-nc-nd/4.0/
Most deep learning models for temporal regression directly output the estimation based on single input images, ignoring the relationships between different images. In this paper, we propose deep relation learning for regression, aiming to learn different relations between a pair of input images. Four non-linear relations are considered: "cumulative relation", "relative relation", "maximal relation" and "minimal relation". These four relations are learned simultaneously from one deep neural network which has two parts: feature extraction and relation regression. We use an efficient convolutional neural network to extract deep features from the pair of input images and apply a Transformer for relation learning. The proposed method is evaluated on a merged dataset with 6,049 subjects with ages of 0-97 years using 5-fold cross-validation for the task of brain age estimation. The experimental results have shown that the proposed method achieved a mean absolute error (MAE) of 2.38 years, which is lower than the MAEs of 8 other state-of-the-art algorithms with statistical significance (p$<$0.05) in paired T-test (two-side).
[ { "created": "Wed, 13 Apr 2022 18:40:34 GMT", "version": "v1" } ]
2022-04-15
[ [ "He", "Sheng", "" ], [ "Feng", "Yanfang", "" ], [ "Grant", "P. Ellen", "" ], [ "Ou", "Yangming", "" ] ]
Most deep learning models for temporal regression directly output the estimation based on single input images, ignoring the relationships between different images. In this paper, we propose deep relation learning for regression, aiming to learn different relations between a pair of input images. Four non-linear relations are considered: "cumulative relation", "relative relation", "maximal relation" and "minimal relation". These four relations are learned simultaneously from one deep neural network which has two parts: feature extraction and relation regression. We use an efficient convolutional neural network to extract deep features from the pair of input images and apply a Transformer for relation learning. The proposed method is evaluated on a merged dataset with 6,049 subjects with ages of 0-97 years using 5-fold cross-validation for the task of brain age estimation. The experimental results have shown that the proposed method achieved a mean absolute error (MAE) of 2.38 years, which is lower than the MAEs of 8 other state-of-the-art algorithms with statistical significance (p$<$0.05) in paired T-test (two-side).
1408.3469
Steven Weber
Nan Xie, John MacLaren Walsh, Steven Weber
Properties of an Aloha-like stability region
28 pages, 9 figures. Submitted August 15, 2014, revised September 21, 2015 and August 31, 2016, and accepted November 06, 2016 for publication in IEEE Transactions on Information Theory. Preliminary results presented at ISIT 2010, ITA 2010, and ITA 2011. DOI: 10.1109/TIT.2016.2640302. Copyright transferred to IEEE. This is last version uploaded by the authors prior to IEEE proofing process
null
10.1109/TIT.2016.2640302
null
cs.IT cs.NI math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
A well-known inner bound on the stability region of the finite-user slotted Aloha protocol is the set of all arrival rates for which there exists some choice of the contention probabilities such that the associated worst-case service rate for each user exceeds the user's arrival rate, denoted $\Lambda$. Although testing membership in $\Lambda$ of a given arrival rate can be posed as a convex program, it is nonetheless of interest to understand the properties of this set. In this paper we develop new results of this nature, including $i)$ an equivalence between membership in $\Lambda$ and the existence of a positive root of a given polynomial, $ii)$ a method to construct a vector of contention probabilities to stabilize any stabilizable arrival rate vector, $iii)$ the volume of $\Lambda$, $iv)$ explicit polyhedral, spherical, and ellipsoid inner and outer bounds on $\Lambda$, and $v)$ characterization of the generalized convexity properties of a natural ``excess rate'' function associated with $\Lambda$, including the convexity of the set of contention probabilities that stabilize a given arrival rate vector.
[ { "created": "Fri, 15 Aug 2014 05:28:52 GMT", "version": "v1" }, { "created": "Wed, 4 Jan 2017 20:20:40 GMT", "version": "v2" } ]
2017-01-06
[ [ "Xie", "Nan", "" ], [ "Walsh", "John MacLaren", "" ], [ "Weber", "Steven", "" ] ]
A well-known inner bound on the stability region of the finite-user slotted Aloha protocol is the set of all arrival rates for which there exists some choice of the contention probabilities such that the associated worst-case service rate for each user exceeds the user's arrival rate, denoted $\Lambda$. Although testing membership in $\Lambda$ of a given arrival rate can be posed as a convex program, it is nonetheless of interest to understand the properties of this set. In this paper we develop new results of this nature, including $i)$ an equivalence between membership in $\Lambda$ and the existence of a positive root of a given polynomial, $ii)$ a method to construct a vector of contention probabilities to stabilize any stabilizable arrival rate vector, $iii)$ the volume of $\Lambda$, $iv)$ explicit polyhedral, spherical, and ellipsoid inner and outer bounds on $\Lambda$, and $v)$ characterization of the generalized convexity properties of a natural ``excess rate'' function associated with $\Lambda$, including the convexity of the set of contention probabilities that stabilize a given arrival rate vector.
1305.4367
Raphael Jolly
Rapha\"el Jolly
Parallelizing Stream with Future
null
null
null
null
cs.DC
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Stream is re-interpreted in terms of a Lazy monad. Future is substituted for Lazy in the obtained construct, resulting in possible parallelization of any algorithm expressible as a Stream computation. The principle is tested against two example algorithms. Performance is evaluated, and a way to improve it briefly discussed.
[ { "created": "Sun, 19 May 2013 15:00:14 GMT", "version": "v1" } ]
2013-05-21
[ [ "Jolly", "Raphaël", "" ] ]
Stream is re-interpreted in terms of a Lazy monad. Future is substituted for Lazy in the obtained construct, resulting in possible parallelization of any algorithm expressible as a Stream computation. The principle is tested against two example algorithms. Performance is evaluated, and a way to improve it briefly discussed.
1904.05005
Xun Yang
Xun Yang, Meng Wang, Dacheng Tao
Person Re-identification with Metric Learning using Privileged Information
Accepted for IEEE TIP
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Despite the promising progress made in recent years, person re-identification remains a challenging task due to complex variations in human appearances from different camera views. This paper presents a logistic discriminant metric learning method for this challenging problem. Different with most existing metric learning algorithms, it exploits both original data and auxiliary data during training, which is motivated by the new machine learning paradigm - Learning Using Privileged Information. Such privileged information is a kind of auxiliary knowledge which is only available during training. Our goal is to learn an optimal distance function by constructing a locally adaptive decision rule with the help of privileged information. We jointly learn two distance metrics by minimizing the empirical loss penalizing the difference between the distance in the original space and that in the privileged space. In our setting, the distance in the privileged space functions as a local decision threshold, which guides the decision making in the original space like a teacher. The metric learned from the original space is used to compute the distance between a probe image and a gallery image during testing. In addition, we extend the proposed approach to a multi-view setting which is able to explore the complementation of multiple feature representations. In the multi-view setting, multiple metrics corresponding to different original features are jointly learned, guided by the same privileged information. Besides, an effective iterative optimization scheme is introduced to simultaneously optimize the metrics and the assigned metric weights. Experiment results on several widely-used datasets demonstrate that the proposed approach is superior to global decision threshold based methods and outperforms most state-of-the-art results.
[ { "created": "Wed, 10 Apr 2019 05:01:28 GMT", "version": "v1" } ]
2019-04-11
[ [ "Yang", "Xun", "" ], [ "Wang", "Meng", "" ], [ "Tao", "Dacheng", "" ] ]
Despite the promising progress made in recent years, person re-identification remains a challenging task due to complex variations in human appearances from different camera views. This paper presents a logistic discriminant metric learning method for this challenging problem. Different with most existing metric learning algorithms, it exploits both original data and auxiliary data during training, which is motivated by the new machine learning paradigm - Learning Using Privileged Information. Such privileged information is a kind of auxiliary knowledge which is only available during training. Our goal is to learn an optimal distance function by constructing a locally adaptive decision rule with the help of privileged information. We jointly learn two distance metrics by minimizing the empirical loss penalizing the difference between the distance in the original space and that in the privileged space. In our setting, the distance in the privileged space functions as a local decision threshold, which guides the decision making in the original space like a teacher. The metric learned from the original space is used to compute the distance between a probe image and a gallery image during testing. In addition, we extend the proposed approach to a multi-view setting which is able to explore the complementation of multiple feature representations. In the multi-view setting, multiple metrics corresponding to different original features are jointly learned, guided by the same privileged information. Besides, an effective iterative optimization scheme is introduced to simultaneously optimize the metrics and the assigned metric weights. Experiment results on several widely-used datasets demonstrate that the proposed approach is superior to global decision threshold based methods and outperforms most state-of-the-art results.
1903.09203
Seyyed Ali Hashemi
Seyyed Ali Hashemi, Carlo Condo, Marco Mondelli, Warren J. Gross
Rate-Flexible Fast Polar Decoders
null
null
10.1109/TSP.2019.2944738
null
cs.IT math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Polar codes have gained extensive attention during the past few years and recently they have been selected for the next generation of wireless communications standards (5G). Successive-cancellation-based (SC-based) decoders, such as SC list (SCL) and SC flip (SCF), provide a reasonable error performance for polar codes at the cost of low decoding speed. Fast SC-based decoders, such as Fast-SSC, Fast-SSCL, and Fast-SSCF, identify the special constituent codes in a polar code graph off-line, produce a list of operations, store the list in memory, and feed the list to the decoder to decode the constituent codes in order efficiently, thus increasing the decoding speed. However, the list of operations is dependent on the code rate and as the rate changes, a new list is produced, making fast SC-based decoders not rate-flexible. In this paper, we propose a completely rate-flexible fast SC-based decoder by creating the list of operations directly in hardware, with low implementation complexity. We further propose a hardware architecture implementing the proposed method and show that the area occupation of the rate-flexible fast SC-based decoder in this paper is only $38\%$ of the total area of the memory-based base-line decoder when 5G code rates are supported.
[ { "created": "Thu, 21 Mar 2019 19:06:51 GMT", "version": "v1" } ]
2020-01-08
[ [ "Hashemi", "Seyyed Ali", "" ], [ "Condo", "Carlo", "" ], [ "Mondelli", "Marco", "" ], [ "Gross", "Warren J.", "" ] ]
Polar codes have gained extensive attention during the past few years and recently they have been selected for the next generation of wireless communications standards (5G). Successive-cancellation-based (SC-based) decoders, such as SC list (SCL) and SC flip (SCF), provide a reasonable error performance for polar codes at the cost of low decoding speed. Fast SC-based decoders, such as Fast-SSC, Fast-SSCL, and Fast-SSCF, identify the special constituent codes in a polar code graph off-line, produce a list of operations, store the list in memory, and feed the list to the decoder to decode the constituent codes in order efficiently, thus increasing the decoding speed. However, the list of operations is dependent on the code rate and as the rate changes, a new list is produced, making fast SC-based decoders not rate-flexible. In this paper, we propose a completely rate-flexible fast SC-based decoder by creating the list of operations directly in hardware, with low implementation complexity. We further propose a hardware architecture implementing the proposed method and show that the area occupation of the rate-flexible fast SC-based decoder in this paper is only $38\%$ of the total area of the memory-based base-line decoder when 5G code rates are supported.
1910.12783
Lingzhou Hong
Lingzhou Hong, Alfredo Garcia, and Ceyhun Eksin
Distributed Networked Learning with Correlated Data
36 pages
null
null
null
cs.LG stat.ML
http://creativecommons.org/licenses/by-nc-sa/4.0/
We consider a distributed estimation method in a setting with heterogeneous streams of correlated data distributed across nodes in a network. In the considered approach, linear models are estimated locally (i.e., with only local data) subject to a network regularization term that penalizes a local model that differs from neighboring models. We analyze computation dynamics (associated with stochastic gradient updates) and information exchange (associated with exchanging current models with neighboring nodes). We provide a finite-time characterization of convergence of the weighted ensemble average estimate and compare this result to federated learning, an alternative approach to estimation wherein a single model is updated by locally generated gradient updates. This comparison highlights the trade-off between speed vs precision: while model updates take place at a faster rate in federated learning, the proposed networked approach to estimation enables the identification of models with higher precision. We illustrate the method's general applicability in two examples: estimating a Markov random field using wireless sensor networks and modeling prey escape behavior of flocking birds based on a publicly available dataset.
[ { "created": "Mon, 28 Oct 2019 16:14:02 GMT", "version": "v1" }, { "created": "Tue, 9 Feb 2021 23:38:45 GMT", "version": "v2" } ]
2021-02-11
[ [ "Hong", "Lingzhou", "" ], [ "Garcia", "Alfredo", "" ], [ "Eksin", "Ceyhun", "" ] ]
We consider a distributed estimation method in a setting with heterogeneous streams of correlated data distributed across nodes in a network. In the considered approach, linear models are estimated locally (i.e., with only local data) subject to a network regularization term that penalizes a local model that differs from neighboring models. We analyze computation dynamics (associated with stochastic gradient updates) and information exchange (associated with exchanging current models with neighboring nodes). We provide a finite-time characterization of convergence of the weighted ensemble average estimate and compare this result to federated learning, an alternative approach to estimation wherein a single model is updated by locally generated gradient updates. This comparison highlights the trade-off between speed vs precision: while model updates take place at a faster rate in federated learning, the proposed networked approach to estimation enables the identification of models with higher precision. We illustrate the method's general applicability in two examples: estimating a Markov random field using wireless sensor networks and modeling prey escape behavior of flocking birds based on a publicly available dataset.
1711.01214
Olivier Auber
Olivier Auber
Refounding legitimacy towards Aethogenesis
Proceedings of 18th International Research Conference in The Planetary Collegium's Series 'Art & consciousness in the post-biological era' Shanghai 2015. 9 pages. 4 figures
Technoetic Arts Volume 14 Number 3 December 2016 pp. 235-249(15)
10.1386/tear.14.3.235_1
null
cs.CY
http://creativecommons.org/licenses/by-sa/4.0/
The fusion of humans and technology takes us into an unknown world described by some authors as populated by quasi living species that would relegate us - ordinary humans - to the rank of alienated agents emptied of our identity and consciousness. I argue instead that our world is woven of simple though invisible perspectives which - if we become aware of them - may renew our ability for making judgments and enhance our autonomy. I became aware of these invisible perspectives by observing and practicing a real time collective net art experiment called the Poietic Generator. As the perspectives unveiled by this experiment are invisible I have called them anoptical perspectives i.e. non-optical by analogy with the optical perspective of the Renaissance. Later I have come to realize that these perspectives obtain their cognitive structure from the political origins of our language. Accordingly it is possible to define certain cognitive criteria for assessing the legitimacy of the anoptical perspectives just like some artists and architects of the Renaissance defined the geometrical criteria that established the legitimacy of the optical one.
[ { "created": "Fri, 3 Nov 2017 15:49:03 GMT", "version": "v1" } ]
2017-11-06
[ [ "Auber", "Olivier", "" ] ]
The fusion of humans and technology takes us into an unknown world described by some authors as populated by quasi living species that would relegate us - ordinary humans - to the rank of alienated agents emptied of our identity and consciousness. I argue instead that our world is woven of simple though invisible perspectives which - if we become aware of them - may renew our ability for making judgments and enhance our autonomy. I became aware of these invisible perspectives by observing and practicing a real time collective net art experiment called the Poietic Generator. As the perspectives unveiled by this experiment are invisible I have called them anoptical perspectives i.e. non-optical by analogy with the optical perspective of the Renaissance. Later I have come to realize that these perspectives obtain their cognitive structure from the political origins of our language. Accordingly it is possible to define certain cognitive criteria for assessing the legitimacy of the anoptical perspectives just like some artists and architects of the Renaissance defined the geometrical criteria that established the legitimacy of the optical one.
2404.16380
Zuocheng Wen
Zuocheng Wen and Lingzhong Guo
Efficient Higher-order Convolution for Small Kernels in Deep Learning
null
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Deep convolutional neural networks (DCNNs) are a class of artificial neural networks, primarily for computer vision tasks such as segmentation and classification. Many nonlinear operations, such as activation functions and pooling strategies, are used in DCNNs to enhance their ability to process different signals with different tasks. Conceptional convolution, a linear filter, is the essential component of DCNNs while nonlinear convolution is generally implemented as higher-order Volterra filters, However, for Volterra filtering, significant memory and computational costs pose a primary limitation for its widespread application in DCNN applications. In this study, we propose a novel method to perform higher-order Volterra filtering with lower memory and computation cost in forward and backward pass in DCNN training. The proposed method demonstrates computational advantages compared with conventional Volterra filter implementation. Furthermore, based on the proposed method, a new attention module called Higher-order Local Attention Block (HLA) is proposed and tested on CIFAR-100 dataset, which shows competitive improvement for classification task. Source code is available at: https://github.com/WinterWen666/Efficient-High-Order-Volterra-Convolution.git
[ { "created": "Thu, 25 Apr 2024 07:42:48 GMT", "version": "v1" } ]
2024-04-26
[ [ "Wen", "Zuocheng", "" ], [ "Guo", "Lingzhong", "" ] ]
Deep convolutional neural networks (DCNNs) are a class of artificial neural networks, primarily for computer vision tasks such as segmentation and classification. Many nonlinear operations, such as activation functions and pooling strategies, are used in DCNNs to enhance their ability to process different signals with different tasks. Conceptional convolution, a linear filter, is the essential component of DCNNs while nonlinear convolution is generally implemented as higher-order Volterra filters, However, for Volterra filtering, significant memory and computational costs pose a primary limitation for its widespread application in DCNN applications. In this study, we propose a novel method to perform higher-order Volterra filtering with lower memory and computation cost in forward and backward pass in DCNN training. The proposed method demonstrates computational advantages compared with conventional Volterra filter implementation. Furthermore, based on the proposed method, a new attention module called Higher-order Local Attention Block (HLA) is proposed and tested on CIFAR-100 dataset, which shows competitive improvement for classification task. Source code is available at: https://github.com/WinterWen666/Efficient-High-Order-Volterra-Convolution.git
2009.11840
Martin Kouteck\'y
Martin Kouteck\'y and Johannes Zink
Complexity of Scheduling Few Types of Jobs on Related and Unrelated Machines
null
null
null
null
cs.DS cs.CC math.OC
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The task of scheduling jobs to machines while minimizing the total makespan, the sum of weighted completion times, or a norm of the load vector, are among the oldest and most fundamental tasks in combinatorial optimization. Since all of these problems are in general NP-hard, much attention has been given to the regime where there is only a small number $k$ of job types, but possibly the number of jobs $n$ is large; this is the few job types, high-multiplicity regime. Despite many positive results, the hardness boundary of this regime was not understood until now. We show that makespan minimization on uniformly related machines ($Q|HM|C_{\max}$) is NP-hard already with $6$ job types, and that the related Cutting Stock problem is NP-hard already with $8$ item types. For the more general unrelated machines model ($R|HM|C_{\max}$), we show that if either the largest job size $p_{\max}$, or the number of jobs $n$ are polynomially bounded in the instance size $|I|$, there are algorithms with complexity $|I|^{\textrm{poly}(k)}$. Our main result is that this is unlikely to be improved, because $Q||C_{\max}$ is W[1]-hard parameterized by $k$ already when $n$, $p_{\max}$, and the numbers describing the speeds are polynomial in $|I|$; the same holds for $R|HM|C_{\max}$ (without speeds) when the job sizes matrix has rank $2$. Our positive and negative results also extend to the objectives $\ell_2$-norm minimization of the load vector and, partially, sum of weighted completion times $\sum w_j C_j$. Along the way, we answer affirmatively the question whether makespan minimization on identical machines ($P||C_{\max}$) is fixed-parameter tractable parameterized by $k$, extending our understanding of this fundamental problem. Together with our hardness results for $Q||C_{\max}$ this implies that the complexity of $P|HM|C_{\max}$ is the only remaining open case.
[ { "created": "Thu, 24 Sep 2020 17:38:31 GMT", "version": "v1" } ]
2020-09-25
[ [ "Koutecký", "Martin", "" ], [ "Zink", "Johannes", "" ] ]
The task of scheduling jobs to machines while minimizing the total makespan, the sum of weighted completion times, or a norm of the load vector, are among the oldest and most fundamental tasks in combinatorial optimization. Since all of these problems are in general NP-hard, much attention has been given to the regime where there is only a small number $k$ of job types, but possibly the number of jobs $n$ is large; this is the few job types, high-multiplicity regime. Despite many positive results, the hardness boundary of this regime was not understood until now. We show that makespan minimization on uniformly related machines ($Q|HM|C_{\max}$) is NP-hard already with $6$ job types, and that the related Cutting Stock problem is NP-hard already with $8$ item types. For the more general unrelated machines model ($R|HM|C_{\max}$), we show that if either the largest job size $p_{\max}$, or the number of jobs $n$ are polynomially bounded in the instance size $|I|$, there are algorithms with complexity $|I|^{\textrm{poly}(k)}$. Our main result is that this is unlikely to be improved, because $Q||C_{\max}$ is W[1]-hard parameterized by $k$ already when $n$, $p_{\max}$, and the numbers describing the speeds are polynomial in $|I|$; the same holds for $R|HM|C_{\max}$ (without speeds) when the job sizes matrix has rank $2$. Our positive and negative results also extend to the objectives $\ell_2$-norm minimization of the load vector and, partially, sum of weighted completion times $\sum w_j C_j$. Along the way, we answer affirmatively the question whether makespan minimization on identical machines ($P||C_{\max}$) is fixed-parameter tractable parameterized by $k$, extending our understanding of this fundamental problem. Together with our hardness results for $Q||C_{\max}$ this implies that the complexity of $P|HM|C_{\max}$ is the only remaining open case.
2208.04713
Sreekrishnan Venkateswaran
Sreekrishnan Venkateswaran
Reflections on the Evolution of Computer Science Education
Preprint Edition of the paper published in ACM SIGSOFT Software Engineering Notes (SEN), Volume 47, Issue 3, July 2022 (https://doi.org/10.1145/3539814.3539817)
ACM SIGSOFT Software Engineering Notes (SEN), Volume 47, Issue 3, July2022
10.1145/3539814.3539817
Volume 47, Issue 3, July 2022
cs.CY cs.SE
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Computer Science education has been evolving over the years to reflect applied realities. Until about a decade ago, theory of computation, algorithm design and system software dominated the curricula. Most courses were considered core and were hence mandatory; the programme structure did not allow much of a choice or variety. This column analyses why this changed Circa 2010 when elective subjects across scores of topics become part of mainstream education to reflect the on-going lateral acceleration of Computer Science. Fundamental discoveries in artificial intelligence, machine learning, virtualization and cloud computing are several decades old. Many core theories in data science are centuries old. Yet their leverage exploded only after Circa 2010, when the stage got set for people-centric problem solving in massive scale. This was due in part to the rush of innovative real-world applications that reached the common man through the ubiquitous smart phone. AI/ML modules arrived in popular programming languages; they could be used to build and train models on powerful - yet affordable - compute on public clouds reachable through high-speed Internet connectivity. Academia responded by adapting Computer Science curricula to align it with the changing technology landscape. The goal of this experiential piece is to trigger a lively discussion on the past and future of Computer Science education.
[ { "created": "Sat, 9 Jul 2022 07:07:12 GMT", "version": "v1" } ]
2022-08-10
[ [ "Venkateswaran", "Sreekrishnan", "" ] ]
Computer Science education has been evolving over the years to reflect applied realities. Until about a decade ago, theory of computation, algorithm design and system software dominated the curricula. Most courses were considered core and were hence mandatory; the programme structure did not allow much of a choice or variety. This column analyses why this changed Circa 2010 when elective subjects across scores of topics become part of mainstream education to reflect the on-going lateral acceleration of Computer Science. Fundamental discoveries in artificial intelligence, machine learning, virtualization and cloud computing are several decades old. Many core theories in data science are centuries old. Yet their leverage exploded only after Circa 2010, when the stage got set for people-centric problem solving in massive scale. This was due in part to the rush of innovative real-world applications that reached the common man through the ubiquitous smart phone. AI/ML modules arrived in popular programming languages; they could be used to build and train models on powerful - yet affordable - compute on public clouds reachable through high-speed Internet connectivity. Academia responded by adapting Computer Science curricula to align it with the changing technology landscape. The goal of this experiential piece is to trigger a lively discussion on the past and future of Computer Science education.
2305.01864
Cem Subakan
Zhepei Wang, Cem Subakan, Krishna Subramani, Junkai Wu, Tiago Tavares, Fabio Ayres, Paris Smaragdis
Unsupervised Improvement of Audio-Text Cross-Modal Representations
Accepted to WASPAA 2023
null
null
null
cs.SD cs.LG eess.AS
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Recent advances in using language models to obtain cross-modal audio-text representations have overcome the limitations of conventional training approaches that use predefined labels. This has allowed the community to make progress in tasks like zero-shot classification, which would otherwise not be possible. However, learning such representations requires a large amount of human-annotated audio-text pairs. In this paper, we study unsupervised approaches to improve the learning framework of such representations with unpaired text and audio. We explore domain-unspecific and domain-specific curation methods to create audio-text pairs that we use to further improve the model. We also show that when domain-specific curation is used in conjunction with a soft-labeled contrastive loss, we are able to obtain significant improvement in terms of zero-shot classification performance on downstream sound event classification or acoustic scene classification tasks.
[ { "created": "Wed, 3 May 2023 02:30:46 GMT", "version": "v1" }, { "created": "Fri, 5 May 2023 02:22:49 GMT", "version": "v2" }, { "created": "Mon, 31 Jul 2023 18:28:36 GMT", "version": "v3" } ]
2023-08-02
[ [ "Wang", "Zhepei", "" ], [ "Subakan", "Cem", "" ], [ "Subramani", "Krishna", "" ], [ "Wu", "Junkai", "" ], [ "Tavares", "Tiago", "" ], [ "Ayres", "Fabio", "" ], [ "Smaragdis", "Paris", "" ] ]
Recent advances in using language models to obtain cross-modal audio-text representations have overcome the limitations of conventional training approaches that use predefined labels. This has allowed the community to make progress in tasks like zero-shot classification, which would otherwise not be possible. However, learning such representations requires a large amount of human-annotated audio-text pairs. In this paper, we study unsupervised approaches to improve the learning framework of such representations with unpaired text and audio. We explore domain-unspecific and domain-specific curation methods to create audio-text pairs that we use to further improve the model. We also show that when domain-specific curation is used in conjunction with a soft-labeled contrastive loss, we are able to obtain significant improvement in terms of zero-shot classification performance on downstream sound event classification or acoustic scene classification tasks.
1309.4508
Abdul Razaque
Nyembo Salama, Christian Bach
Introduction of 6th Generation Smart Phone combining the features of both Apple and Android smart phone
10 pages, i figure
null
null
null
cs.HC
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In this paper, we present our novel contribution methodology based on the results of case study that has been implemented in our research environment to test with new technique the usability of both intelligent Android and Apple phones. This analysis of the case study stands for features similar to applications, operating system, hardware and software structure, battery life, and online based websites. Multiple interrogations were applied to collect user answers ongoing features. Users directly react by responding based on their daily used product experience. Consequently, the estimation is based on the data that has been unregistered from the user. The most recent results will end up by introducing a combination of ideal features on both products to build a wonderful extended product in the future.
[ { "created": "Wed, 18 Sep 2013 00:22:34 GMT", "version": "v1" } ]
2013-09-19
[ [ "Salama", "Nyembo", "" ], [ "Bach", "Christian", "" ] ]
In this paper, we present our novel contribution methodology based on the results of case study that has been implemented in our research environment to test with new technique the usability of both intelligent Android and Apple phones. This analysis of the case study stands for features similar to applications, operating system, hardware and software structure, battery life, and online based websites. Multiple interrogations were applied to collect user answers ongoing features. Users directly react by responding based on their daily used product experience. Consequently, the estimation is based on the data that has been unregistered from the user. The most recent results will end up by introducing a combination of ideal features on both products to build a wonderful extended product in the future.
2406.17106
David Mezey
David Mezey, Renaud Bastien, Yating Zheng, Neal McKee, David Stoll, Heiko Hamann, Pawel Romanczuk
Purely vision-based collective movement of robots
null
null
null
null
cs.RO
http://creativecommons.org/licenses/by/4.0/
Collective movement inspired by animal groups promises inherited benefits for robot swarms, such as enhanced sensing and efficiency. However, while animals move in groups using only their local senses, robots often obey central control or use direct communication, introducing systemic weaknesses to the swarm. In the hope of addressing such vulnerabilities, developing bio-inspired decentralized swarms has been a major focus in recent decades. Yet, creating robots that move efficiently together using only local sensory information remains an extraordinary challenge. In this work, we present a decentralized, purely vision-based swarm of terrestrial robots. Within this novel framework robots achieve collisionless, polarized motion exclusively through minimal visual interactions, computing everything on board based on their individual camera streams, making central processing or direct communication obsolete. With agent-based simulations, we further show that using this model, even with a strictly limited field of view and within confined spaces, ordered group motion can emerge, while also highlighting key limitations. Our results offer a multitude of practical applications from hybrid societies coordinating collective movement without any common communication protocol, to advanced, decentralized vision-based robot swarms capable of diverse tasks in ever-changing environments.
[ { "created": "Mon, 24 Jun 2024 19:47:13 GMT", "version": "v1" } ]
2024-06-26
[ [ "Mezey", "David", "" ], [ "Bastien", "Renaud", "" ], [ "Zheng", "Yating", "" ], [ "McKee", "Neal", "" ], [ "Stoll", "David", "" ], [ "Hamann", "Heiko", "" ], [ "Romanczuk", "Pawel", "" ] ]
Collective movement inspired by animal groups promises inherited benefits for robot swarms, such as enhanced sensing and efficiency. However, while animals move in groups using only their local senses, robots often obey central control or use direct communication, introducing systemic weaknesses to the swarm. In the hope of addressing such vulnerabilities, developing bio-inspired decentralized swarms has been a major focus in recent decades. Yet, creating robots that move efficiently together using only local sensory information remains an extraordinary challenge. In this work, we present a decentralized, purely vision-based swarm of terrestrial robots. Within this novel framework robots achieve collisionless, polarized motion exclusively through minimal visual interactions, computing everything on board based on their individual camera streams, making central processing or direct communication obsolete. With agent-based simulations, we further show that using this model, even with a strictly limited field of view and within confined spaces, ordered group motion can emerge, while also highlighting key limitations. Our results offer a multitude of practical applications from hybrid societies coordinating collective movement without any common communication protocol, to advanced, decentralized vision-based robot swarms capable of diverse tasks in ever-changing environments.
2205.15508
Jianheng Tang
Jianheng Tang, Jiajin Li, Ziqi Gao, Jia Li
Rethinking Graph Neural Networks for Anomaly Detection
Accepted by ICML 2022. Our code and data are released at https://github.com/squareRoot3/Rethinking-Anomaly-Detection
null
null
null
cs.LG eess.SP
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Graph Neural Networks (GNNs) are widely applied for graph anomaly detection. As one of the key components for GNN design is to select a tailored spectral filter, we take the first step towards analyzing anomalies via the lens of the graph spectrum. Our crucial observation is the existence of anomalies will lead to the `right-shift' phenomenon, that is, the spectral energy distribution concentrates less on low frequencies and more on high frequencies. This fact motivates us to propose the Beta Wavelet Graph Neural Network (BWGNN). Indeed, BWGNN has spectral and spatial localized band-pass filters to better handle the `right-shift' phenomenon in anomalies. We demonstrate the effectiveness of BWGNN on four large-scale anomaly detection datasets. Our code and data are released at https://github.com/squareRoot3/Rethinking-Anomaly-Detection
[ { "created": "Tue, 31 May 2022 02:39:05 GMT", "version": "v1" } ]
2022-06-01
[ [ "Tang", "Jianheng", "" ], [ "Li", "Jiajin", "" ], [ "Gao", "Ziqi", "" ], [ "Li", "Jia", "" ] ]
Graph Neural Networks (GNNs) are widely applied for graph anomaly detection. As one of the key components for GNN design is to select a tailored spectral filter, we take the first step towards analyzing anomalies via the lens of the graph spectrum. Our crucial observation is the existence of anomalies will lead to the `right-shift' phenomenon, that is, the spectral energy distribution concentrates less on low frequencies and more on high frequencies. This fact motivates us to propose the Beta Wavelet Graph Neural Network (BWGNN). Indeed, BWGNN has spectral and spatial localized band-pass filters to better handle the `right-shift' phenomenon in anomalies. We demonstrate the effectiveness of BWGNN on four large-scale anomaly detection datasets. Our code and data are released at https://github.com/squareRoot3/Rethinking-Anomaly-Detection
1701.03753
Lifeng Wang
Anqi He, Lifeng Wang, Yue Chen, Kai-Kit Wong, and Maged Elkashlan
Spectral and Energy Efficiency of Uplink D2D Underlaid Massive MIMO Cellular Networks
Accepted by IEEE Transactions on Communications
null
null
null
cs.IT math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
One of key 5G scenarios is that device-to-device (D2D) and massive multiple-input multiple-output (MIMO) will be co-existed. However, interference in the uplink D2D underlaid massive MIMO cellular networks needs to be coordinated, due to the vast cellular and D2D transmissions. To this end, this paper introduces a spatially dynamic power control solution for mitigating the cellular-to-D2D and D2D-to-cellular interference. In particular, the proposed D2D power control policy is rather flexible including the special cases of no D2D links or using maximum transmit power. Under the considered power control, an analytical approach is developed to evaluate the spectral efficiency (SE) and energy efficiency (EE) in such networks. Thus, the exact expressions of SE for a cellular user or D2D transmitter are derived, which quantify the impacts of key system parameters such as massive MIMO antennas and D2D density. Moreover, the D2D scale properties are obtained, which provide the sufficient conditions for achieving the anticipated SE. Numerical results corroborate our analysis and show that the proposed power control solution can efficiently mitigate interference between the cellular and D2D tier. The results demonstrate that there exists the optimal D2D density for maximizing the area SE of D2D tier. In addition, the achievable EE of a cellular user can be comparable to that of a D2D user.
[ { "created": "Fri, 13 Jan 2017 17:48:30 GMT", "version": "v1" }, { "created": "Wed, 17 May 2017 15:36:27 GMT", "version": "v2" }, { "created": "Mon, 29 May 2017 16:20:58 GMT", "version": "v3" } ]
2017-05-30
[ [ "He", "Anqi", "" ], [ "Wang", "Lifeng", "" ], [ "Chen", "Yue", "" ], [ "Wong", "Kai-Kit", "" ], [ "Elkashlan", "Maged", "" ] ]
One of key 5G scenarios is that device-to-device (D2D) and massive multiple-input multiple-output (MIMO) will be co-existed. However, interference in the uplink D2D underlaid massive MIMO cellular networks needs to be coordinated, due to the vast cellular and D2D transmissions. To this end, this paper introduces a spatially dynamic power control solution for mitigating the cellular-to-D2D and D2D-to-cellular interference. In particular, the proposed D2D power control policy is rather flexible including the special cases of no D2D links or using maximum transmit power. Under the considered power control, an analytical approach is developed to evaluate the spectral efficiency (SE) and energy efficiency (EE) in such networks. Thus, the exact expressions of SE for a cellular user or D2D transmitter are derived, which quantify the impacts of key system parameters such as massive MIMO antennas and D2D density. Moreover, the D2D scale properties are obtained, which provide the sufficient conditions for achieving the anticipated SE. Numerical results corroborate our analysis and show that the proposed power control solution can efficiently mitigate interference between the cellular and D2D tier. The results demonstrate that there exists the optimal D2D density for maximizing the area SE of D2D tier. In addition, the achievable EE of a cellular user can be comparable to that of a D2D user.
2106.06614
Harald Woracek
Ana Sokolova, Harald Woracek
Nawrotzki's Algorithm for the Countable Splitting Lemma, Constructively
null
null
null
null
cs.LO math.PR
http://creativecommons.org/licenses/by-nc-sa/4.0/
We reprove the countable splitting lemma by adapting Nawrotzki's algorithm which produces a sequence that converges to a solution. Our algorithm combines Nawrotzki's approach with taking finite cuts. It is constructive in the sense that each term of the iteratively built approximating sequence as well as the error between the approximants and the solution is computable with finitely many algebraic operations.
[ { "created": "Fri, 11 Jun 2021 21:18:44 GMT", "version": "v1" } ]
2021-06-15
[ [ "Sokolova", "Ana", "" ], [ "Woracek", "Harald", "" ] ]
We reprove the countable splitting lemma by adapting Nawrotzki's algorithm which produces a sequence that converges to a solution. Our algorithm combines Nawrotzki's approach with taking finite cuts. It is constructive in the sense that each term of the iteratively built approximating sequence as well as the error between the approximants and the solution is computable with finitely many algebraic operations.
2407.09015
Van-Giang Trinh
Van-Giang Trinh, Belaid Benhamou
Static Analysis of Logic Programs via Boolean Networks
null
null
null
null
cs.LO cs.AI
http://creativecommons.org/licenses/by/4.0/
Answer Set Programming (ASP) is a declarative problem solving paradigm that can be used to encode a combinatorial problem as a logic program whose stable models correspond to the solutions of the considered problem. ASP has been widely applied to various domains in AI and beyond. The question "What can be said about stable models of a logic program from its static information?" has been investigated and proved useful in many circumstances. In this work, we dive into this direction more deeply by making the connection between a logic program and a Boolean network, which is a prominent modeling framework with applications to various areas. The proposed connection can bring the existing results in the rich history on static analysis of Boolean networks to explore and prove more theoretical results on ASP, making it become a unified and powerful tool to further study the static analysis of ASP. In particular, the newly obtained insights have the potential to benefit many problems in the field of ASP.
[ { "created": "Fri, 12 Jul 2024 06:07:05 GMT", "version": "v1" } ]
2024-07-15
[ [ "Trinh", "Van-Giang", "" ], [ "Benhamou", "Belaid", "" ] ]
Answer Set Programming (ASP) is a declarative problem solving paradigm that can be used to encode a combinatorial problem as a logic program whose stable models correspond to the solutions of the considered problem. ASP has been widely applied to various domains in AI and beyond. The question "What can be said about stable models of a logic program from its static information?" has been investigated and proved useful in many circumstances. In this work, we dive into this direction more deeply by making the connection between a logic program and a Boolean network, which is a prominent modeling framework with applications to various areas. The proposed connection can bring the existing results in the rich history on static analysis of Boolean networks to explore and prove more theoretical results on ASP, making it become a unified and powerful tool to further study the static analysis of ASP. In particular, the newly obtained insights have the potential to benefit many problems in the field of ASP.
0808.1641
Sudhakar Sahoo
Sudhakar Sahoo, Pabitra Pal Choudhury, Mithun Chakraborty
Characterization Of any Non-linear Boolean function Using A Set of Linear Operators
12 pages, 4 figures, 2 table. Submitted for possible publication in the International Journal of Computer Mathematics and Applications, July 2008
null
null
null
cs.CC nlin.CG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Global dynamics of a non-linear Cellular Automata is, in general irregular, asymmetric and unpredictable as opposed to that of a linear CA, which is highly systematic and tractable. In the past efforts have been made to systematize non-linear CA evolutions in the light of Boolean derivatives and Jacobian Matrices. In this paper two different efforts have been made: first we try to systematize non-linear CA evolution in the light of deviant states and non-deviant states. For all the non-deviant states the nearest linear rule matrix is applicable where as for the deviant states we have a set of other matrices. Second using algebraic manipulation, an efficient algorithm is proposed by which every Non-linear Boolean function can be characterized by a sequence of binary matrices.
[ { "created": "Tue, 12 Aug 2008 11:04:47 GMT", "version": "v1" } ]
2008-08-13
[ [ "Sahoo", "Sudhakar", "" ], [ "Choudhury", "Pabitra Pal", "" ], [ "Chakraborty", "Mithun", "" ] ]
Global dynamics of a non-linear Cellular Automata is, in general irregular, asymmetric and unpredictable as opposed to that of a linear CA, which is highly systematic and tractable. In the past efforts have been made to systematize non-linear CA evolutions in the light of Boolean derivatives and Jacobian Matrices. In this paper two different efforts have been made: first we try to systematize non-linear CA evolution in the light of deviant states and non-deviant states. For all the non-deviant states the nearest linear rule matrix is applicable where as for the deviant states we have a set of other matrices. Second using algebraic manipulation, an efficient algorithm is proposed by which every Non-linear Boolean function can be characterized by a sequence of binary matrices.
1407.1667
Sumit Nain
Sumit Nain (Rice University), Yoad Lustig (Rice University), Moshe Y Vardi (Rice University)
Synthesis from Probabilistic Components
null
Logical Methods in Computer Science, Volume 10, Issue 2 (June 30, 2014) lmcs:1181
10.2168/LMCS-10(2:17)2014
null
cs.LO
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Synthesis is the automatic construction of a system from its specification. In classical synthesis algorithms, it is always assumed that the system is "constructed from scratch" rather than composed from reusable components. This, of course, rarely happens in real life, where almost every non-trivial commercial software system relies heavily on using libraries of reusable components. Furthermore, other contexts, such as web-service orchestration, can be modeled as synthesis of a system from a library of components. Recently, Lustig and Vardi introduced dataflow and control-flow synthesis from libraries of reusable components. They proved that dataflow synthesis is undecidable, while control-flow synthesis is decidable. In this work, we consider the problem of control-flow synthesis from libraries of probabilistic components . We show that this more general problem is also decidable.
[ { "created": "Mon, 7 Jul 2014 11:10:16 GMT", "version": "v1" }, { "created": "Wed, 9 Jul 2014 12:52:57 GMT", "version": "v2" } ]
2015-07-01
[ [ "Nain", "Sumit", "", "Rice University" ], [ "Lustig", "Yoad", "", "Rice University" ], [ "Vardi", "Moshe Y", "", "Rice University" ] ]
Synthesis is the automatic construction of a system from its specification. In classical synthesis algorithms, it is always assumed that the system is "constructed from scratch" rather than composed from reusable components. This, of course, rarely happens in real life, where almost every non-trivial commercial software system relies heavily on using libraries of reusable components. Furthermore, other contexts, such as web-service orchestration, can be modeled as synthesis of a system from a library of components. Recently, Lustig and Vardi introduced dataflow and control-flow synthesis from libraries of reusable components. They proved that dataflow synthesis is undecidable, while control-flow synthesis is decidable. In this work, we consider the problem of control-flow synthesis from libraries of probabilistic components . We show that this more general problem is also decidable.
2203.10290
Shi Hu
Shi Hu, Eric Nalisnick, Max Welling
Adversarial Defense via Image Denoising with Chaotic Encryption
null
null
null
null
cs.LG cs.CR
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In the literature on adversarial examples, white box and black box attacks have received the most attention. The adversary is assumed to have either full (white) or no (black) access to the defender's model. In this work, we focus on the equally practical gray box setting, assuming an attacker has partial information. We propose a novel defense that assumes everything but a private key will be made available to the attacker. Our framework uses an image denoising procedure coupled with encryption via a discretized Baker map. Extensive testing against adversarial images (e.g. FGSM, PGD) crafted using various gradients shows that our defense achieves significantly better results on CIFAR-10 and CIFAR-100 than the state-of-the-art gray box defenses in both natural and adversarial accuracy.
[ { "created": "Sat, 19 Mar 2022 10:25:02 GMT", "version": "v1" } ]
2022-03-22
[ [ "Hu", "Shi", "" ], [ "Nalisnick", "Eric", "" ], [ "Welling", "Max", "" ] ]
In the literature on adversarial examples, white box and black box attacks have received the most attention. The adversary is assumed to have either full (white) or no (black) access to the defender's model. In this work, we focus on the equally practical gray box setting, assuming an attacker has partial information. We propose a novel defense that assumes everything but a private key will be made available to the attacker. Our framework uses an image denoising procedure coupled with encryption via a discretized Baker map. Extensive testing against adversarial images (e.g. FGSM, PGD) crafted using various gradients shows that our defense achieves significantly better results on CIFAR-10 and CIFAR-100 than the state-of-the-art gray box defenses in both natural and adversarial accuracy.
1710.00811
Aaron Tuor
Aaron Tuor, Samuel Kaplan, Brian Hutchinson, Nicole Nichols, Sean Robinson
Deep Learning for Unsupervised Insider Threat Detection in Structured Cybersecurity Data Streams
Proceedings of AI for Cyber Security Workshop at AAAI 2017
null
null
null
cs.NE cs.CR cs.LG stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Analysis of an organization's computer network activity is a key component of early detection and mitigation of insider threat, a growing concern for many organizations. Raw system logs are a prototypical example of streaming data that can quickly scale beyond the cognitive power of a human analyst. As a prospective filter for the human analyst, we present an online unsupervised deep learning approach to detect anomalous network activity from system logs in real time. Our models decompose anomaly scores into the contributions of individual user behavior features for increased interpretability to aid analysts reviewing potential cases of insider threat. Using the CERT Insider Threat Dataset v6.2 and threat detection recall as our performance metric, our novel deep and recurrent neural network models outperform Principal Component Analysis, Support Vector Machine and Isolation Forest based anomaly detection baselines. For our best model, the events labeled as insider threat activity in our dataset had an average anomaly score in the 95.53 percentile, demonstrating our approach's potential to greatly reduce analyst workloads.
[ { "created": "Mon, 2 Oct 2017 17:54:28 GMT", "version": "v1" }, { "created": "Fri, 15 Dec 2017 20:53:03 GMT", "version": "v2" } ]
2017-12-19
[ [ "Tuor", "Aaron", "" ], [ "Kaplan", "Samuel", "" ], [ "Hutchinson", "Brian", "" ], [ "Nichols", "Nicole", "" ], [ "Robinson", "Sean", "" ] ]
Analysis of an organization's computer network activity is a key component of early detection and mitigation of insider threat, a growing concern for many organizations. Raw system logs are a prototypical example of streaming data that can quickly scale beyond the cognitive power of a human analyst. As a prospective filter for the human analyst, we present an online unsupervised deep learning approach to detect anomalous network activity from system logs in real time. Our models decompose anomaly scores into the contributions of individual user behavior features for increased interpretability to aid analysts reviewing potential cases of insider threat. Using the CERT Insider Threat Dataset v6.2 and threat detection recall as our performance metric, our novel deep and recurrent neural network models outperform Principal Component Analysis, Support Vector Machine and Isolation Forest based anomaly detection baselines. For our best model, the events labeled as insider threat activity in our dataset had an average anomaly score in the 95.53 percentile, demonstrating our approach's potential to greatly reduce analyst workloads.
2009.06381
Anas Blasi
Mohammed A. Alsuwaiket, Anas H. Blasi, Khawla Altarawneh
Refining Student Marks based on Enrolled Modules Assessment Methods using Data Mining Techniques
arXiv admin note: substantial text overlap with arXiv:2008.13255
null
null
null
cs.CY
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Choosing the right and effective way to assess students is one of the most important tasks of higher education. Many studies have shown that students tend to receive higher scores during their studies when assessed by different study methods which include units that are fully assessed by varying the duration of study or a combination of courses and exams than by exams alone. Many Educational Data Mining studies process data in advance through traditional data extraction, including the data preparation process. In this paper, we propose a different data preparation process by investigating more than 230000 student records for the preparation of scores. The data have been processed through diverse stages in order to extract a categorical factor through which students module marks are refined during the data preparation stage. The results of this work show that students final marks should not be isolated from the nature of the enrolled module assessment methods. They must rather be investigated thoroughly and considered during EDM data preprocessing stage. More generally, educational data should not be prepared in the same way normal data are due to the differences in data sources, applications, and error types. The effect of Module Assessment Index on the prediction process using Random Forest and Naive Bayes classification techniques were investigated. It was shown that considering MAI as attribute increases the accuracy of predicting students second year averages based on their first year averages.
[ { "created": "Sun, 30 Aug 2020 19:47:45 GMT", "version": "v1" } ]
2020-09-15
[ [ "Alsuwaiket", "Mohammed A.", "" ], [ "Blasi", "Anas H.", "" ], [ "Altarawneh", "Khawla", "" ] ]
Choosing the right and effective way to assess students is one of the most important tasks of higher education. Many studies have shown that students tend to receive higher scores during their studies when assessed by different study methods which include units that are fully assessed by varying the duration of study or a combination of courses and exams than by exams alone. Many Educational Data Mining studies process data in advance through traditional data extraction, including the data preparation process. In this paper, we propose a different data preparation process by investigating more than 230000 student records for the preparation of scores. The data have been processed through diverse stages in order to extract a categorical factor through which students module marks are refined during the data preparation stage. The results of this work show that students final marks should not be isolated from the nature of the enrolled module assessment methods. They must rather be investigated thoroughly and considered during EDM data preprocessing stage. More generally, educational data should not be prepared in the same way normal data are due to the differences in data sources, applications, and error types. The effect of Module Assessment Index on the prediction process using Random Forest and Naive Bayes classification techniques were investigated. It was shown that considering MAI as attribute increases the accuracy of predicting students second year averages based on their first year averages.
2402.09845
Maik Ender
Maik Ender and Felix Hahn and Marc Fyrbiak and Amir Moradi and Christof Paar
JustSTART: How to Find an RSA Authentication Bypass on Xilinx UltraScale(+) with Fuzzing
null
null
null
null
cs.CR
http://creativecommons.org/licenses/by/4.0/
Fuzzing is a well-established technique in the software domain to uncover bugs and vulnerabilities. Yet, applications of fuzzing for security vulnerabilities in hardware systems are scarce, as principal reasons are requirements for design information access (HDL source code). Moreover, observation of internal hardware state during runtime is typically an ineffective information source, as its documentation is often not publicly available. In addition, such observation during runtime is also inefficient due to bandwidth-limited analysis interfaces (JTAG, and minimal introspection of internal modules). In this work, we investigate fuzzing for 7-Series and UltraScale(+) FPGA configuration engines, the control plane governing the (secure) bitstream configuration within the FPGA. Our goal is to examine the effectiveness of fuzzing to analyze and document the opaque inner workings of FPGA configuration engines, with a primary emphasis on identifying security vulnerabilities. Using only the publicly available chip and dispersed documentation, we first design and implement ConFuzz, an advanced FPGA configuration engine fuzzing and rapid prototyping framework. Based on our detailed understanding of the bitstream file format, we then systematically define 3 novel key fuzzing strategies for Xilinx configuration engines. Moreover, our strategies are executed through mutational structure-aware fuzzers and incorporate various novel custom-tailored, FPGA-specific optimizations. Our evaluation reveals previously undocumented behavior within the configuration engine, including critical findings such as system crashes leading to unresponsive states of the FPGA. In addition, our investigations not only lead to the rediscovery of the starbleed attack but also uncover JustSTART (CVE-2023-20570), capable of circumventing RSA authentication for Xilinx UltraScale(+). Note that we also discuss countermeasures.
[ { "created": "Thu, 15 Feb 2024 10:03:35 GMT", "version": "v1" } ]
2024-02-16
[ [ "Ender", "Maik", "" ], [ "Hahn", "Felix", "" ], [ "Fyrbiak", "Marc", "" ], [ "Moradi", "Amir", "" ], [ "Paar", "Christof", "" ] ]
Fuzzing is a well-established technique in the software domain to uncover bugs and vulnerabilities. Yet, applications of fuzzing for security vulnerabilities in hardware systems are scarce, as principal reasons are requirements for design information access (HDL source code). Moreover, observation of internal hardware state during runtime is typically an ineffective information source, as its documentation is often not publicly available. In addition, such observation during runtime is also inefficient due to bandwidth-limited analysis interfaces (JTAG, and minimal introspection of internal modules). In this work, we investigate fuzzing for 7-Series and UltraScale(+) FPGA configuration engines, the control plane governing the (secure) bitstream configuration within the FPGA. Our goal is to examine the effectiveness of fuzzing to analyze and document the opaque inner workings of FPGA configuration engines, with a primary emphasis on identifying security vulnerabilities. Using only the publicly available chip and dispersed documentation, we first design and implement ConFuzz, an advanced FPGA configuration engine fuzzing and rapid prototyping framework. Based on our detailed understanding of the bitstream file format, we then systematically define 3 novel key fuzzing strategies for Xilinx configuration engines. Moreover, our strategies are executed through mutational structure-aware fuzzers and incorporate various novel custom-tailored, FPGA-specific optimizations. Our evaluation reveals previously undocumented behavior within the configuration engine, including critical findings such as system crashes leading to unresponsive states of the FPGA. In addition, our investigations not only lead to the rediscovery of the starbleed attack but also uncover JustSTART (CVE-2023-20570), capable of circumventing RSA authentication for Xilinx UltraScale(+). Note that we also discuss countermeasures.
1103.3099
Rahul Urgaonkar
Rahul Urgaonkar, Bhuvan Urgaonkar, Michael J. Neely, Anand Sivasubramaniam
Optimal Power Cost Management Using Stored Energy in Data Centers
Full version of Sigmetrics 2011 paper
null
null
null
cs.PF cs.SY math.OC
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Since the electricity bill of a data center constitutes a significant portion of its overall operational costs, reducing this has become important. We investigate cost reduction opportunities that arise by the use of uninterrupted power supply (UPS) units as energy storage devices. This represents a deviation from the usual use of these devices as mere transitional fail-over mechanisms between utility and captive sources such as diesel generators. We consider the problem of opportunistically using these devices to reduce the time average electric utility bill in a data center. Using the technique of Lyapunov optimization, we develop an online control algorithm that can optimally exploit these devices to minimize the time average cost. This algorithm operates without any knowledge of the statistics of the workload or electricity cost processes, making it attractive in the presence of workload and pricing uncertainties. An interesting feature of our algorithm is that its deviation from optimality reduces as the storage capacity is increased. Our work opens up a new area in data center power management.
[ { "created": "Wed, 16 Mar 2011 05:37:18 GMT", "version": "v1" }, { "created": "Sat, 19 Mar 2011 12:58:09 GMT", "version": "v2" } ]
2011-03-22
[ [ "Urgaonkar", "Rahul", "" ], [ "Urgaonkar", "Bhuvan", "" ], [ "Neely", "Michael J.", "" ], [ "Sivasubramaniam", "Anand", "" ] ]
Since the electricity bill of a data center constitutes a significant portion of its overall operational costs, reducing this has become important. We investigate cost reduction opportunities that arise by the use of uninterrupted power supply (UPS) units as energy storage devices. This represents a deviation from the usual use of these devices as mere transitional fail-over mechanisms between utility and captive sources such as diesel generators. We consider the problem of opportunistically using these devices to reduce the time average electric utility bill in a data center. Using the technique of Lyapunov optimization, we develop an online control algorithm that can optimally exploit these devices to minimize the time average cost. This algorithm operates without any knowledge of the statistics of the workload or electricity cost processes, making it attractive in the presence of workload and pricing uncertainties. An interesting feature of our algorithm is that its deviation from optimality reduces as the storage capacity is increased. Our work opens up a new area in data center power management.
2011.13614
Shanshan Wang
Kehan Qi, Yu Gong, Xinfeng Liu, Xin Liu, Hairong Zheng, Shanshan Wang
Multi-task MR Imaging with Iterative Teacher Forcing and Re-weighted Deep Learning
null
null
null
null
cs.CV cs.AI
http://creativecommons.org/publicdomain/zero/1.0/
Noises, artifacts, and loss of information caused by the magnetic resonance (MR) reconstruction may compromise the final performance of the downstream applications. In this paper, we develop a re-weighted multi-task deep learning method to learn prior knowledge from the existing big dataset and then utilize them to assist simultaneous MR reconstruction and segmentation from the under-sampled k-space data. The multi-task deep learning framework is equipped with two network sub-modules, which are integrated and trained by our designed iterative teacher forcing scheme (ITFS) under the dynamic re-weighted loss constraint (DRLC). The ITFS is designed to avoid error accumulation by injecting the fully-sampled data into the training process. The DRLC is proposed to dynamically balance the contributions from the reconstruction and segmentation sub-modules so as to co-prompt the multi-task accuracy. The proposed method has been evaluated on two open datasets and one in vivo in-house dataset and compared to six state-of-the-art methods. Results show that the proposed method possesses encouraging capabilities for simultaneous and accurate MR reconstruction and segmentation.
[ { "created": "Fri, 27 Nov 2020 09:08:05 GMT", "version": "v1" } ]
2020-11-30
[ [ "Qi", "Kehan", "" ], [ "Gong", "Yu", "" ], [ "Liu", "Xinfeng", "" ], [ "Liu", "Xin", "" ], [ "Zheng", "Hairong", "" ], [ "Wang", "Shanshan", "" ] ]
Noises, artifacts, and loss of information caused by the magnetic resonance (MR) reconstruction may compromise the final performance of the downstream applications. In this paper, we develop a re-weighted multi-task deep learning method to learn prior knowledge from the existing big dataset and then utilize them to assist simultaneous MR reconstruction and segmentation from the under-sampled k-space data. The multi-task deep learning framework is equipped with two network sub-modules, which are integrated and trained by our designed iterative teacher forcing scheme (ITFS) under the dynamic re-weighted loss constraint (DRLC). The ITFS is designed to avoid error accumulation by injecting the fully-sampled data into the training process. The DRLC is proposed to dynamically balance the contributions from the reconstruction and segmentation sub-modules so as to co-prompt the multi-task accuracy. The proposed method has been evaluated on two open datasets and one in vivo in-house dataset and compared to six state-of-the-art methods. Results show that the proposed method possesses encouraging capabilities for simultaneous and accurate MR reconstruction and segmentation.
2207.05959
Jianghong Ma
Tianjun Wei, Jianghong Ma, Tommy W. S. Chow
Fine-tuning Partition-aware Item Similarities for Efficient and Scalable Recommendation
Accepted by The 2023 ACM Web Conference (WWW 2023)
null
10.1145/3543507.3583240
null
cs.IR cs.LG
http://creativecommons.org/licenses/by-sa/4.0/
Collaborative filtering (CF) is widely searched in recommendation with various types of solutions. Recent success of Graph Convolution Networks (GCN) in CF demonstrates the effectiveness of modeling high-order relationships through graphs, while repetitive graph convolution and iterative batch optimization limit their efficiency. Instead, item similarity models attempt to construct direct relationships through efficient interaction encoding. Despite their great performance, the growing item numbers result in quadratic growth in similarity modeling process, posing critical scalability problems. In this paper, we investigate the graph sampling strategy adopted in latest GCN model for efficiency improving, and identify the potential item group structure in the sampled graph. Based on this, we propose a novel item similarity model which introduces graph partitioning to restrict the item similarity modeling within each partition. Specifically, we show that the spectral information of the original graph is well in preserving global-level information. Then, it is added to fine-tune local item similarities with a new data augmentation strategy acted as partition-aware prior knowledge, jointly to cope with the information loss brought by partitioning. Experiments carried out on 4 datasets show that the proposed model outperforms state-of-the-art GCN models with 10x speed-up and item similarity models with 95\% parameter storage savings.
[ { "created": "Wed, 13 Jul 2022 04:37:48 GMT", "version": "v1" }, { "created": "Fri, 10 Feb 2023 07:10:42 GMT", "version": "v2" } ]
2023-02-13
[ [ "Wei", "Tianjun", "" ], [ "Ma", "Jianghong", "" ], [ "Chow", "Tommy W. S.", "" ] ]
Collaborative filtering (CF) is widely searched in recommendation with various types of solutions. Recent success of Graph Convolution Networks (GCN) in CF demonstrates the effectiveness of modeling high-order relationships through graphs, while repetitive graph convolution and iterative batch optimization limit their efficiency. Instead, item similarity models attempt to construct direct relationships through efficient interaction encoding. Despite their great performance, the growing item numbers result in quadratic growth in similarity modeling process, posing critical scalability problems. In this paper, we investigate the graph sampling strategy adopted in latest GCN model for efficiency improving, and identify the potential item group structure in the sampled graph. Based on this, we propose a novel item similarity model which introduces graph partitioning to restrict the item similarity modeling within each partition. Specifically, we show that the spectral information of the original graph is well in preserving global-level information. Then, it is added to fine-tune local item similarities with a new data augmentation strategy acted as partition-aware prior knowledge, jointly to cope with the information loss brought by partitioning. Experiments carried out on 4 datasets show that the proposed model outperforms state-of-the-art GCN models with 10x speed-up and item similarity models with 95\% parameter storage savings.
1310.2322
Morgan Chopin
Janka Chleb\'ikov\'a and Morgan Chopin
The Firefighter Problem: A Structural Analysis
null
null
null
null
cs.DM cs.DS math.CO
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We consider the complexity of the firefighter problem where b>=1 firefighters are available at each time step. This problem is proved NP-complete even on trees of degree at most three and budget one (Finbow et al.,2007) and on trees of bounded degree b+3 for any fixed budget b>=2 (Bazgan et al.,2012). In this paper, we provide further insight into the complexity landscape of the problem by showing that the pathwidth and the maximum degree of the input graph govern its complexity. More precisely, we first prove that the problem is NP-complete even on trees of pathwidth at most three for any fixed budget b>=1. We then show that the problem turns out to be fixed parameter-tractable with respect to the combined parameter "pathwidth" and "maximum degree" of the input graph.
[ { "created": "Wed, 9 Oct 2013 01:39:10 GMT", "version": "v1" }, { "created": "Mon, 28 Apr 2014 07:49:59 GMT", "version": "v2" } ]
2014-04-29
[ [ "Chlebíková", "Janka", "" ], [ "Chopin", "Morgan", "" ] ]
We consider the complexity of the firefighter problem where b>=1 firefighters are available at each time step. This problem is proved NP-complete even on trees of degree at most three and budget one (Finbow et al.,2007) and on trees of bounded degree b+3 for any fixed budget b>=2 (Bazgan et al.,2012). In this paper, we provide further insight into the complexity landscape of the problem by showing that the pathwidth and the maximum degree of the input graph govern its complexity. More precisely, we first prove that the problem is NP-complete even on trees of pathwidth at most three for any fixed budget b>=1. We then show that the problem turns out to be fixed parameter-tractable with respect to the combined parameter "pathwidth" and "maximum degree" of the input graph.
2308.02043
Zulqarnain Rashid Dr
Zulqarnain Rashid, Amos A Folarin, Yatharth Ranjan, Pauline Conde, Heet Sankesara, Yuezhou Zhang, Shaoxiong Sun, Callum Stewart, Petroula Laiou, Richard JB Dobson
Disease Insight through Digital Biomarkers Developed by Remotely Collected Wearables and Smartphone Data
null
null
null
null
cs.CY cs.AI
http://creativecommons.org/licenses/by-nc-sa/4.0/
Digital Biomarkers and remote patient monitoring can provide valuable and timely insights into how a patient is coping with their condition (disease progression, treatment response, etc.), complementing treatment in traditional healthcare settings.Smartphones with embedded and connected sensors have immense potential for improving healthcare through various apps and mHealth (mobile health) platforms. This capability could enable the development of reliable digital biomarkers from long-term longitudinal data collected remotely from patients. We built an open-source platform, RADAR-base, to support large-scale data collection in remote monitoring studies. RADAR-base is a modern remote data collection platform built around Confluent's Apache Kafka, to support scalability, extensibility, security, privacy and quality of data. It provides support for study design and set-up, active (eg PROMs) and passive (eg. phone sensors, wearable devices and IoT) remote data collection capabilities with feature generation (eg. behavioural, environmental and physiological markers). The backend enables secure data transmission, and scalable solutions for data storage, management and data access. The platform has successfully collected longitudinal data for various cohorts in a number of disease areas including Multiple Sclerosis, Depression, Epilepsy, ADHD, Alzheimer, Autism and Lung diseases. Digital biomarkers developed through collected data are providing useful insights into different diseases. RADAR-base provides a modern open-source, community-driven solution for remote monitoring, data collection, and digital phenotyping of physical and mental health diseases. Clinicians can use digital biomarkers to augment their decision making for the prevention, personalisation and early intervention of disease.
[ { "created": "Thu, 3 Aug 2023 22:44:48 GMT", "version": "v1" } ]
2023-08-07
[ [ "Rashid", "Zulqarnain", "" ], [ "Folarin", "Amos A", "" ], [ "Ranjan", "Yatharth", "" ], [ "Conde", "Pauline", "" ], [ "Sankesara", "Heet", "" ], [ "Zhang", "Yuezhou", "" ], [ "Sun", "Shaoxiong", "" ], [ "Stewart", "Callum", "" ], [ "Laiou", "Petroula", "" ], [ "Dobson", "Richard JB", "" ] ]
Digital Biomarkers and remote patient monitoring can provide valuable and timely insights into how a patient is coping with their condition (disease progression, treatment response, etc.), complementing treatment in traditional healthcare settings.Smartphones with embedded and connected sensors have immense potential for improving healthcare through various apps and mHealth (mobile health) platforms. This capability could enable the development of reliable digital biomarkers from long-term longitudinal data collected remotely from patients. We built an open-source platform, RADAR-base, to support large-scale data collection in remote monitoring studies. RADAR-base is a modern remote data collection platform built around Confluent's Apache Kafka, to support scalability, extensibility, security, privacy and quality of data. It provides support for study design and set-up, active (eg PROMs) and passive (eg. phone sensors, wearable devices and IoT) remote data collection capabilities with feature generation (eg. behavioural, environmental and physiological markers). The backend enables secure data transmission, and scalable solutions for data storage, management and data access. The platform has successfully collected longitudinal data for various cohorts in a number of disease areas including Multiple Sclerosis, Depression, Epilepsy, ADHD, Alzheimer, Autism and Lung diseases. Digital biomarkers developed through collected data are providing useful insights into different diseases. RADAR-base provides a modern open-source, community-driven solution for remote monitoring, data collection, and digital phenotyping of physical and mental health diseases. Clinicians can use digital biomarkers to augment their decision making for the prevention, personalisation and early intervention of disease.
2005.00842
Tatsuki Kuribayashi
Tatsuki Kuribayashi, Takumi Ito, Jun Suzuki, Kentaro Inui
Language Models as an Alternative Evaluator of Word Order Hypotheses: A Case Study in Japanese
Accepted by ACL2020
null
null
null
cs.CL
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We examine a methodology using neural language models (LMs) for analyzing the word order of language. This LM-based method has the potential to overcome the difficulties existing methods face, such as the propagation of preprocessor errors in count-based methods. In this study, we explore whether the LM-based method is valid for analyzing the word order. As a case study, this study focuses on Japanese due to its complex and flexible word order. To validate the LM-based method, we test (i) parallels between LMs and human word order preference, and (ii) consistency of the results obtained using the LM-based method with previous linguistic studies. Through our experiments, we tentatively conclude that LMs display sufficient word order knowledge for usage as an analysis tool. Finally, using the LM-based method, we demonstrate the relationship between the canonical word order and topicalization, which had yet to be analyzed by large-scale experiments.
[ { "created": "Sat, 2 May 2020 14:32:40 GMT", "version": "v1" } ]
2020-05-05
[ [ "Kuribayashi", "Tatsuki", "" ], [ "Ito", "Takumi", "" ], [ "Suzuki", "Jun", "" ], [ "Inui", "Kentaro", "" ] ]
We examine a methodology using neural language models (LMs) for analyzing the word order of language. This LM-based method has the potential to overcome the difficulties existing methods face, such as the propagation of preprocessor errors in count-based methods. In this study, we explore whether the LM-based method is valid for analyzing the word order. As a case study, this study focuses on Japanese due to its complex and flexible word order. To validate the LM-based method, we test (i) parallels between LMs and human word order preference, and (ii) consistency of the results obtained using the LM-based method with previous linguistic studies. Through our experiments, we tentatively conclude that LMs display sufficient word order knowledge for usage as an analysis tool. Finally, using the LM-based method, we demonstrate the relationship between the canonical word order and topicalization, which had yet to be analyzed by large-scale experiments.
2212.11725
Fabrice Rossi
Aichetou Bouchareb (SAMM), Marc Boull\'e, Fabrice Cl\'erot, Fabrice Rossi (CEREMADE)
Model Based Co-clustering of Mixed Numerical and Binary Data
null
Advances in Knowledge Discovery and Management, 834, Springer International Publishing, pp.3-22, 2019, Studies in Computational Intelligence
10.1007/978-3-030-18129-1_1
null
cs.LG math.ST stat.ML stat.TH
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Co-clustering is a data mining technique used to extract the underlying block structure between the rows and columns of a data matrix. Many approaches have been studied and have shown their capacity to extract such structures in continuous, binary or contingency tables. However, very little work has been done to perform co-clustering on mixed type data. In this article, we extend the latent block models based co-clustering to the case of mixed data (continuous and binary variables). We then evaluate the effectiveness of the proposed approach on simulated data and we discuss its advantages and potential limits.
[ { "created": "Thu, 22 Dec 2022 14:16:08 GMT", "version": "v1" } ]
2022-12-23
[ [ "Bouchareb", "Aichetou", "", "SAMM" ], [ "Boullé", "Marc", "", "CEREMADE" ], [ "Clérot", "Fabrice", "", "CEREMADE" ], [ "Rossi", "Fabrice", "", "CEREMADE" ] ]
Co-clustering is a data mining technique used to extract the underlying block structure between the rows and columns of a data matrix. Many approaches have been studied and have shown their capacity to extract such structures in continuous, binary or contingency tables. However, very little work has been done to perform co-clustering on mixed type data. In this article, we extend the latent block models based co-clustering to the case of mixed data (continuous and binary variables). We then evaluate the effectiveness of the proposed approach on simulated data and we discuss its advantages and potential limits.