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
2306.08403
Oumaima El Khettari
Oumaima El Khettari, Solen Quiniou, Samuel Chaffron
Building a Corpus for Biomedical Relation Extraction of Species Mentions
Accepted in BioNLP@ACL 2023
null
null
null
cs.CL
http://creativecommons.org/licenses/by/4.0/
We present a manually annotated corpus, Species-Species Interaction, for extracting meaningful binary relations between species, in biomedical texts, at sentence level, with a focus on the gut microbiota. The corpus leverages PubTator to annotate species in full-text articles after evaluating different Named Entity Recognition species taggers. Our first results are promising for extracting relations between species using BERT and its biomedical variants.
[ { "created": "Wed, 14 Jun 2023 09:56:32 GMT", "version": "v1" } ]
2023-06-16
[ [ "Khettari", "Oumaima El", "" ], [ "Quiniou", "Solen", "" ], [ "Chaffron", "Samuel", "" ] ]
We present a manually annotated corpus, Species-Species Interaction, for extracting meaningful binary relations between species, in biomedical texts, at sentence level, with a focus on the gut microbiota. The corpus leverages PubTator to annotate species in full-text articles after evaluating different Named Entity Recognition species taggers. Our first results are promising for extracting relations between species using BERT and its biomedical variants.
2310.03702
Samuel Taggart
Jason Hartline, Darrell Hoy, and Samuel Taggart
Robust Analysis of Auction Equilibria
This paper provides an economic interpretation on results presented in an extended abstract under the title "Price of Anarchy for Auction Revenue" at the fifteenth ACM Conference on Economics and Computation
null
null
null
cs.GT
http://creativecommons.org/licenses/by/4.0/
Equilibria in auctions can be very difficult to analyze, beyond the symmetric environments where revenue equivalence renders the analysis straightforward. This paper takes a robust approach to evaluating the equilibria of auctions. Rather than identify the equilibria of an auction under specific environmental conditions, it considers worst-case analysis, where an auction is evaluated according to the worst environment and worst equilibrium in that environment. It identifies a non-equilibrium property of auctions that governs whether or not their worst-case equilibria are good for welfare and revenue. This property is easy to analyze, can be refined from data, and composes across markets where multiple auctions are run simultaneously.
[ { "created": "Thu, 5 Oct 2023 17:23:09 GMT", "version": "v1" } ]
2023-10-06
[ [ "Hartline", "Jason", "" ], [ "Hoy", "Darrell", "" ], [ "Taggart", "Samuel", "" ] ]
Equilibria in auctions can be very difficult to analyze, beyond the symmetric environments where revenue equivalence renders the analysis straightforward. This paper takes a robust approach to evaluating the equilibria of auctions. Rather than identify the equilibria of an auction under specific environmental conditions, it considers worst-case analysis, where an auction is evaluated according to the worst environment and worst equilibrium in that environment. It identifies a non-equilibrium property of auctions that governs whether or not their worst-case equilibria are good for welfare and revenue. This property is easy to analyze, can be refined from data, and composes across markets where multiple auctions are run simultaneously.
1709.02285
Darius Burschka
Darius Burschka
Monocular Navigation in Large Scale Dynamic Environments
2017 British Machine Vision Conference, London (BMVC 2017)
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We present a processing technique for a robust reconstruction of motion properties for single points in large scale, dynamic environments. We assume that the acquisition camera is moving and that there are other independently moving agents in a large environment, like road scenarios. The separation of direction and magnitude of the reconstructed motion allows for robust reconstruction of the dynamic state of the objects in situations, where conventional binocular systems fail due to a small signal (disparity) from the images due to a constant detection error, and where structure from motion approaches fail due to unobserved motion of other agents between the camera frames. We present the mathematical framework and the sensitivity analysis for the resulting system.
[ { "created": "Thu, 7 Sep 2017 14:46:45 GMT", "version": "v1" } ]
2017-09-08
[ [ "Burschka", "Darius", "" ] ]
We present a processing technique for a robust reconstruction of motion properties for single points in large scale, dynamic environments. We assume that the acquisition camera is moving and that there are other independently moving agents in a large environment, like road scenarios. The separation of direction and magnitude of the reconstructed motion allows for robust reconstruction of the dynamic state of the objects in situations, where conventional binocular systems fail due to a small signal (disparity) from the images due to a constant detection error, and where structure from motion approaches fail due to unobserved motion of other agents between the camera frames. We present the mathematical framework and the sensitivity analysis for the resulting system.
1211.6988
Florian Meyer
Florian Meyer, Erwin Riegler, Ondrej Hlinka, and Franz Hlawatsch
Simultaneous Distributed Sensor Self-Localization and Target Tracking Using Belief Propagation and Likelihood Consensus
10 pages, 5 figures
null
null
null
cs.NI cs.IT math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We introduce the framework of cooperative simultaneous localization and tracking (CoSLAT), which provides a consistent combination of cooperative self-localization (CSL) and distributed target tracking (DTT) in sensor networks without a fusion center. CoSLAT extends simultaneous localization and tracking (SLAT) in that it uses also intersensor measurements. Starting from a factor graph formulation of the CoSLAT problem, we develop a particle-based, distributed message passing algorithm for CoSLAT that combines nonparametric belief propagation with the likelihood consensus scheme. The proposed CoSLAT algorithm improves on state-of-the-art CSL and DTT algorithms by exchanging probabilistic information between CSL and DTT. Simulation results demonstrate substantial improvements in both self-localization and tracking performance.
[ { "created": "Thu, 29 Nov 2012 17:14:55 GMT", "version": "v1" } ]
2012-11-30
[ [ "Meyer", "Florian", "" ], [ "Riegler", "Erwin", "" ], [ "Hlinka", "Ondrej", "" ], [ "Hlawatsch", "Franz", "" ] ]
We introduce the framework of cooperative simultaneous localization and tracking (CoSLAT), which provides a consistent combination of cooperative self-localization (CSL) and distributed target tracking (DTT) in sensor networks without a fusion center. CoSLAT extends simultaneous localization and tracking (SLAT) in that it uses also intersensor measurements. Starting from a factor graph formulation of the CoSLAT problem, we develop a particle-based, distributed message passing algorithm for CoSLAT that combines nonparametric belief propagation with the likelihood consensus scheme. The proposed CoSLAT algorithm improves on state-of-the-art CSL and DTT algorithms by exchanging probabilistic information between CSL and DTT. Simulation results demonstrate substantial improvements in both self-localization and tracking performance.
1406.1528
Dustin Lang
Dustin Lang, David W. Hogg, and Bernhard Scholkopf
Towards building a Crowd-Sourced Sky Map
Appeared at AI-STATS 2014
JMLR Workshop and Conference Proceedings, 33 (AI & Statistics 2014), 549
null
null
cs.CV astro-ph.IM
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We describe a system that builds a high dynamic-range and wide-angle image of the night sky by combining a large set of input images. The method makes use of pixel-rank information in the individual input images to improve a "consensus" pixel rank in the combined image. Because it only makes use of ranks and the complexity of the algorithm is linear in the number of images, the method is useful for large sets of uncalibrated images that might have undergone unknown non-linear tone mapping transformations for visualization or aesthetic reasons. We apply the method to images of the night sky (of unknown provenance) discovered on the Web. The method permits discovery of astronomical objects or features that are not visible in any of the input images taken individually. More importantly, however, it permits scientific exploitation of a huge source of astronomical images that would not be available to astronomical research without our automatic system.
[ { "created": "Thu, 5 Jun 2014 21:18:44 GMT", "version": "v1" } ]
2014-06-09
[ [ "Lang", "Dustin", "" ], [ "Hogg", "David W.", "" ], [ "Scholkopf", "Bernhard", "" ] ]
We describe a system that builds a high dynamic-range and wide-angle image of the night sky by combining a large set of input images. The method makes use of pixel-rank information in the individual input images to improve a "consensus" pixel rank in the combined image. Because it only makes use of ranks and the complexity of the algorithm is linear in the number of images, the method is useful for large sets of uncalibrated images that might have undergone unknown non-linear tone mapping transformations for visualization or aesthetic reasons. We apply the method to images of the night sky (of unknown provenance) discovered on the Web. The method permits discovery of astronomical objects or features that are not visible in any of the input images taken individually. More importantly, however, it permits scientific exploitation of a huge source of astronomical images that would not be available to astronomical research without our automatic system.
1806.02366
Alex James Dr
Kamilya Smagulova and Kazybek Adam and Olga Krestinskaya and Alex Pappachen James
Design of CMOS-memristor Circuits for LSTM architecture
null
IEEE International Conferences on Electron Devices and Solid-State Circuits, 2018
null
null
cs.ET
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Long Short-Term memory (LSTM) architecture is a well-known approach for building recurrent neural networks (RNN) useful in sequential processing of data in application to natural language processing. The near-sensor hardware implementation of LSTM is challenged due to large parallelism and complexity. We propose a 0.18 m CMOS, GST memristor LSTM hardware architecture for near-sensor processing. The proposed system is validated in a forecasting problem based on Keras model.
[ { "created": "Wed, 6 Jun 2018 18:14:59 GMT", "version": "v1" } ]
2018-06-08
[ [ "Smagulova", "Kamilya", "" ], [ "Adam", "Kazybek", "" ], [ "Krestinskaya", "Olga", "" ], [ "James", "Alex Pappachen", "" ] ]
Long Short-Term memory (LSTM) architecture is a well-known approach for building recurrent neural networks (RNN) useful in sequential processing of data in application to natural language processing. The near-sensor hardware implementation of LSTM is challenged due to large parallelism and complexity. We propose a 0.18 m CMOS, GST memristor LSTM hardware architecture for near-sensor processing. The proposed system is validated in a forecasting problem based on Keras model.
1603.02130
Jing (Janet) Liu
Jing Liu and John D. Backes and Darren Cofer and Andrew Gacek
From Design Contracts to Component Requirements Verification
15 pages, 2 figures, conference submission
null
null
null
cs.SE
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
During the development and verification of complex airborne systems, a variety of languages and development environments are used for different levels of the system hierarchy. As a result, there may be manual steps to translate requirements between these different environments. This paper presents a tool-supported export technique that translates high-level requirements from the software architecture modeling environment into observers of requirements that can be used for verification in the software component environment. This allows efficient verification that the component designs comply with their high-level requirements. It also provides an automated tool chain supporting formal verification from system requirements down to low-level software requirements that is consistent with certification guidance for avionics systems. The effectiveness of the technique has been evaluated and demonstrated on a medical infusion pump and an aircraft wheel braking system.
[ { "created": "Mon, 7 Mar 2016 16:09:42 GMT", "version": "v1" }, { "created": "Fri, 22 Apr 2016 21:28:19 GMT", "version": "v2" } ]
2016-04-26
[ [ "Liu", "Jing", "" ], [ "Backes", "John D.", "" ], [ "Cofer", "Darren", "" ], [ "Gacek", "Andrew", "" ] ]
During the development and verification of complex airborne systems, a variety of languages and development environments are used for different levels of the system hierarchy. As a result, there may be manual steps to translate requirements between these different environments. This paper presents a tool-supported export technique that translates high-level requirements from the software architecture modeling environment into observers of requirements that can be used for verification in the software component environment. This allows efficient verification that the component designs comply with their high-level requirements. It also provides an automated tool chain supporting formal verification from system requirements down to low-level software requirements that is consistent with certification guidance for avionics systems. The effectiveness of the technique has been evaluated and demonstrated on a medical infusion pump and an aircraft wheel braking system.
2304.10664
Miriam J\"ager
Miriam J\"ager, Patrick H\"ubner, Dennis Haitz, Boris Jutzi
A Comparative Neural Radiance Field (NeRF) 3D Analysis of Camera Poses from HoloLens Trajectories and Structure from Motion
7 pages, 5 figures. Will be published in the ISPRS The International Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences
null
null
null
cs.CV
http://creativecommons.org/licenses/by/4.0/
Neural Radiance Fields (NeRFs) are trained using a set of camera poses and associated images as input to estimate density and color values for each position. The position-dependent density learning is of particular interest for photogrammetry, enabling 3D reconstruction by querying and filtering the NeRF coordinate system based on the object density. While traditional methods like Structure from Motion are commonly used for camera pose calculation in pre-processing for NeRFs, the HoloLens offers an interesting interface for extracting the required input data directly. We present a workflow for high-resolution 3D reconstructions almost directly from HoloLens data using NeRFs. Thereby, different investigations are considered: Internal camera poses from the HoloLens trajectory via a server application, and external camera poses from Structure from Motion, both with an enhanced variant applied through pose refinement. Results show that the internal camera poses lead to NeRF convergence with a PSNR of 25\,dB with a simple rotation around the x-axis and enable a 3D reconstruction. Pose refinement enables comparable quality compared to external camera poses, resulting in improved training process with a PSNR of 27\,dB and a better 3D reconstruction. Overall, NeRF reconstructions outperform the conventional photogrammetric dense reconstruction using Multi-View Stereo in terms of completeness and level of detail.
[ { "created": "Thu, 20 Apr 2023 22:17:28 GMT", "version": "v1" } ]
2023-04-24
[ [ "Jäger", "Miriam", "" ], [ "Hübner", "Patrick", "" ], [ "Haitz", "Dennis", "" ], [ "Jutzi", "Boris", "" ] ]
Neural Radiance Fields (NeRFs) are trained using a set of camera poses and associated images as input to estimate density and color values for each position. The position-dependent density learning is of particular interest for photogrammetry, enabling 3D reconstruction by querying and filtering the NeRF coordinate system based on the object density. While traditional methods like Structure from Motion are commonly used for camera pose calculation in pre-processing for NeRFs, the HoloLens offers an interesting interface for extracting the required input data directly. We present a workflow for high-resolution 3D reconstructions almost directly from HoloLens data using NeRFs. Thereby, different investigations are considered: Internal camera poses from the HoloLens trajectory via a server application, and external camera poses from Structure from Motion, both with an enhanced variant applied through pose refinement. Results show that the internal camera poses lead to NeRF convergence with a PSNR of 25\,dB with a simple rotation around the x-axis and enable a 3D reconstruction. Pose refinement enables comparable quality compared to external camera poses, resulting in improved training process with a PSNR of 27\,dB and a better 3D reconstruction. Overall, NeRF reconstructions outperform the conventional photogrammetric dense reconstruction using Multi-View Stereo in terms of completeness and level of detail.
1805.00976
Saba Ahmadian
Saba Ahmadian, Onur Mutlu, and Hossein Asadi
ECI-Cache: A High-Endurance and Cost-Efficient I/O Caching Scheme for Virtualized Platforms
null
Proceedings of the ACM on Measurement and Analysis of Computing Systems 2.1 (2018): 9
10.1145/3179412
null
cs.AR
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In recent years, high interest in using Virtual Machines (VMs) in data centers and Cloud computing has significantly increased the demand for high-performance data storage systems. Recent studies suggest using SSDs as a caching layer for HDD-based storage subsystems in virtualization platforms. Such studies neglect to address the endurance and cost of SSDs, which can significantly affect the efficiency of I/O caching. Moreover, previous studies only configure the cache size to provide the required performance level for each VM, while neglecting other important parameters such as cache write policy and request type, which can adversely affect both performance-per-cost and endurance. In this paper, we present a new high-Endurance and Cost-efficient I/O Caching (ECI-Cache) scheme for virtualized platforms, which can significantly improve both the performance-per-cost and endurance of storage subsystems as opposed to previously proposed I/O caching schemes. Unlike traditional I/O caching schemes which allocate cache size only based on reuse distance of accesses, we propose a new metric, Useful Reuse Distance (URD), which considers the request type in reuse distance calculation, resulting in improved performance-per-cost and endurance for the SSD cache. Via online characterization of workloads and using URD, ECI-Cache partitions the SSD cache across VMs and is able to dynamically adjust the cache size and write policy for each VM. To evaluate the proposed scheme, we have implemented ECI-Cache in an open source hypervisor, QEMU (version 2.8.0), on a server running the CentOS 7 operating system (kernel version 3.10.0-327). Experimental results show that our proposed scheme improves the performance, performance-per-cost, and endurance of the SSD cache by 17%, 30% and 65%, respectively, compared to the state-of-the-art dynamic cache partitioning scheme.
[ { "created": "Wed, 2 May 2018 18:41:58 GMT", "version": "v1" } ]
2018-05-04
[ [ "Ahmadian", "Saba", "" ], [ "Mutlu", "Onur", "" ], [ "Asadi", "Hossein", "" ] ]
In recent years, high interest in using Virtual Machines (VMs) in data centers and Cloud computing has significantly increased the demand for high-performance data storage systems. Recent studies suggest using SSDs as a caching layer for HDD-based storage subsystems in virtualization platforms. Such studies neglect to address the endurance and cost of SSDs, which can significantly affect the efficiency of I/O caching. Moreover, previous studies only configure the cache size to provide the required performance level for each VM, while neglecting other important parameters such as cache write policy and request type, which can adversely affect both performance-per-cost and endurance. In this paper, we present a new high-Endurance and Cost-efficient I/O Caching (ECI-Cache) scheme for virtualized platforms, which can significantly improve both the performance-per-cost and endurance of storage subsystems as opposed to previously proposed I/O caching schemes. Unlike traditional I/O caching schemes which allocate cache size only based on reuse distance of accesses, we propose a new metric, Useful Reuse Distance (URD), which considers the request type in reuse distance calculation, resulting in improved performance-per-cost and endurance for the SSD cache. Via online characterization of workloads and using URD, ECI-Cache partitions the SSD cache across VMs and is able to dynamically adjust the cache size and write policy for each VM. To evaluate the proposed scheme, we have implemented ECI-Cache in an open source hypervisor, QEMU (version 2.8.0), on a server running the CentOS 7 operating system (kernel version 3.10.0-327). Experimental results show that our proposed scheme improves the performance, performance-per-cost, and endurance of the SSD cache by 17%, 30% and 65%, respectively, compared to the state-of-the-art dynamic cache partitioning scheme.
2210.14582
Yan Huang
Xiang Long, Yan Huang, Zhendong Liu, Lansheng Han, Haili Sun, Jingyuan He
WebCrack: Dynamic Dictionary Adjustment for Web Weak Password Detection based on Blasting Response Event Discrimination
22 pages, 6 figures, 4 tables
null
null
null
cs.CR cs.DS
http://creativecommons.org/licenses/by/4.0/
The feature diversity of different web systems in page elements, submission contents and return information makes it difficult to detect weak password automatically. To solve this problem, multi-factor correlation detection method as integrated in the DBKER algorithm is proposed to achieve automatic detection of web weak passwords and universal passwords. It generates password dictionaries based on PCFG algorithm, proposes to judge blasting result via 4 steps with traditional static keyword features and dynamic page feature information. Then the blasting failure events are discriminated and the usernames are blasted based on response time. Thereafter the weak password dictionary is dynamically adjusted according to the hints provided by the response failure page. Based on the algorithm, this paper implements a detection system named WebCrack. Experimental results of two blasting tests on DedeCMS and Discuz! systems as well as a random backend test show that the proposed method can detect weak passwords and universal passwords of various web systems with an average accuracy rate of about 93.75%, providing security advisories for users' password settings with strong practicability.
[ { "created": "Wed, 26 Oct 2022 09:34:41 GMT", "version": "v1" } ]
2022-10-27
[ [ "Long", "Xiang", "" ], [ "Huang", "Yan", "" ], [ "Liu", "Zhendong", "" ], [ "Han", "Lansheng", "" ], [ "Sun", "Haili", "" ], [ "He", "Jingyuan", "" ] ]
The feature diversity of different web systems in page elements, submission contents and return information makes it difficult to detect weak password automatically. To solve this problem, multi-factor correlation detection method as integrated in the DBKER algorithm is proposed to achieve automatic detection of web weak passwords and universal passwords. It generates password dictionaries based on PCFG algorithm, proposes to judge blasting result via 4 steps with traditional static keyword features and dynamic page feature information. Then the blasting failure events are discriminated and the usernames are blasted based on response time. Thereafter the weak password dictionary is dynamically adjusted according to the hints provided by the response failure page. Based on the algorithm, this paper implements a detection system named WebCrack. Experimental results of two blasting tests on DedeCMS and Discuz! systems as well as a random backend test show that the proposed method can detect weak passwords and universal passwords of various web systems with an average accuracy rate of about 93.75%, providing security advisories for users' password settings with strong practicability.
1703.07534
Dong Liu
Jingxian Zhang and Dong Liu
Visual Analyses of Music History: A User-Centric Approach
null
null
null
null
cs.HC
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Music history, referring to the records of users' listening or downloading history in online music services, is the primary source for music service providers to analyze users' preferences on music and thus to provide personalized recommendations to users. In order to engage users into the service and to improve user experience, it would be beneficial to provide visual analyses of one user's music history as well as visualized recommendations to that user. In this paper, we take a user-centric approach to the design of such visual analyses. We start by investigating user needs on such visual analyses and recommendations, then propose several different visualization schemes, and perform a pilot study to collect user feedback on the designed schemes. We further conduct user studies to verify the utility of the proposed schemes, and the results not only demonstrate the effectiveness of our proposed visualization, but also provide important insights to guide the visualization design in the future.
[ { "created": "Wed, 22 Mar 2017 05:37:19 GMT", "version": "v1" } ]
2017-03-23
[ [ "Zhang", "Jingxian", "" ], [ "Liu", "Dong", "" ] ]
Music history, referring to the records of users' listening or downloading history in online music services, is the primary source for music service providers to analyze users' preferences on music and thus to provide personalized recommendations to users. In order to engage users into the service and to improve user experience, it would be beneficial to provide visual analyses of one user's music history as well as visualized recommendations to that user. In this paper, we take a user-centric approach to the design of such visual analyses. We start by investigating user needs on such visual analyses and recommendations, then propose several different visualization schemes, and perform a pilot study to collect user feedback on the designed schemes. We further conduct user studies to verify the utility of the proposed schemes, and the results not only demonstrate the effectiveness of our proposed visualization, but also provide important insights to guide the visualization design in the future.
2403.11752
Vigneshwaran Shankaran
Aditya Narayan Sankaran, Vigneshwaran Shankaran, Sampath Lonka, Rajesh Sharma
Revisiting The Classics: A Study on Identifying and Rectifying Gender Stereotypes in Rhymes and Poems
Accepted to appear at LREC-COLING 2024
null
null
null
cs.CL
http://creativecommons.org/licenses/by/4.0/
Rhymes and poems are a powerful medium for transmitting cultural norms and societal roles. However, the pervasive existence of gender stereotypes in these works perpetuates biased perceptions and limits the scope of individuals' identities. Past works have shown that stereotyping and prejudice emerge in early childhood, and developmental research on causal mechanisms is critical for understanding and controlling stereotyping and prejudice. This work contributes by gathering a dataset of rhymes and poems to identify gender stereotypes and propose a model with 97% accuracy to identify gender bias. Gender stereotypes were rectified using a Large Language Model (LLM) and its effectiveness was evaluated in a comparative survey against human educator rectifications. To summarize, this work highlights the pervasive nature of gender stereotypes in literary works and reveals the potential of LLMs to rectify gender stereotypes. This study raises awareness and promotes inclusivity within artistic expressions, making a significant contribution to the discourse on gender equality.
[ { "created": "Mon, 18 Mar 2024 13:02:02 GMT", "version": "v1" }, { "created": "Mon, 25 Mar 2024 16:33:12 GMT", "version": "v2" } ]
2024-03-26
[ [ "Sankaran", "Aditya Narayan", "" ], [ "Shankaran", "Vigneshwaran", "" ], [ "Lonka", "Sampath", "" ], [ "Sharma", "Rajesh", "" ] ]
Rhymes and poems are a powerful medium for transmitting cultural norms and societal roles. However, the pervasive existence of gender stereotypes in these works perpetuates biased perceptions and limits the scope of individuals' identities. Past works have shown that stereotyping and prejudice emerge in early childhood, and developmental research on causal mechanisms is critical for understanding and controlling stereotyping and prejudice. This work contributes by gathering a dataset of rhymes and poems to identify gender stereotypes and propose a model with 97% accuracy to identify gender bias. Gender stereotypes were rectified using a Large Language Model (LLM) and its effectiveness was evaluated in a comparative survey against human educator rectifications. To summarize, this work highlights the pervasive nature of gender stereotypes in literary works and reveals the potential of LLMs to rectify gender stereotypes. This study raises awareness and promotes inclusivity within artistic expressions, making a significant contribution to the discourse on gender equality.
1608.08505
Carla Binucci
Carla Binucci, Markus Chimani, Walter Didimo, Giuseppe Liotta, Fabrizio Montecchiani
Placing Arrows in Directed Graph Drawings
Appears in the Proceedings of the 24th International Symposium on Graph Drawing and Network Visualization (GD 2016)
null
null
null
cs.DS
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We consider the problem of placing arrow heads in directed graph drawings without them overlapping other drawn objects. This gives drawings where edge directions can be deduced unambiguously. We show hardness of the problem, present exact and heuristic algorithms, and report on a practical study.
[ { "created": "Tue, 30 Aug 2016 15:33:56 GMT", "version": "v1" } ]
2016-08-31
[ [ "Binucci", "Carla", "" ], [ "Chimani", "Markus", "" ], [ "Didimo", "Walter", "" ], [ "Liotta", "Giuseppe", "" ], [ "Montecchiani", "Fabrizio", "" ] ]
We consider the problem of placing arrow heads in directed graph drawings without them overlapping other drawn objects. This gives drawings where edge directions can be deduced unambiguously. We show hardness of the problem, present exact and heuristic algorithms, and report on a practical study.
2308.09604
Xiaokang Pan
Jin Liu, Xiaokang Pan, Junwen Duan, Hongdong Li, Youqi Li, Zhe Qu
Faster Stochastic Variance Reduction Methods for Compositional MiniMax Optimization
null
null
null
null
cs.LG math.OC
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This paper delves into the realm of stochastic optimization for compositional minimax optimization - a pivotal challenge across various machine learning domains, including deep AUC and reinforcement learning policy evaluation. Despite its significance, the problem of compositional minimax optimization is still under-explored. Adding to the complexity, current methods of compositional minimax optimization are plagued by sub-optimal complexities or heavy reliance on sizable batch sizes. To respond to these constraints, this paper introduces a novel method, called Nested STOchastic Recursive Momentum (NSTORM), which can achieve the optimal sample complexity of $O(\kappa^3 /\epsilon^3 )$ to obtain the $\epsilon$-accuracy solution. We also demonstrate that NSTORM can achieve the same sample complexity under the Polyak-\L ojasiewicz (PL)-condition - an insightful extension of its capabilities. Yet, NSTORM encounters an issue with its requirement for low learning rates, potentially constraining its real-world applicability in machine learning. To overcome this hurdle, we present ADAptive NSTORM (ADA-NSTORM) with adaptive learning rates. We demonstrate that ADA-NSTORM can achieve the same sample complexity but the experimental results show its more effectiveness. All the proposed complexities indicate that our proposed methods can match lower bounds to existing minimax optimizations, without requiring a large batch size in each iteration. Extensive experiments support the efficiency of our proposed methods.
[ { "created": "Fri, 18 Aug 2023 14:57:21 GMT", "version": "v1" }, { "created": "Tue, 12 Dec 2023 05:28:51 GMT", "version": "v2" } ]
2023-12-13
[ [ "Liu", "Jin", "" ], [ "Pan", "Xiaokang", "" ], [ "Duan", "Junwen", "" ], [ "Li", "Hongdong", "" ], [ "Li", "Youqi", "" ], [ "Qu", "Zhe", "" ] ]
This paper delves into the realm of stochastic optimization for compositional minimax optimization - a pivotal challenge across various machine learning domains, including deep AUC and reinforcement learning policy evaluation. Despite its significance, the problem of compositional minimax optimization is still under-explored. Adding to the complexity, current methods of compositional minimax optimization are plagued by sub-optimal complexities or heavy reliance on sizable batch sizes. To respond to these constraints, this paper introduces a novel method, called Nested STOchastic Recursive Momentum (NSTORM), which can achieve the optimal sample complexity of $O(\kappa^3 /\epsilon^3 )$ to obtain the $\epsilon$-accuracy solution. We also demonstrate that NSTORM can achieve the same sample complexity under the Polyak-\L ojasiewicz (PL)-condition - an insightful extension of its capabilities. Yet, NSTORM encounters an issue with its requirement for low learning rates, potentially constraining its real-world applicability in machine learning. To overcome this hurdle, we present ADAptive NSTORM (ADA-NSTORM) with adaptive learning rates. We demonstrate that ADA-NSTORM can achieve the same sample complexity but the experimental results show its more effectiveness. All the proposed complexities indicate that our proposed methods can match lower bounds to existing minimax optimizations, without requiring a large batch size in each iteration. Extensive experiments support the efficiency of our proposed methods.
2208.12850
Michael Baddeley Dr
Michael Baddeley, Yevgen Gyl, Markus Schuss, Xiaoyuan Ma, and Carlo Alberto Boano
OSF: An Open-Source Framework for Synchronous Flooding over Multiple Physical Layers
null
null
null
null
cs.NI
http://creativecommons.org/licenses/by/4.0/
Flooding protocols based on concurrent transmissions are regarded as the most reliable way to collect or disseminate data across a multi-hop low-power wireless mesh network. Recent works have shown that such protocols are effective for narrowband communication not only over IEEE 802.15.4, but also over the BLE 5 physical layers (PHYs). However, to date, existing literature has only built synchronous flooding solutions on top of a single PHY, and there has been no attempt to leverage different PHYs at runtime to increase performance. This paper fills this gap and presents OSF, an open-source framework that enables the design of multi-PHY synchronous flooding solutions thanks to a novel radio driver and middle-ware architecture capable of dynamically switching the underlying physical layer. This allows exploitation of the specific benefits of each PHY (e.g., higher data-rate, increased robustness) on-demand during each flood, increasing performance. We tailor OSF to the off-the-shelf nRF52840 platform, and showcase its benefits by comparing single-PHY and multi-PHY synchronous flooding solutions on a real-world testbed.
[ { "created": "Fri, 26 Aug 2022 19:40:29 GMT", "version": "v1" } ]
2022-08-30
[ [ "Baddeley", "Michael", "" ], [ "Gyl", "Yevgen", "" ], [ "Schuss", "Markus", "" ], [ "Ma", "Xiaoyuan", "" ], [ "Boano", "Carlo Alberto", "" ] ]
Flooding protocols based on concurrent transmissions are regarded as the most reliable way to collect or disseminate data across a multi-hop low-power wireless mesh network. Recent works have shown that such protocols are effective for narrowband communication not only over IEEE 802.15.4, but also over the BLE 5 physical layers (PHYs). However, to date, existing literature has only built synchronous flooding solutions on top of a single PHY, and there has been no attempt to leverage different PHYs at runtime to increase performance. This paper fills this gap and presents OSF, an open-source framework that enables the design of multi-PHY synchronous flooding solutions thanks to a novel radio driver and middle-ware architecture capable of dynamically switching the underlying physical layer. This allows exploitation of the specific benefits of each PHY (e.g., higher data-rate, increased robustness) on-demand during each flood, increasing performance. We tailor OSF to the off-the-shelf nRF52840 platform, and showcase its benefits by comparing single-PHY and multi-PHY synchronous flooding solutions on a real-world testbed.
2210.09604
Xiaoning Liu
Xiaoning Liu
Perceptual Multi-Exposure Fusion
The current version is our previous work rejected by IEEE TMM. I'm very sorry and I want to withdraw this submitted version. I will resubmit it when I improve it in the future. The version involves some ideas we are doing
null
null
null
cs.CV eess.IV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
As an ever-increasing demand for high dynamic range (HDR) scene shooting, multi-exposure image fusion (MEF) technology has abounded. In recent years, multi-scale exposure fusion approaches based on detail-enhancement have led the way for improvement in highlight and shadow details. Most of such methods, however, are too computationally expensive to be deployed on mobile devices. This paper presents a perceptual multi-exposure fusion method that not just ensures fine shadow/highlight details but with lower complexity than detailenhanced methods. We analyze the potential defects of three classical exposure measures in lieu of using detail-enhancement component and improve two of them, namely adaptive Wellexposedness (AWE) and the gradient of color images (3-D gradient). AWE designed in YCbCr color space considers the difference between varying exposure images. 3-D gradient is employed to extract fine details. We build a large-scale multiexposure benchmark dataset suitable for static scenes, which contains 167 image sequences all told. Experiments on the constructed dataset demonstrate that the proposed method exceeds existing eight state-of-the-art approaches in terms of visually and MEF-SSIM value. Moreover, our approach can achieve a better improvement for current image enhancement techniques, ensuring fine detail in bright light.
[ { "created": "Tue, 18 Oct 2022 05:34:58 GMT", "version": "v1" }, { "created": "Wed, 19 Oct 2022 06:58:48 GMT", "version": "v2" } ]
2022-10-20
[ [ "Liu", "Xiaoning", "" ] ]
As an ever-increasing demand for high dynamic range (HDR) scene shooting, multi-exposure image fusion (MEF) technology has abounded. In recent years, multi-scale exposure fusion approaches based on detail-enhancement have led the way for improvement in highlight and shadow details. Most of such methods, however, are too computationally expensive to be deployed on mobile devices. This paper presents a perceptual multi-exposure fusion method that not just ensures fine shadow/highlight details but with lower complexity than detailenhanced methods. We analyze the potential defects of three classical exposure measures in lieu of using detail-enhancement component and improve two of them, namely adaptive Wellexposedness (AWE) and the gradient of color images (3-D gradient). AWE designed in YCbCr color space considers the difference between varying exposure images. 3-D gradient is employed to extract fine details. We build a large-scale multiexposure benchmark dataset suitable for static scenes, which contains 167 image sequences all told. Experiments on the constructed dataset demonstrate that the proposed method exceeds existing eight state-of-the-art approaches in terms of visually and MEF-SSIM value. Moreover, our approach can achieve a better improvement for current image enhancement techniques, ensuring fine detail in bright light.
1904.01782
Rui Liu
Rui Liu, Yu Liu, Xinyu Gong, Xiaogang Wang, Hongsheng Li
Conditional Adversarial Generative Flow for Controllable Image Synthesis
Accepted by CVPR 2019
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Flow-based generative models show great potential in image synthesis due to its reversible pipeline and exact log-likelihood target, yet it suffers from weak ability for conditional image synthesis, especially for multi-label or unaware conditions. This is because the potential distribution of image conditions is hard to measure precisely from its latent variable $z$. In this paper, based on modeling a joint probabilistic density of an image and its conditions, we propose a novel flow-based generative model named conditional adversarial generative flow (CAGlow). Instead of disentangling attributes from latent space, we blaze a new trail for learning an encoder to estimate the mapping from condition space to latent space in an adversarial manner. Given a specific condition $c$, CAGlow can encode it to a sampled $z$, and then enable robust conditional image synthesis in complex situations like combining person identity with multiple attributes. The proposed CAGlow can be implemented in both supervised and unsupervised manners, thus can synthesize images with conditional information like categories, attributes, and even some unknown properties. Extensive experiments show that CAGlow ensures the independence of different conditions and outperforms regular Glow to a significant extent.
[ { "created": "Wed, 3 Apr 2019 05:58:01 GMT", "version": "v1" } ]
2019-04-04
[ [ "Liu", "Rui", "" ], [ "Liu", "Yu", "" ], [ "Gong", "Xinyu", "" ], [ "Wang", "Xiaogang", "" ], [ "Li", "Hongsheng", "" ] ]
Flow-based generative models show great potential in image synthesis due to its reversible pipeline and exact log-likelihood target, yet it suffers from weak ability for conditional image synthesis, especially for multi-label or unaware conditions. This is because the potential distribution of image conditions is hard to measure precisely from its latent variable $z$. In this paper, based on modeling a joint probabilistic density of an image and its conditions, we propose a novel flow-based generative model named conditional adversarial generative flow (CAGlow). Instead of disentangling attributes from latent space, we blaze a new trail for learning an encoder to estimate the mapping from condition space to latent space in an adversarial manner. Given a specific condition $c$, CAGlow can encode it to a sampled $z$, and then enable robust conditional image synthesis in complex situations like combining person identity with multiple attributes. The proposed CAGlow can be implemented in both supervised and unsupervised manners, thus can synthesize images with conditional information like categories, attributes, and even some unknown properties. Extensive experiments show that CAGlow ensures the independence of different conditions and outperforms regular Glow to a significant extent.
1511.08063
Julien Mineraud
Julien Mineraud and Sasu Tarkoma
Toward interoperability for the Internet of Things with meta-hubs
7 pages, 4 figures
null
null
null
cs.SE
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The Internet of Things (IoT) envisions that objects may be connected to the Internet, producing and consuming data in real-time. Today, numerous middleware platforms are available to facilitate the communication with these objects. Unfortunately, the interoperability of these platforms is very limited because it requires to "manually" connect the services proposed by each platform. One key design goal for our contribution is not to build yet another middleware, but rather to augment the functionalities of existing systems via an extension to support their integration into a network of heterogeneous IoT hubs. The extension includes a RESTful API to manipulate the basic component of our extension, the IoT feeds. The IoT feeds allow the platform's owner to dynamically marshal the IoT features connected to the platform, as well as the data that they produce. Furthermore, the feeds enable the owner to manage and control the data flows before connecting them to his applications. Subsequently, these feeds may also be published to meta-hubs in order to expose them to third parties. We evaluated an implementation our extension for Android systems to show the feasibility of managing the data flows using the RESTful API on this platform.
[ { "created": "Wed, 25 Nov 2015 13:57:08 GMT", "version": "v1" } ]
2015-11-26
[ [ "Mineraud", "Julien", "" ], [ "Tarkoma", "Sasu", "" ] ]
The Internet of Things (IoT) envisions that objects may be connected to the Internet, producing and consuming data in real-time. Today, numerous middleware platforms are available to facilitate the communication with these objects. Unfortunately, the interoperability of these platforms is very limited because it requires to "manually" connect the services proposed by each platform. One key design goal for our contribution is not to build yet another middleware, but rather to augment the functionalities of existing systems via an extension to support their integration into a network of heterogeneous IoT hubs. The extension includes a RESTful API to manipulate the basic component of our extension, the IoT feeds. The IoT feeds allow the platform's owner to dynamically marshal the IoT features connected to the platform, as well as the data that they produce. Furthermore, the feeds enable the owner to manage and control the data flows before connecting them to his applications. Subsequently, these feeds may also be published to meta-hubs in order to expose them to third parties. We evaluated an implementation our extension for Android systems to show the feasibility of managing the data flows using the RESTful API on this platform.
2203.10579
\`Alex R. Atrio
\`Alex R. Atrio, Andrei Popescu-Belis
Small Batch Sizes Improve Training of Low-Resource Neural MT
To be published in 18th International Conference on Natural Language Processing (ICON 2021)
null
null
null
cs.CL cs.AI cs.LG
http://creativecommons.org/licenses/by/4.0/
We study the role of an essential hyper-parameter that governs the training of Transformers for neural machine translation in a low-resource setting: the batch size. Using theoretical insights and experimental evidence, we argue against the widespread belief that batch size should be set as large as allowed by the memory of the GPUs. We show that in a low-resource setting, a smaller batch size leads to higher scores in a shorter training time, and argue that this is due to better regularization of the gradients during training.
[ { "created": "Sun, 20 Mar 2022 15:14:39 GMT", "version": "v1" } ]
2022-03-22
[ [ "Atrio", "Àlex R.", "" ], [ "Popescu-Belis", "Andrei", "" ] ]
We study the role of an essential hyper-parameter that governs the training of Transformers for neural machine translation in a low-resource setting: the batch size. Using theoretical insights and experimental evidence, we argue against the widespread belief that batch size should be set as large as allowed by the memory of the GPUs. We show that in a low-resource setting, a smaller batch size leads to higher scores in a shorter training time, and argue that this is due to better regularization of the gradients during training.
0805.4323
Martin Hoefer
Ulrik Brandes, Martin Hoefer, Bobo Nick
Network Connection Games with Disconnected Equilibria
18 pages, 4 figures, extended abstract in WINE 2008
null
null
null
cs.GT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In this paper we extend a popular non-cooperative network creation game (NCG) to allow for disconnected equilibrium networks. There are n players, each is a vertex in a graph, and a strategy is a subset of players to build edges to. For each edge a player must pay a cost \alpha, and the individual cost for a player represents a trade-off between edge costs and shortest path lengths to all other players. We extend the model to a penalized game (PCG), for which we reduce the penalty counted towards the individual cost for a pair of disconnected players to a finite value \beta. Our analysis concentrates on existence, structure, and cost of disconnected Nash and strong equilibria. Although the PCG is not a potential game, pure Nash equilibria always and pure strong equilibria very often exist. We provide tight conditions under which disconnected Nash (strong) equilibria can evolve. Components of these equilibria must be Nash (strong) equilibria of a smaller NCG. However, in contrast to the NCG, for almost all parameter values no tree is a stable component. Finally, we present a detailed characterization of the price of anarchy that reveals cases in which the price of anarchy is \Theta(n) and thus several orders of magnitude larger than in the NCG. Perhaps surprisingly, the strong price of anarchy increases to at most 4. This indicates that global communication and coordination can be extremely valuable to overcome socially inferior topologies in distributed selfish network design.
[ { "created": "Wed, 28 May 2008 12:09:15 GMT", "version": "v1" }, { "created": "Mon, 27 Oct 2008 22:12:35 GMT", "version": "v2" } ]
2008-10-28
[ [ "Brandes", "Ulrik", "" ], [ "Hoefer", "Martin", "" ], [ "Nick", "Bobo", "" ] ]
In this paper we extend a popular non-cooperative network creation game (NCG) to allow for disconnected equilibrium networks. There are n players, each is a vertex in a graph, and a strategy is a subset of players to build edges to. For each edge a player must pay a cost \alpha, and the individual cost for a player represents a trade-off between edge costs and shortest path lengths to all other players. We extend the model to a penalized game (PCG), for which we reduce the penalty counted towards the individual cost for a pair of disconnected players to a finite value \beta. Our analysis concentrates on existence, structure, and cost of disconnected Nash and strong equilibria. Although the PCG is not a potential game, pure Nash equilibria always and pure strong equilibria very often exist. We provide tight conditions under which disconnected Nash (strong) equilibria can evolve. Components of these equilibria must be Nash (strong) equilibria of a smaller NCG. However, in contrast to the NCG, for almost all parameter values no tree is a stable component. Finally, we present a detailed characterization of the price of anarchy that reveals cases in which the price of anarchy is \Theta(n) and thus several orders of magnitude larger than in the NCG. Perhaps surprisingly, the strong price of anarchy increases to at most 4. This indicates that global communication and coordination can be extremely valuable to overcome socially inferior topologies in distributed selfish network design.
1903.01905
Bernhard Kainz
Daniel Grzech, Lo\"ic le Folgoc, Mattias P. Heinrich, Bishesh Khanal, Jakub Moll, Julia A. Schnabel, Ben Glocker, Bernhard Kainz
FastReg: Fast Non-Rigid Registration via Accelerated Optimisation on the Manifold of Diffeomorphisms
There is an ongoing dispute about the presentation of this paper. It will be withdrawn until the dispute is resoved
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We present an implementation of a new approach to diffeomorphic non-rigid registration of medical images. The method is based on optical flow and warps images via gradient flow with the standard $L^2$ inner product. To compute the transformation, we rely on accelerated optimisation on the manifold of diffeomorphisms. We achieve regularity properties of Sobolev gradient flows, which are expensive to compute, owing to a novel method of averaging the gradients in time rather than space. We successfully register brain MRI and challenging abdominal CT scans at speeds orders of magnitude faster than previous approaches. We make our code available in a public repository: https://github.com/dgrzech/fastreg
[ { "created": "Tue, 5 Mar 2019 15:41:47 GMT", "version": "v1" }, { "created": "Tue, 23 Apr 2019 15:37:43 GMT", "version": "v2" }, { "created": "Wed, 24 Apr 2019 10:02:27 GMT", "version": "v3" } ]
2019-04-25
[ [ "Grzech", "Daniel", "" ], [ "Folgoc", "Loïc le", "" ], [ "Heinrich", "Mattias P.", "" ], [ "Khanal", "Bishesh", "" ], [ "Moll", "Jakub", "" ], [ "Schnabel", "Julia A.", "" ], [ "Glocker", "Ben", "" ], [ "Kainz", "Bernhard", "" ] ]
We present an implementation of a new approach to diffeomorphic non-rigid registration of medical images. The method is based on optical flow and warps images via gradient flow with the standard $L^2$ inner product. To compute the transformation, we rely on accelerated optimisation on the manifold of diffeomorphisms. We achieve regularity properties of Sobolev gradient flows, which are expensive to compute, owing to a novel method of averaging the gradients in time rather than space. We successfully register brain MRI and challenging abdominal CT scans at speeds orders of magnitude faster than previous approaches. We make our code available in a public repository: https://github.com/dgrzech/fastreg
2105.02318
Kishor Jothimurugan
Kishor Jothimurugan, Matthew Andrews, Jeongran Lee and Lorenzo Maggi
Learning Algorithms for Regenerative Stopping Problems with Applications to Shipping Consolidation in Logistics
null
null
null
null
cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We study regenerative stopping problems in which the system starts anew whenever the controller decides to stop and the long-term average cost is to be minimized. Traditional model-based solutions involve estimating the underlying process from data and computing strategies for the estimated model. In this paper, we compare such solutions to deep reinforcement learning and imitation learning which involve learning a neural network policy from simulations. We evaluate the different approaches on a real-world problem of shipping consolidation in logistics and demonstrate that deep learning can be effectively used to solve such problems.
[ { "created": "Wed, 5 May 2021 20:45:46 GMT", "version": "v1" } ]
2021-05-07
[ [ "Jothimurugan", "Kishor", "" ], [ "Andrews", "Matthew", "" ], [ "Lee", "Jeongran", "" ], [ "Maggi", "Lorenzo", "" ] ]
We study regenerative stopping problems in which the system starts anew whenever the controller decides to stop and the long-term average cost is to be minimized. Traditional model-based solutions involve estimating the underlying process from data and computing strategies for the estimated model. In this paper, we compare such solutions to deep reinforcement learning and imitation learning which involve learning a neural network policy from simulations. We evaluate the different approaches on a real-world problem of shipping consolidation in logistics and demonstrate that deep learning can be effectively used to solve such problems.
1410.8127
Ali Soltani Tehrani
Ali Soltani Tehrani, Jessica Chani, Thomas Eriksson, and Christian Fager
Investigation of Parameter Adaptation in RF Power Amplifier Behavioral Models
null
null
null
null
cs.SY
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This paper presents an investigation into parameter adaptation in behavioral model--based digital predistortion for radio frequency power amplifiers. A novel measurement setup framework that emulates real--time adaptation in transmitters is developed that allows evaluation of different parameters, configurations and adaptation algorithms. This setup relieves the need for full feedback loops for parameter adaptation while providing the flexibility needed in the design process of parameter adaptation. Issues such as convergence speed, sensitivity to quantization noise in the feedback loop and predistortion performance are investigated for some different parameter update algorithms using the proposed measurement setup. The approach presented in this paper allows the possibility to analyze different aspects of digital predistortion adaptation algorithms, and is an important enabling step for further research on parameter adaptation before the real--time hardware is implemented for use.
[ { "created": "Wed, 29 Oct 2014 19:59:34 GMT", "version": "v1" } ]
2014-10-30
[ [ "Tehrani", "Ali Soltani", "" ], [ "Chani", "Jessica", "" ], [ "Eriksson", "Thomas", "" ], [ "Fager", "Christian", "" ] ]
This paper presents an investigation into parameter adaptation in behavioral model--based digital predistortion for radio frequency power amplifiers. A novel measurement setup framework that emulates real--time adaptation in transmitters is developed that allows evaluation of different parameters, configurations and adaptation algorithms. This setup relieves the need for full feedback loops for parameter adaptation while providing the flexibility needed in the design process of parameter adaptation. Issues such as convergence speed, sensitivity to quantization noise in the feedback loop and predistortion performance are investigated for some different parameter update algorithms using the proposed measurement setup. The approach presented in this paper allows the possibility to analyze different aspects of digital predistortion adaptation algorithms, and is an important enabling step for further research on parameter adaptation before the real--time hardware is implemented for use.
2211.13577
Cheng Feng
Cheng Feng and Pingge Hu
Learning Invariant Rules from Data for Interpretable Anomaly Detection
null
null
null
null
cs.LG cs.AI
http://creativecommons.org/licenses/by-nc-sa/4.0/
In the research area of anomaly detection, novel and promising methods are frequently developed. However, most existing studies exclusively focus on the detection task only and ignore the interpretability of the underlying models as well as their detection results. Nevertheless, anomaly interpretation, which aims to provide explanation of why specific data instances are identified as anomalies, is an equally important task in many real-world applications. In this work, we propose a novel framework which synergizes several machine learning and data mining techniques to automatically learn invariant rules that are consistently satisfied in a given dataset. The learned invariant rules can provide explicit explanation of anomaly detection results in the inference phase and thus are extremely useful for subsequent decision-making regarding reported anomalies. Furthermore, our empirical evaluation shows that the proposed method can also achieve comparable or even better performance in terms of AUC and partial AUC on public benchmark datasets across various application domains compared with start-of-the-art anomaly detection models.
[ { "created": "Thu, 24 Nov 2022 13:03:20 GMT", "version": "v1" }, { "created": "Fri, 6 Jan 2023 03:51:28 GMT", "version": "v2" }, { "created": "Fri, 13 Jan 2023 03:57:35 GMT", "version": "v3" } ]
2023-01-16
[ [ "Feng", "Cheng", "" ], [ "Hu", "Pingge", "" ] ]
In the research area of anomaly detection, novel and promising methods are frequently developed. However, most existing studies exclusively focus on the detection task only and ignore the interpretability of the underlying models as well as their detection results. Nevertheless, anomaly interpretation, which aims to provide explanation of why specific data instances are identified as anomalies, is an equally important task in many real-world applications. In this work, we propose a novel framework which synergizes several machine learning and data mining techniques to automatically learn invariant rules that are consistently satisfied in a given dataset. The learned invariant rules can provide explicit explanation of anomaly detection results in the inference phase and thus are extremely useful for subsequent decision-making regarding reported anomalies. Furthermore, our empirical evaluation shows that the proposed method can also achieve comparable or even better performance in terms of AUC and partial AUC on public benchmark datasets across various application domains compared with start-of-the-art anomaly detection models.
1709.05861
Narotam Singh
Narotam Singh (1), Nittin Singh (1), Abhinav Dhall (1) ((1) Indian Institute of Technology Ropar)
Continuous Multimodal Emotion Recognition Approach for AVEC 2017
4 pages, 3 figures, arXiv:1605.06778, arXiv:1512.03385
null
null
null
cs.CV cs.LG cs.MM
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This paper reports the analysis of audio and visual features in predicting the continuous emotion dimensions under the seventh Audio/Visual Emotion Challenge (AVEC 2017), which was done as part of a B.Tech. 2nd year internship project. For visual features we used the HOG (Histogram of Gradients) features, Fisher encodings of SIFT (Scale-Invariant Feature Transform) features based on Gaussian mixture model (GMM) and some pretrained Convolutional Neural Network layers as features; all these extracted for each video clip. For audio features we used the Bag-of-audio-words (BoAW) representation of the LLDs (low-level descriptors) generated by openXBOW provided by the organisers of the event. Then we trained fully connected neural network regression model on the dataset for all these different modalities. We applied multimodal fusion on the output models to get the Concordance correlation coefficient on Development set as well as Test set.
[ { "created": "Mon, 18 Sep 2017 11:01:43 GMT", "version": "v1" }, { "created": "Tue, 24 Oct 2017 12:08:09 GMT", "version": "v2" } ]
2017-10-25
[ [ "Singh", "Narotam", "" ], [ "Singh", "Nittin", "" ], [ "Dhall", "Abhinav", "" ] ]
This paper reports the analysis of audio and visual features in predicting the continuous emotion dimensions under the seventh Audio/Visual Emotion Challenge (AVEC 2017), which was done as part of a B.Tech. 2nd year internship project. For visual features we used the HOG (Histogram of Gradients) features, Fisher encodings of SIFT (Scale-Invariant Feature Transform) features based on Gaussian mixture model (GMM) and some pretrained Convolutional Neural Network layers as features; all these extracted for each video clip. For audio features we used the Bag-of-audio-words (BoAW) representation of the LLDs (low-level descriptors) generated by openXBOW provided by the organisers of the event. Then we trained fully connected neural network regression model on the dataset for all these different modalities. We applied multimodal fusion on the output models to get the Concordance correlation coefficient on Development set as well as Test set.
1908.10149
Michael Barz
Michael Barz and Daniel Sonntag
Incremental Improvement of a Question Answering System by Re-ranking Answer Candidates using Machine Learning
Accepted for oral presentation at tenth International Workshop on Spoken Dialogue Systems Technology (IWSDS) 2019
null
10.1007/978-981-15-9323-9_34
null
cs.LG cs.CL cs.IR stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We implement a method for re-ranking top-10 results of a state-of-the-art question answering (QA) system. The goal of our re-ranking approach is to improve the answer selection given the user question and the top-10 candidates. We focus on improving deployed QA systems that do not allow re-training or re-training comes at a high cost. Our re-ranking approach learns a similarity function using n-gram based features using the query, the answer and the initial system confidence as input. Our contributions are: (1) we generate a QA training corpus starting from 877 answers from the customer care domain of T-Mobile Austria, (2) we implement a state-of-the-art QA pipeline using neural sentence embeddings that encode queries in the same space than the answer index, and (3) we evaluate the QA pipeline and our re-ranking approach using a separately provided test set. The test set can be considered to be available after deployment of the system, e.g., based on feedback of users. Our results show that the system performance, in terms of top-n accuracy and the mean reciprocal rank, benefits from re-ranking using gradient boosted regression trees. On average, the mean reciprocal rank improves by 9.15%.
[ { "created": "Tue, 27 Aug 2019 11:54:23 GMT", "version": "v1" } ]
2021-06-17
[ [ "Barz", "Michael", "" ], [ "Sonntag", "Daniel", "" ] ]
We implement a method for re-ranking top-10 results of a state-of-the-art question answering (QA) system. The goal of our re-ranking approach is to improve the answer selection given the user question and the top-10 candidates. We focus on improving deployed QA systems that do not allow re-training or re-training comes at a high cost. Our re-ranking approach learns a similarity function using n-gram based features using the query, the answer and the initial system confidence as input. Our contributions are: (1) we generate a QA training corpus starting from 877 answers from the customer care domain of T-Mobile Austria, (2) we implement a state-of-the-art QA pipeline using neural sentence embeddings that encode queries in the same space than the answer index, and (3) we evaluate the QA pipeline and our re-ranking approach using a separately provided test set. The test set can be considered to be available after deployment of the system, e.g., based on feedback of users. Our results show that the system performance, in terms of top-n accuracy and the mean reciprocal rank, benefits from re-ranking using gradient boosted regression trees. On average, the mean reciprocal rank improves by 9.15%.
2308.11373
Zesen Liu
Zesen Liu, Meng Guo, Weimin Bao and Zhongkui Li
Fast and Adaptive Multi-agent Planning under Collaborative Temporal Logic Tasks via Poset Products
16 pages, 9 figures
null
null
null
cs.RO
http://creativecommons.org/licenses/by-nc-nd/4.0/
Efficient coordination and planning is essential for large-scale multi-agent systems that collaborate in a shared dynamic environment. Heuristic search methods or learning-based approaches often lack the guarantee on correctness and performance. Moreover, when the collaborative tasks contain both spatial and temporal requirements, e.g., as Linear Temporal Logic (LTL) formulas, formal methods provide a verifiable framework for task planning. However, since the planning complexity grows exponentially with the number of agents and the length of the task formula, existing studies are mostly limited to small artificial cases. To address this issue, a new planning paradigm is proposed in this work for system-wide temporal task formulas that are released online and continually. It avoids two common bottlenecks in the traditional methods, i.e., (i) the direct translation of the complete task formula to the associated B\"uchi automaton; and (ii) the synchronized product between the B\"uchi automaton and the transition models of all agents. Instead, an adaptive planning algorithm is proposed that computes the product of relaxed partially-ordered sets (R-posets) on-the-fly, and assigns these subtasks to the agents subject to the ordering constraints. It is shown that the first valid plan can be derived with a polynomial time and memory complexity w.r.t. the system size and the formula length. Our method can take into account task formulas with a length of more than 400 and a fleet with more than $400$ agents, while most existing methods fail at the formula length of 25 within a reasonable duration. The proposed method is validated on large fleets of service robots in both simulation and hardware experiments.
[ { "created": "Tue, 22 Aug 2023 11:56:15 GMT", "version": "v1" }, { "created": "Tue, 9 Apr 2024 14:01:16 GMT", "version": "v2" } ]
2024-04-10
[ [ "Liu", "Zesen", "" ], [ "Guo", "Meng", "" ], [ "Bao", "Weimin", "" ], [ "Li", "Zhongkui", "" ] ]
Efficient coordination and planning is essential for large-scale multi-agent systems that collaborate in a shared dynamic environment. Heuristic search methods or learning-based approaches often lack the guarantee on correctness and performance. Moreover, when the collaborative tasks contain both spatial and temporal requirements, e.g., as Linear Temporal Logic (LTL) formulas, formal methods provide a verifiable framework for task planning. However, since the planning complexity grows exponentially with the number of agents and the length of the task formula, existing studies are mostly limited to small artificial cases. To address this issue, a new planning paradigm is proposed in this work for system-wide temporal task formulas that are released online and continually. It avoids two common bottlenecks in the traditional methods, i.e., (i) the direct translation of the complete task formula to the associated B\"uchi automaton; and (ii) the synchronized product between the B\"uchi automaton and the transition models of all agents. Instead, an adaptive planning algorithm is proposed that computes the product of relaxed partially-ordered sets (R-posets) on-the-fly, and assigns these subtasks to the agents subject to the ordering constraints. It is shown that the first valid plan can be derived with a polynomial time and memory complexity w.r.t. the system size and the formula length. Our method can take into account task formulas with a length of more than 400 and a fleet with more than $400$ agents, while most existing methods fail at the formula length of 25 within a reasonable duration. The proposed method is validated on large fleets of service robots in both simulation and hardware experiments.
1005.0600
Manuel Kauers
Manuel Kauers and Veronika Pillwein
When can we decide that a P-finite sequence is positive?
null
null
null
null
cs.SC
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We consider two algorithms which can be used for proving positivity of sequences that are defined by a linear recurrence equation with polynomial coefficients (P-finite sequences). Both algorithms have in common that while they do succeed on a great many examples, there is no guarantee for them to terminate, and they do in fact not terminate for every input. For some restricted classes of P-finite recurrence equations of order up to three we provide a priori criteria that assert the termination of the algorithms.
[ { "created": "Tue, 4 May 2010 18:24:19 GMT", "version": "v1" } ]
2010-05-05
[ [ "Kauers", "Manuel", "" ], [ "Pillwein", "Veronika", "" ] ]
We consider two algorithms which can be used for proving positivity of sequences that are defined by a linear recurrence equation with polynomial coefficients (P-finite sequences). Both algorithms have in common that while they do succeed on a great many examples, there is no guarantee for them to terminate, and they do in fact not terminate for every input. For some restricted classes of P-finite recurrence equations of order up to three we provide a priori criteria that assert the termination of the algorithms.
1510.02395
Mohammed Gollapalli Dr.
Mohammed Gollapalli
Literature Review Of Attribute Level And Structure Level Data Linkage Techniques
20 pages
International Journal of Data Mining & Knowledge Management Process (IJDKP) Vol.5, No.5, September 2015
null
null
cs.DB
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Data Linkage is an important step that can provide valuable insights for evidence-based decision making, especially for crucial events. Performing sensible queries across heterogeneous databases containing millions of records is a complex task that requires a complete understanding of each contributing databases schema to define the structure of its information. The key aim is to approximate the structure and content of the induced data into a concise synopsis in order to extract and link meaningful data-driven facts. We identify such problems as four major research issues in Data Linkage: associated costs in pair-wise matching, record matching overheads, semantic flow of information restrictions, and single order classification limitations. In this paper, we give a literature review of research in Data Linkage. The purpose for this review is to establish a basic understanding of Data Linkage, and to discuss the background in the Data Linkage research domain. Particularly, we focus on the literature related to the recent advancements in Approximate Matching algorithms at Attribute Level and Structure Level. Their efficiency, functionality and limitations are critically analysed and open-ended problems have been exposed.
[ { "created": "Wed, 7 Oct 2015 12:38:24 GMT", "version": "v1" } ]
2015-10-09
[ [ "Gollapalli", "Mohammed", "" ] ]
Data Linkage is an important step that can provide valuable insights for evidence-based decision making, especially for crucial events. Performing sensible queries across heterogeneous databases containing millions of records is a complex task that requires a complete understanding of each contributing databases schema to define the structure of its information. The key aim is to approximate the structure and content of the induced data into a concise synopsis in order to extract and link meaningful data-driven facts. We identify such problems as four major research issues in Data Linkage: associated costs in pair-wise matching, record matching overheads, semantic flow of information restrictions, and single order classification limitations. In this paper, we give a literature review of research in Data Linkage. The purpose for this review is to establish a basic understanding of Data Linkage, and to discuss the background in the Data Linkage research domain. Particularly, we focus on the literature related to the recent advancements in Approximate Matching algorithms at Attribute Level and Structure Level. Their efficiency, functionality and limitations are critically analysed and open-ended problems have been exposed.
2205.08891
Jingqing Zhang
Jingqing Zhang, Atri Sharma, Luis Bolanos, Tong Li, Ashwani Tanwar, Vibhor Gupta, Yike Guo
A Scalable Workflow to Build Machine Learning Classifiers with Clinician-in-the-Loop to Identify Patients in Specific Diseases
Under review
null
null
null
cs.CL
http://creativecommons.org/licenses/by/4.0/
Clinicians may rely on medical coding systems such as International Classification of Diseases (ICD) to identify patients with diseases from Electronic Health Records (EHRs). However, due to the lack of detail and specificity as well as a probability of miscoding, recent studies suggest the ICD codes often cannot characterise patients accurately for specific diseases in real clinical practice, and as a result, using them to find patients for studies or trials can result in high failure rates and missing out on uncoded patients. Manual inspection of all patients at scale is not feasible as it is highly costly and slow. This paper proposes a scalable workflow which leverages both structured data and unstructured textual notes from EHRs with techniques including NLP, AutoML and Clinician-in-the-Loop mechanism to build machine learning classifiers to identify patients at scale with given diseases, especially those who might currently be miscoded or missed by ICD codes. Case studies in the MIMIC-III dataset were conducted where the proposed workflow demonstrates a higher classification performance in terms of F1 scores compared to simply using ICD codes on gold testing subset to identify patients with Ovarian Cancer (0.901 vs 0.814), Lung Cancer (0.859 vs 0.828), Cancer Cachexia (0.862 vs 0.650), and Lupus Nephritis (0.959 vs 0.855). Also, the proposed workflow that leverages unstructured notes consistently outperforms the baseline that uses structured data only with an increase of F1 (Ovarian Cancer 0.901 vs 0.719, Lung Cancer 0.859 vs 0.787, Cancer Cachexia 0.862 vs 0.838 and Lupus Nephritis 0.959 vs 0.785). Experiments on the large testing set also demonstrate the proposed workflow can find more patients who are miscoded or missed by ICD codes. Moreover, interpretability studies are also conducted to clinically validate the top impact features of the classifiers.
[ { "created": "Wed, 18 May 2022 12:24:07 GMT", "version": "v1" } ]
2022-05-19
[ [ "Zhang", "Jingqing", "" ], [ "Sharma", "Atri", "" ], [ "Bolanos", "Luis", "" ], [ "Li", "Tong", "" ], [ "Tanwar", "Ashwani", "" ], [ "Gupta", "Vibhor", "" ], [ "Guo", "Yike", "" ] ]
Clinicians may rely on medical coding systems such as International Classification of Diseases (ICD) to identify patients with diseases from Electronic Health Records (EHRs). However, due to the lack of detail and specificity as well as a probability of miscoding, recent studies suggest the ICD codes often cannot characterise patients accurately for specific diseases in real clinical practice, and as a result, using them to find patients for studies or trials can result in high failure rates and missing out on uncoded patients. Manual inspection of all patients at scale is not feasible as it is highly costly and slow. This paper proposes a scalable workflow which leverages both structured data and unstructured textual notes from EHRs with techniques including NLP, AutoML and Clinician-in-the-Loop mechanism to build machine learning classifiers to identify patients at scale with given diseases, especially those who might currently be miscoded or missed by ICD codes. Case studies in the MIMIC-III dataset were conducted where the proposed workflow demonstrates a higher classification performance in terms of F1 scores compared to simply using ICD codes on gold testing subset to identify patients with Ovarian Cancer (0.901 vs 0.814), Lung Cancer (0.859 vs 0.828), Cancer Cachexia (0.862 vs 0.650), and Lupus Nephritis (0.959 vs 0.855). Also, the proposed workflow that leverages unstructured notes consistently outperforms the baseline that uses structured data only with an increase of F1 (Ovarian Cancer 0.901 vs 0.719, Lung Cancer 0.859 vs 0.787, Cancer Cachexia 0.862 vs 0.838 and Lupus Nephritis 0.959 vs 0.785). Experiments on the large testing set also demonstrate the proposed workflow can find more patients who are miscoded or missed by ICD codes. Moreover, interpretability studies are also conducted to clinically validate the top impact features of the classifiers.
2405.19701
Lavanya Prahallad
Lavanya Prahallad, Radhika Mamidi
Significance of Chain of Thought in Gender Bias Mitigation for English-Dravidian Machine Translation
6 pages
null
null
null
cs.CL cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Gender bias in machine translation (MT) sys- tems poses a significant challenge to achieving accurate and inclusive translations. This paper examines gender bias in machine translation systems for languages such as Telugu and Kan- nada from the Dravidian family, analyzing how gender inflections affect translation accuracy and neutrality using Google Translate and Chat- GPT. It finds that while plural forms can reduce bias, individual-centric sentences often main- tain the bias due to historical stereotypes. The study evaluates the Chain of Thought process- ing, noting significant bias mitigation from 80% to 4% in Telugu and from 40% to 0% in Kan- nada. It also compares Telugu and Kannada translations, emphasizing the need for language specific strategies to address these challenges and suggesting directions for future research to enhance fairness in both data preparation and prompts during inference.
[ { "created": "Thu, 30 May 2024 05:26:57 GMT", "version": "v1" }, { "created": "Mon, 3 Jun 2024 15:59:34 GMT", "version": "v2" } ]
2024-06-04
[ [ "Prahallad", "Lavanya", "" ], [ "Mamidi", "Radhika", "" ] ]
Gender bias in machine translation (MT) sys- tems poses a significant challenge to achieving accurate and inclusive translations. This paper examines gender bias in machine translation systems for languages such as Telugu and Kan- nada from the Dravidian family, analyzing how gender inflections affect translation accuracy and neutrality using Google Translate and Chat- GPT. It finds that while plural forms can reduce bias, individual-centric sentences often main- tain the bias due to historical stereotypes. The study evaluates the Chain of Thought process- ing, noting significant bias mitigation from 80% to 4% in Telugu and from 40% to 0% in Kan- nada. It also compares Telugu and Kannada translations, emphasizing the need for language specific strategies to address these challenges and suggesting directions for future research to enhance fairness in both data preparation and prompts during inference.
1312.5912
Andrea Cal\`i PhD
Andrea Cal\`i and Riccardo Torlone
Containment of Schema Mappings for Data Exchange (Preliminary Report)
11 pages, no figures
null
null
null
cs.DB
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In data exchange, data are materialised from a source schema to a target schema, according to suitable source-to-target constraints. Constraints are also expressed on the target schema to represent the domain of interest. A schema mapping is the union of the source-to-target and of the target constraints. In this paper, we address the problem of containment of schema mappings for data exchange, which has been recently proposed in this framework as a step towards the optimization of data exchange settings. We refer to a natural notion of containment that relies on the behaviour of schema mappings with respect to conjunctive query answering, in the presence of so-called LAV TGDs as target constraints. Our contribution is a practical technique for testing the containment based on the existence of a homomorphism between special "dummy" instances, which can be easily built from schema mappings. We argue that containment of schema mappings is decidable for most practical cases, and we set the basis for further investigations in the topic. This paper extends our preliminary results.
[ { "created": "Fri, 20 Dec 2013 12:13:11 GMT", "version": "v1" }, { "created": "Tue, 31 Dec 2013 09:51:14 GMT", "version": "v2" } ]
2014-01-03
[ [ "Calì", "Andrea", "" ], [ "Torlone", "Riccardo", "" ] ]
In data exchange, data are materialised from a source schema to a target schema, according to suitable source-to-target constraints. Constraints are also expressed on the target schema to represent the domain of interest. A schema mapping is the union of the source-to-target and of the target constraints. In this paper, we address the problem of containment of schema mappings for data exchange, which has been recently proposed in this framework as a step towards the optimization of data exchange settings. We refer to a natural notion of containment that relies on the behaviour of schema mappings with respect to conjunctive query answering, in the presence of so-called LAV TGDs as target constraints. Our contribution is a practical technique for testing the containment based on the existence of a homomorphism between special "dummy" instances, which can be easily built from schema mappings. We argue that containment of schema mappings is decidable for most practical cases, and we set the basis for further investigations in the topic. This paper extends our preliminary results.
1411.1607
Alan Edelman
Jeff Bezanson, Alan Edelman, Stefan Karpinski, Viral B. Shah
Julia: A Fresh Approach to Numerical Computing
37 pages
null
null
null
cs.MS
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Bridging cultures that have often been distant, Julia combines expertise from the diverse fields of computer science and computational science to create a new approach to numerical computing. Julia is designed to be easy and fast. Julia questions notions generally held as "laws of nature" by practitioners of numerical computing: 1. High-level dynamic programs have to be slow. 2. One must prototype in one language and then rewrite in another language for speed or deployment, and 3. There are parts of a system for the programmer, and other parts best left untouched as they are built by the experts. We introduce the Julia programming language and its design --- a dance between specialization and abstraction. Specialization allows for custom treatment. Multiple dispatch, a technique from computer science, picks the right algorithm for the right circumstance. Abstraction, what good computation is really about, recognizes what remains the same after differences are stripped away. Abstractions in mathematics are captured as code through another technique from computer science, generic programming. Julia shows that one can have machine performance without sacrificing human convenience.
[ { "created": "Thu, 6 Nov 2014 13:39:40 GMT", "version": "v1" }, { "created": "Fri, 7 Nov 2014 11:19:21 GMT", "version": "v2" }, { "created": "Fri, 12 Dec 2014 22:40:09 GMT", "version": "v3" }, { "created": "Sun, 19 Jul 2015 19:58:28 GMT", "version": "v4" } ]
2015-07-21
[ [ "Bezanson", "Jeff", "" ], [ "Edelman", "Alan", "" ], [ "Karpinski", "Stefan", "" ], [ "Shah", "Viral B.", "" ] ]
Bridging cultures that have often been distant, Julia combines expertise from the diverse fields of computer science and computational science to create a new approach to numerical computing. Julia is designed to be easy and fast. Julia questions notions generally held as "laws of nature" by practitioners of numerical computing: 1. High-level dynamic programs have to be slow. 2. One must prototype in one language and then rewrite in another language for speed or deployment, and 3. There are parts of a system for the programmer, and other parts best left untouched as they are built by the experts. We introduce the Julia programming language and its design --- a dance between specialization and abstraction. Specialization allows for custom treatment. Multiple dispatch, a technique from computer science, picks the right algorithm for the right circumstance. Abstraction, what good computation is really about, recognizes what remains the same after differences are stripped away. Abstractions in mathematics are captured as code through another technique from computer science, generic programming. Julia shows that one can have machine performance without sacrificing human convenience.
2208.06946
Fangyi Yu
Fangyi Yu and Miguel Vargas Martin
Targeted Honeyword Generation with Language Models
8 pages, 7 tables, 2 figures
null
null
null
cs.AI cs.CR
http://creativecommons.org/licenses/by/4.0/
Honeywords are fictitious passwords inserted into databases in order to identify password breaches. The major difficulty is how to produce honeywords that are difficult to distinguish from real passwords. Although the generation of honeywords has been widely investigated in the past, the majority of existing research assumes attackers have no knowledge of the users. These honeyword generating techniques (HGTs) may utterly fail if attackers exploit users' personally identifiable information (PII) and the real passwords include users' PII. In this paper, we propose to build a more secure and trustworthy authentication system that employs off-the-shelf pre-trained language models which require no further training on real passwords to produce honeywords while retaining the PII of the associated real password, therefore significantly raising the bar for attackers. We conducted a pilot experiment in which individuals are asked to distinguish between authentic passwords and honeywords when the username is provided for GPT-3 and a tweaking technique. Results show that it is extremely difficult to distinguish the real passwords from the artifical ones for both techniques. We speculate that a larger sample size could reveal a significant difference between the two HGT techniques, favouring our proposed approach.
[ { "created": "Mon, 15 Aug 2022 00:06:29 GMT", "version": "v1" }, { "created": "Tue, 23 Aug 2022 16:12:27 GMT", "version": "v2" } ]
2022-08-24
[ [ "Yu", "Fangyi", "" ], [ "Martin", "Miguel Vargas", "" ] ]
Honeywords are fictitious passwords inserted into databases in order to identify password breaches. The major difficulty is how to produce honeywords that are difficult to distinguish from real passwords. Although the generation of honeywords has been widely investigated in the past, the majority of existing research assumes attackers have no knowledge of the users. These honeyword generating techniques (HGTs) may utterly fail if attackers exploit users' personally identifiable information (PII) and the real passwords include users' PII. In this paper, we propose to build a more secure and trustworthy authentication system that employs off-the-shelf pre-trained language models which require no further training on real passwords to produce honeywords while retaining the PII of the associated real password, therefore significantly raising the bar for attackers. We conducted a pilot experiment in which individuals are asked to distinguish between authentic passwords and honeywords when the username is provided for GPT-3 and a tweaking technique. Results show that it is extremely difficult to distinguish the real passwords from the artifical ones for both techniques. We speculate that a larger sample size could reveal a significant difference between the two HGT techniques, favouring our proposed approach.
1903.02255
Peter Boyvalenkov
Peter Boyvalenkov, Danyo Danev
Linear Programming Bounds
22 pages
null
null
null
cs.IT math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This chapter is written for the forthcoming book "A Concise Encyclopedia of Coding Theory" (CRC press), edited by W. Cary Huffman, Jon-Lark Kim, and Patrick Sol\'e. This book will collect short but foundational articles, emphasizing definitions, examples, exhaustive references, and basic facts. The target audience of the Encyclopedia is upper level undergraduates and graduate students.
[ { "created": "Wed, 6 Mar 2019 09:16:13 GMT", "version": "v1" } ]
2019-03-07
[ [ "Boyvalenkov", "Peter", "" ], [ "Danev", "Danyo", "" ] ]
This chapter is written for the forthcoming book "A Concise Encyclopedia of Coding Theory" (CRC press), edited by W. Cary Huffman, Jon-Lark Kim, and Patrick Sol\'e. This book will collect short but foundational articles, emphasizing definitions, examples, exhaustive references, and basic facts. The target audience of the Encyclopedia is upper level undergraduates and graduate students.
2203.16952
Swalpa Kumar Roy Dr.
Swalpa Kumar Roy, Ankur Deria, Danfeng Hong, Behnood Rasti, Antonio Plaza, Jocelyn Chanussot
Multimodal Fusion Transformer for Remote Sensing Image Classification
Published in IEEE Transactions on Geoscience and Remote Sensing
null
10.1109/TGRS.2023.3286826
null
cs.CV cs.LG eess.IV
http://creativecommons.org/licenses/by/4.0/
Vision transformers (ViTs) have been trending in image classification tasks due to their promising performance when compared to convolutional neural networks (CNNs). As a result, many researchers have tried to incorporate ViTs in hyperspectral image (HSI) classification tasks. To achieve satisfactory performance, close to that of CNNs, transformers need fewer parameters. ViTs and other similar transformers use an external classification (CLS) token which is randomly initialized and often fails to generalize well, whereas other sources of multimodal datasets, such as light detection and ranging (LiDAR) offer the potential to improve these models by means of a CLS. In this paper, we introduce a new multimodal fusion transformer (MFT) network which comprises a multihead cross patch attention (mCrossPA) for HSI land-cover classification. Our mCrossPA utilizes other sources of complementary information in addition to the HSI in the transformer encoder to achieve better generalization. The concept of tokenization is used to generate CLS and HSI patch tokens, helping to learn a {distinctive representation} in a reduced and hierarchical feature space. Extensive experiments are carried out on {widely used benchmark} datasets {i.e.,} the University of Houston, Trento, University of Southern Mississippi Gulfpark (MUUFL), and Augsburg. We compare the results of the proposed MFT model with other state-of-the-art transformers, classical CNNs, and conventional classifiers models. The superior performance achieved by the proposed model is due to the use of multihead cross patch attention. The source code will be made available publicly at \url{https://github.com/AnkurDeria/MFT}.}
[ { "created": "Thu, 31 Mar 2022 11:18:41 GMT", "version": "v1" }, { "created": "Tue, 20 Jun 2023 17:58:25 GMT", "version": "v2" } ]
2023-06-21
[ [ "Roy", "Swalpa Kumar", "" ], [ "Deria", "Ankur", "" ], [ "Hong", "Danfeng", "" ], [ "Rasti", "Behnood", "" ], [ "Plaza", "Antonio", "" ], [ "Chanussot", "Jocelyn", "" ] ]
Vision transformers (ViTs) have been trending in image classification tasks due to their promising performance when compared to convolutional neural networks (CNNs). As a result, many researchers have tried to incorporate ViTs in hyperspectral image (HSI) classification tasks. To achieve satisfactory performance, close to that of CNNs, transformers need fewer parameters. ViTs and other similar transformers use an external classification (CLS) token which is randomly initialized and often fails to generalize well, whereas other sources of multimodal datasets, such as light detection and ranging (LiDAR) offer the potential to improve these models by means of a CLS. In this paper, we introduce a new multimodal fusion transformer (MFT) network which comprises a multihead cross patch attention (mCrossPA) for HSI land-cover classification. Our mCrossPA utilizes other sources of complementary information in addition to the HSI in the transformer encoder to achieve better generalization. The concept of tokenization is used to generate CLS and HSI patch tokens, helping to learn a {distinctive representation} in a reduced and hierarchical feature space. Extensive experiments are carried out on {widely used benchmark} datasets {i.e.,} the University of Houston, Trento, University of Southern Mississippi Gulfpark (MUUFL), and Augsburg. We compare the results of the proposed MFT model with other state-of-the-art transformers, classical CNNs, and conventional classifiers models. The superior performance achieved by the proposed model is due to the use of multihead cross patch attention. The source code will be made available publicly at \url{https://github.com/AnkurDeria/MFT}.}
cs/0701118
Mohammad Ali Maddah-Ali Mr.
Mohammad Ali Maddah-Ali, Hajar Mahdavi-Doost, and Amir K. Khandani
Optimal Order of Decoding for Max-Min Fairness in $K$-User Memoryless Interference Channels
11 Pages, Submitted to IEEE International Symposium on Information Theory(ISIT 2007)
null
10.1109/ISIT.2007.4557653
null
cs.IT math.IT
null
A $K$-user memoryless interference channel is considered where each receiver sequentially decodes the data of a subset of transmitters before it decodes the data of the designated transmitter. Therefore, the data rate of each transmitter depends on (i) the subset of receivers which decode the data of that transmitter, (ii) the decoding order, employed at each of these receivers. In this paper, a greedy algorithm is developed to find the users which are decoded at each receiver and the corresponding decoding order such that the minimum rate of the users is maximized. It is proven that the proposed algorithm is optimal.
[ { "created": "Thu, 18 Jan 2007 20:54:03 GMT", "version": "v1" } ]
2016-11-15
[ [ "Maddah-Ali", "Mohammad Ali", "" ], [ "Mahdavi-Doost", "Hajar", "" ], [ "Khandani", "Amir K.", "" ] ]
A $K$-user memoryless interference channel is considered where each receiver sequentially decodes the data of a subset of transmitters before it decodes the data of the designated transmitter. Therefore, the data rate of each transmitter depends on (i) the subset of receivers which decode the data of that transmitter, (ii) the decoding order, employed at each of these receivers. In this paper, a greedy algorithm is developed to find the users which are decoded at each receiver and the corresponding decoding order such that the minimum rate of the users is maximized. It is proven that the proposed algorithm is optimal.
2201.04205
Waleed Yousef
Waleed A.Yousef, Hisham E. Mohammed, Andrew A. Naguib, Rafat S. Eid, Sherif E. Emabrak, Ahmed F. Hamed, Yusuf M. Khalifa, Shrouk T. AbdElrheem, Eman A. Awad, Sara G. Gaafar, Alaa M. Mamdoh, Nada A. Shawky
JSOL: JavaScript Open-source Library for Grammar of Graphics
null
null
null
null
cs.GR
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In this paper, we introduce the JavaScript Open-source Library (\libname), a high-level grammar for representing data in visualization graphs and plots. \libname~perspective on the grammar of graphics is unique; it provides state-of-art rules for encoding visual primitives that can be used to generate a known scene or to invent a new one. \libname~has ton rules developed specifically for data-munging, mapping, and visualization through many layers, such as algebra, scales, and geometries. Additionally, it has a compiler that incorporates and combines all rules specified by a user and put them in a flow to validate it as a visualization grammar and check its requisites. Users can customize scenes through a pipeline that either puts customized rules or comes with new ones. We evaluated \libname~on a multitude of plots to check rules specification of customizing a specific plot. Although the project is still under development and many enhancements are under construction, this paper describes the first developed version of \libname, circa 2016, where an open-source version of it is available. One immediate practical deployment for JSOl is to be integrated with the open-source version of the Data Visualization Platform (DVP) \citep{Yousef2019DVP-arxiv}
[ { "created": "Tue, 11 Jan 2022 21:23:23 GMT", "version": "v1" } ]
2022-01-13
[ [ "Yousef", "Waleed A.", "" ], [ "Mohammed", "Hisham E.", "" ], [ "Naguib", "Andrew A.", "" ], [ "Eid", "Rafat S.", "" ], [ "Emabrak", "Sherif E.", "" ], [ "Hamed", "Ahmed F.", "" ], [ "Khalifa", "Yusuf M.", "" ], [ "AbdElrheem", "Shrouk T.", "" ], [ "Awad", "Eman A.", "" ], [ "Gaafar", "Sara G.", "" ], [ "Mamdoh", "Alaa M.", "" ], [ "Shawky", "Nada A.", "" ] ]
In this paper, we introduce the JavaScript Open-source Library (\libname), a high-level grammar for representing data in visualization graphs and plots. \libname~perspective on the grammar of graphics is unique; it provides state-of-art rules for encoding visual primitives that can be used to generate a known scene or to invent a new one. \libname~has ton rules developed specifically for data-munging, mapping, and visualization through many layers, such as algebra, scales, and geometries. Additionally, it has a compiler that incorporates and combines all rules specified by a user and put them in a flow to validate it as a visualization grammar and check its requisites. Users can customize scenes through a pipeline that either puts customized rules or comes with new ones. We evaluated \libname~on a multitude of plots to check rules specification of customizing a specific plot. Although the project is still under development and many enhancements are under construction, this paper describes the first developed version of \libname, circa 2016, where an open-source version of it is available. One immediate practical deployment for JSOl is to be integrated with the open-source version of the Data Visualization Platform (DVP) \citep{Yousef2019DVP-arxiv}
1912.11855
Aditi Sharma
Aditi Sharma, Ravi Ranjan
Software Effort Estimation using Neuro Fuzzy Inference System: Past and Present
null
International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169 2017
null
null
cs.SE cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Most important reason for project failure is poor effort estimation. Software development effort estimation is needed for assigning appropriate team members for development, allocating resources for software development, binding etc. Inaccurate software estimation may lead to delay in project, over-budget or cancellation of the project. But the effort estimation models are not very efficient. In this paper, we are analyzing the new approach for estimation i.e. Neuro Fuzzy Inference System (NFIS). It is a mixture model that consolidates the components of artificial neural network with fuzzy logic for giving a better estimation.
[ { "created": "Thu, 26 Dec 2019 12:55:38 GMT", "version": "v1" } ]
2019-12-30
[ [ "Sharma", "Aditi", "" ], [ "Ranjan", "Ravi", "" ] ]
Most important reason for project failure is poor effort estimation. Software development effort estimation is needed for assigning appropriate team members for development, allocating resources for software development, binding etc. Inaccurate software estimation may lead to delay in project, over-budget or cancellation of the project. But the effort estimation models are not very efficient. In this paper, we are analyzing the new approach for estimation i.e. Neuro Fuzzy Inference System (NFIS). It is a mixture model that consolidates the components of artificial neural network with fuzzy logic for giving a better estimation.
2012.04580
James Jordon
James Jordon, Alan Wilson and Mihaela van der Schaar
Synthetic Data: Opening the data floodgates to enable faster, more directed development of machine learning methods
null
null
null
null
cs.LG cs.CY
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Many ground-breaking advancements in machine learning can be attributed to the availability of a large volume of rich data. Unfortunately, many large-scale datasets are highly sensitive, such as healthcare data, and are not widely available to the machine learning community. Generating synthetic data with privacy guarantees provides one such solution, allowing meaningful research to be carried out "at scale" - by allowing the entirety of the machine learning community to potentially accelerate progress within a given field. In this article, we provide a high-level view of synthetic data: what it means, how we might evaluate it and how we might use it.
[ { "created": "Tue, 8 Dec 2020 17:26:10 GMT", "version": "v1" } ]
2020-12-09
[ [ "Jordon", "James", "" ], [ "Wilson", "Alan", "" ], [ "van der Schaar", "Mihaela", "" ] ]
Many ground-breaking advancements in machine learning can be attributed to the availability of a large volume of rich data. Unfortunately, many large-scale datasets are highly sensitive, such as healthcare data, and are not widely available to the machine learning community. Generating synthetic data with privacy guarantees provides one such solution, allowing meaningful research to be carried out "at scale" - by allowing the entirety of the machine learning community to potentially accelerate progress within a given field. In this article, we provide a high-level view of synthetic data: what it means, how we might evaluate it and how we might use it.
1910.08888
Carlo Zaniolo
Carlo Zaniolo, Ariyam Das, Jiaqi Gu, Youfu Li, Mingda li, Jin Wang
Monotonic Properties of Completed Aggregates in Recursive Queries
null
null
null
null
cs.DB
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The use of aggregates in recursion enables efficient and scalable support for a wide range of BigData algorithms, including those used in graph applications, KDD applications, and ML applications, which have proven difficult to be expressed and supported efficiently in BigData systems supporting Datalog or SQL. The problem with these languages and systems is that, to avoid the semantic and computational issues created by non-monotonic constructs in recursion, they only allow programs that are stratified with respect to negation and aggregates. Now, while this crippling restriction is well-justified for negation, it is frequently unjustified for aggregates, since (i) aggregates are often monotonic in the standard lattice of set-containment, (ii) the PreM property guarantees that programs with extrema in recursion are equivalent to stratified programs where extrema are used as post-constraints, and (iii) any program computing any aggregates on sets of facts of predictable cardinality tantamounts to stratified programs where the precomputation of the cardinality of the set is followed by a stratum where recursive rules only use monotonic constructs. With (i) and (ii) covered in previous papers, this paper focuses on (iii) using examples of great practical interest. For such examples, we provide a formal semantics that is conducive to efficient and scalable implementations via well-known techniques such as semi-naive fixpoint currently supported by most Datalog and SQL3 systems.
[ { "created": "Sun, 20 Oct 2019 03:52:40 GMT", "version": "v1" } ]
2019-10-22
[ [ "Zaniolo", "Carlo", "" ], [ "Das", "Ariyam", "" ], [ "Gu", "Jiaqi", "" ], [ "Li", "Youfu", "" ], [ "li", "Mingda", "" ], [ "Wang", "Jin", "" ] ]
The use of aggregates in recursion enables efficient and scalable support for a wide range of BigData algorithms, including those used in graph applications, KDD applications, and ML applications, which have proven difficult to be expressed and supported efficiently in BigData systems supporting Datalog or SQL. The problem with these languages and systems is that, to avoid the semantic and computational issues created by non-monotonic constructs in recursion, they only allow programs that are stratified with respect to negation and aggregates. Now, while this crippling restriction is well-justified for negation, it is frequently unjustified for aggregates, since (i) aggregates are often monotonic in the standard lattice of set-containment, (ii) the PreM property guarantees that programs with extrema in recursion are equivalent to stratified programs where extrema are used as post-constraints, and (iii) any program computing any aggregates on sets of facts of predictable cardinality tantamounts to stratified programs where the precomputation of the cardinality of the set is followed by a stratum where recursive rules only use monotonic constructs. With (i) and (ii) covered in previous papers, this paper focuses on (iii) using examples of great practical interest. For such examples, we provide a formal semantics that is conducive to efficient and scalable implementations via well-known techniques such as semi-naive fixpoint currently supported by most Datalog and SQL3 systems.
2407.15865
Craig Pirie
Craig Pirie, Harsha Kalutarage, Muhammad Shadi Hajar, Nirmalie Wiratunga, Subodha Charles, Geeth Sandaru Madhushan, Priyantha Buddhika, Supun Wijesiriwardana, Akila Dimantha, Kithdara Hansamal, Shalitha Pathiranage
A Survey of AI-Powered Mini-Grid Solutions for a Sustainable Future in Rural Communities
null
null
null
null
cs.LG cs.AI cs.CE
http://creativecommons.org/licenses/by/4.0/
This paper presents a comprehensive survey of AI-driven mini-grid solutions aimed at enhancing sustainable energy access. It emphasises the potential of mini-grids, which can operate independently or in conjunction with national power grids, to provide reliable and affordable electricity to remote communities. Given the inherent unpredictability of renewable energy sources such as solar and wind, the necessity for accurate energy forecasting and management is discussed, highlighting the role of advanced AI techniques in forecasting energy supply and demand, optimising grid operations, and ensuring sustainable energy distribution. This paper reviews various forecasting models, including statistical methods, machine learning algorithms, and hybrid approaches, evaluating their effectiveness for both short-term and long-term predictions. Additionally, it explores public datasets and tools such as Prophet, NeuralProphet, and N-BEATS for model implementation and validation. The survey concludes with recommendations for future research, addressing challenges in model adaptation and optimisation for real-world applications.
[ { "created": "Wed, 17 Jul 2024 20:23:38 GMT", "version": "v1" } ]
2024-07-24
[ [ "Pirie", "Craig", "" ], [ "Kalutarage", "Harsha", "" ], [ "Hajar", "Muhammad Shadi", "" ], [ "Wiratunga", "Nirmalie", "" ], [ "Charles", "Subodha", "" ], [ "Madhushan", "Geeth Sandaru", "" ], [ "Buddhika", "Priyantha", "" ], [ "Wijesiriwardana", "Supun", "" ], [ "Dimantha", "Akila", "" ], [ "Hansamal", "Kithdara", "" ], [ "Pathiranage", "Shalitha", "" ] ]
This paper presents a comprehensive survey of AI-driven mini-grid solutions aimed at enhancing sustainable energy access. It emphasises the potential of mini-grids, which can operate independently or in conjunction with national power grids, to provide reliable and affordable electricity to remote communities. Given the inherent unpredictability of renewable energy sources such as solar and wind, the necessity for accurate energy forecasting and management is discussed, highlighting the role of advanced AI techniques in forecasting energy supply and demand, optimising grid operations, and ensuring sustainable energy distribution. This paper reviews various forecasting models, including statistical methods, machine learning algorithms, and hybrid approaches, evaluating their effectiveness for both short-term and long-term predictions. Additionally, it explores public datasets and tools such as Prophet, NeuralProphet, and N-BEATS for model implementation and validation. The survey concludes with recommendations for future research, addressing challenges in model adaptation and optimisation for real-world applications.
2304.12152
Haitian Jiang
Haitian Jiang, Dongliang Xiong, Xiaowen Jiang, Li Ding, Liang Chen, Kai Huang
Efficient Halftoning via Deep Reinforcement Learning
null
IEEE Transactions on Image Processing (TIP), 2023
10.1109/TIP.2023.3318937
null
cs.CV cs.GR
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Halftoning aims to reproduce a continuous-tone image with pixels whose intensities are constrained to two discrete levels. This technique has been deployed on every printer, and the majority of them adopt fast methods (e.g., ordered dithering, error diffusion) that fail to render structural details, which determine halftone's quality. Other prior methods of pursuing visual pleasure by searching for the optimal halftone solution, on the contrary, suffer from their high computational cost. In this paper, we propose a fast and structure-aware halftoning method via a data-driven approach. Specifically, we formulate halftoning as a reinforcement learning problem, in which each binary pixel's value is regarded as an action chosen by a virtual agent with a shared fully convolutional neural network (CNN) policy. In the offline phase, an effective gradient estimator is utilized to train the agents in producing high-quality halftones in one action step. Then, halftones can be generated online by one fast CNN inference. Besides, we propose a novel anisotropy suppressing loss function, which brings the desirable blue-noise property. Finally, we find that optimizing SSIM could result in holes in flat areas, which can be avoided by weighting the metric with the contone's contrast map. Experiments show that our framework can effectively train a light-weight CNN, which is 15x faster than previous structure-aware methods, to generate blue-noise halftones with satisfactory visual quality. We also present a prototype of deep multitoning to demonstrate the extensibility of our method.
[ { "created": "Mon, 24 Apr 2023 15:03:37 GMT", "version": "v1" }, { "created": "Fri, 13 Oct 2023 03:40:42 GMT", "version": "v2" } ]
2023-10-16
[ [ "Jiang", "Haitian", "" ], [ "Xiong", "Dongliang", "" ], [ "Jiang", "Xiaowen", "" ], [ "Ding", "Li", "" ], [ "Chen", "Liang", "" ], [ "Huang", "Kai", "" ] ]
Halftoning aims to reproduce a continuous-tone image with pixels whose intensities are constrained to two discrete levels. This technique has been deployed on every printer, and the majority of them adopt fast methods (e.g., ordered dithering, error diffusion) that fail to render structural details, which determine halftone's quality. Other prior methods of pursuing visual pleasure by searching for the optimal halftone solution, on the contrary, suffer from their high computational cost. In this paper, we propose a fast and structure-aware halftoning method via a data-driven approach. Specifically, we formulate halftoning as a reinforcement learning problem, in which each binary pixel's value is regarded as an action chosen by a virtual agent with a shared fully convolutional neural network (CNN) policy. In the offline phase, an effective gradient estimator is utilized to train the agents in producing high-quality halftones in one action step. Then, halftones can be generated online by one fast CNN inference. Besides, we propose a novel anisotropy suppressing loss function, which brings the desirable blue-noise property. Finally, we find that optimizing SSIM could result in holes in flat areas, which can be avoided by weighting the metric with the contone's contrast map. Experiments show that our framework can effectively train a light-weight CNN, which is 15x faster than previous structure-aware methods, to generate blue-noise halftones with satisfactory visual quality. We also present a prototype of deep multitoning to demonstrate the extensibility of our method.
2011.14058
Wei He
Zhongzhan Huang, Senwei Liang, Mingfu Liang, Wei He, Haizhao Yang
Efficient Attention Network: Accelerate Attention by Searching Where to Plug
null
null
null
null
cs.CV cs.AI cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Recently, many plug-and-play self-attention modules are proposed to enhance the model generalization by exploiting the internal information of deep convolutional neural networks (CNNs). Previous works lay an emphasis on the design of attention module for specific functionality, e.g., light-weighted or task-oriented attention. However, they ignore the importance of where to plug in the attention module since they connect the modules individually with each block of the entire CNN backbone for granted, leading to incremental computational cost and number of parameters with the growth of network depth. Thus, we propose a framework called Efficient Attention Network (EAN) to improve the efficiency for the existing attention modules. In EAN, we leverage the sharing mechanism (Huang et al. 2020) to share the attention module within the backbone and search where to connect the shared attention module via reinforcement learning. Finally, we obtain the attention network with sparse connections between the backbone and modules, while (1) maintaining accuracy (2) reducing extra parameter increment and (3) accelerating inference. Extensive experiments on widely-used benchmarks and popular attention networks show the effectiveness of EAN. Furthermore, we empirically illustrate that our EAN has the capacity of transferring to other tasks and capturing the informative features. The code is available at https://github.com/gbup-group/EAN-efficient-attention-network.
[ { "created": "Sat, 28 Nov 2020 03:31:08 GMT", "version": "v1" }, { "created": "Sun, 11 Jul 2021 12:44:58 GMT", "version": "v2" } ]
2021-07-13
[ [ "Huang", "Zhongzhan", "" ], [ "Liang", "Senwei", "" ], [ "Liang", "Mingfu", "" ], [ "He", "Wei", "" ], [ "Yang", "Haizhao", "" ] ]
Recently, many plug-and-play self-attention modules are proposed to enhance the model generalization by exploiting the internal information of deep convolutional neural networks (CNNs). Previous works lay an emphasis on the design of attention module for specific functionality, e.g., light-weighted or task-oriented attention. However, they ignore the importance of where to plug in the attention module since they connect the modules individually with each block of the entire CNN backbone for granted, leading to incremental computational cost and number of parameters with the growth of network depth. Thus, we propose a framework called Efficient Attention Network (EAN) to improve the efficiency for the existing attention modules. In EAN, we leverage the sharing mechanism (Huang et al. 2020) to share the attention module within the backbone and search where to connect the shared attention module via reinforcement learning. Finally, we obtain the attention network with sparse connections between the backbone and modules, while (1) maintaining accuracy (2) reducing extra parameter increment and (3) accelerating inference. Extensive experiments on widely-used benchmarks and popular attention networks show the effectiveness of EAN. Furthermore, we empirically illustrate that our EAN has the capacity of transferring to other tasks and capturing the informative features. The code is available at https://github.com/gbup-group/EAN-efficient-attention-network.
2307.12517
Pourya Shamsolmoali
Pourya Shamsolmoali, Masoumeh Zareapoor
Entropy Transformer Networks: A Learning Approach via Tangent Bundle Data Manifold
null
null
null
null
cs.CV
http://creativecommons.org/licenses/by/4.0/
This paper focuses on an accurate and fast interpolation approach for image transformation employed in the design of CNN architectures. Standard Spatial Transformer Networks (STNs) use bilinear or linear interpolation as their interpolation, with unrealistic assumptions about the underlying data distributions, which leads to poor performance under scale variations. Moreover, STNs do not preserve the norm of gradients in propagation due to their dependency on sparse neighboring pixels. To address this problem, a novel Entropy STN (ESTN) is proposed that interpolates on the data manifold distributions. In particular, random samples are generated for each pixel in association with the tangent space of the data manifold and construct a linear approximation of their intensity values with an entropy regularizer to compute the transformer parameters. A simple yet effective technique is also proposed to normalize the non-zero values of the convolution operation, to fine-tune the layers for gradients' norm-regularization during training. Experiments on challenging benchmarks show that the proposed ESTN can improve predictive accuracy over a range of computer vision tasks, including image reconstruction, and classification, while reducing the computational cost.
[ { "created": "Mon, 24 Jul 2023 04:21:51 GMT", "version": "v1" } ]
2023-07-25
[ [ "Shamsolmoali", "Pourya", "" ], [ "Zareapoor", "Masoumeh", "" ] ]
This paper focuses on an accurate and fast interpolation approach for image transformation employed in the design of CNN architectures. Standard Spatial Transformer Networks (STNs) use bilinear or linear interpolation as their interpolation, with unrealistic assumptions about the underlying data distributions, which leads to poor performance under scale variations. Moreover, STNs do not preserve the norm of gradients in propagation due to their dependency on sparse neighboring pixels. To address this problem, a novel Entropy STN (ESTN) is proposed that interpolates on the data manifold distributions. In particular, random samples are generated for each pixel in association with the tangent space of the data manifold and construct a linear approximation of their intensity values with an entropy regularizer to compute the transformer parameters. A simple yet effective technique is also proposed to normalize the non-zero values of the convolution operation, to fine-tune the layers for gradients' norm-regularization during training. Experiments on challenging benchmarks show that the proposed ESTN can improve predictive accuracy over a range of computer vision tasks, including image reconstruction, and classification, while reducing the computational cost.
1605.04359
Aman Madaan
Aman Madaan, Sunita Sarawagi
Occurrence Statistics of Entities, Relations and Types on the Web
null
null
null
null
cs.CL
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The problem of collecting reliable estimates of occurrence of entities on the open web forms the premise for this report. The models learned for tagging entities cannot be expected to perform well when deployed on the web. This is owing to the severe mismatch in the distributions of such entities on the web and in the relatively diminutive training data. In this report, we build up the case for maximum mean discrepancy for estimation of occurrence statistics of entities on the web, taking a review of named entity disambiguation techniques and related concepts along the way.
[ { "created": "Sat, 14 May 2016 01:13:48 GMT", "version": "v1" } ]
2016-05-17
[ [ "Madaan", "Aman", "" ], [ "Sarawagi", "Sunita", "" ] ]
The problem of collecting reliable estimates of occurrence of entities on the open web forms the premise for this report. The models learned for tagging entities cannot be expected to perform well when deployed on the web. This is owing to the severe mismatch in the distributions of such entities on the web and in the relatively diminutive training data. In this report, we build up the case for maximum mean discrepancy for estimation of occurrence statistics of entities on the web, taking a review of named entity disambiguation techniques and related concepts along the way.
2002.04711
Said Hanafi
Fred Glover, Said Hanafi, and Gintaras Palubeckis
Bi-objective Optimization of Biclustering with Binary Data
37 pages
null
null
null
cs.AI cs.DM math.CO math.OC
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Clustering consists of partitioning data objects into subsets called clusters according to some similarity criteria. This paper addresses a generalization called quasi-clustering that allows overlapping of clusters, and which we link to biclustering. Biclustering simultaneously groups the objects and features so that a specific group of objects has a special group of features. In recent years, biclustering has received a lot of attention in several practical applications. In this paper we consider a bi-objective optimization of biclustering problem with binary data. First we present an integer programing formulations for the bi-objective optimization biclustering. Next we propose a constructive heuristic based on the set intersection operation and its efficient implementation for solving a series of mono-objective problems used inside the Epsilon-constraint method (obtained by keeping only one objective function and the other objective function is integrated into constraints). Finally, our experimental results show that using CPLEX solver as an exact algorithm for finding an optimal solution drastically increases the computational cost for large instances, while our proposed heuristic provides very good results and significantly reduces the computational expense.
[ { "created": "Sun, 9 Feb 2020 21:49:26 GMT", "version": "v1" } ]
2020-02-13
[ [ "Glover", "Fred", "" ], [ "Hanafi", "Said", "" ], [ "Palubeckis", "Gintaras", "" ] ]
Clustering consists of partitioning data objects into subsets called clusters according to some similarity criteria. This paper addresses a generalization called quasi-clustering that allows overlapping of clusters, and which we link to biclustering. Biclustering simultaneously groups the objects and features so that a specific group of objects has a special group of features. In recent years, biclustering has received a lot of attention in several practical applications. In this paper we consider a bi-objective optimization of biclustering problem with binary data. First we present an integer programing formulations for the bi-objective optimization biclustering. Next we propose a constructive heuristic based on the set intersection operation and its efficient implementation for solving a series of mono-objective problems used inside the Epsilon-constraint method (obtained by keeping only one objective function and the other objective function is integrated into constraints). Finally, our experimental results show that using CPLEX solver as an exact algorithm for finding an optimal solution drastically increases the computational cost for large instances, while our proposed heuristic provides very good results and significantly reduces the computational expense.
2108.02694
Liming Xu
Liming Xu, Dave Towey, Andrew French, Steve Benford, Zhi Quan Zhou and Tsong Yueh Chen
Using Metamorphic Relations to Verify and Enhance Artcode Classification
32 pages, 11 figures
null
null
null
cs.SE cs.LG
http://creativecommons.org/licenses/by/4.0/
Software testing is often hindered where it is impossible or impractical to determine the correctness of the behaviour or output of the software under test (SUT), a situation known as the oracle problem. An example of an area facing the oracle problem is automatic image classification, using machine learning to classify an input image as one of a set of predefined classes. An approach to software testing that alleviates the oracle problem is metamorphic testing (MT). While traditional software testing examines the correctness of individual test cases, MT instead examines the relations amongst multiple executions of test cases and their outputs. These relations are called metamorphic relations (MRs): if an MR is found to be violated, then a fault must exist in the SUT. This paper examines the problem of classifying images containing visually hidden markers called Artcodes, and applies MT to verify and enhance the trained classifiers. This paper further examines two MRs, Separation and Occlusion, and reports on their capability in verifying the image classification using one-way analysis of variance (ANOVA) in conjunction with three other statistical analysis methods: t-test (for unequal variances), Kruskal-Wallis test, and Dunnett's test. In addition to our previously-studied classifier, that used Random Forests, we introduce a new classifier that uses a support vector machine, and present its MR-augmented version. Experimental evaluations across a number of performance metrics show that the augmented classifiers can achieve better performance than non-augmented classifiers. This paper also analyses how the enhanced performance is obtained.
[ { "created": "Thu, 5 Aug 2021 15:54:56 GMT", "version": "v1" } ]
2021-08-06
[ [ "Xu", "Liming", "" ], [ "Towey", "Dave", "" ], [ "French", "Andrew", "" ], [ "Benford", "Steve", "" ], [ "Zhou", "Zhi Quan", "" ], [ "Chen", "Tsong Yueh", "" ] ]
Software testing is often hindered where it is impossible or impractical to determine the correctness of the behaviour or output of the software under test (SUT), a situation known as the oracle problem. An example of an area facing the oracle problem is automatic image classification, using machine learning to classify an input image as one of a set of predefined classes. An approach to software testing that alleviates the oracle problem is metamorphic testing (MT). While traditional software testing examines the correctness of individual test cases, MT instead examines the relations amongst multiple executions of test cases and their outputs. These relations are called metamorphic relations (MRs): if an MR is found to be violated, then a fault must exist in the SUT. This paper examines the problem of classifying images containing visually hidden markers called Artcodes, and applies MT to verify and enhance the trained classifiers. This paper further examines two MRs, Separation and Occlusion, and reports on their capability in verifying the image classification using one-way analysis of variance (ANOVA) in conjunction with three other statistical analysis methods: t-test (for unequal variances), Kruskal-Wallis test, and Dunnett's test. In addition to our previously-studied classifier, that used Random Forests, we introduce a new classifier that uses a support vector machine, and present its MR-augmented version. Experimental evaluations across a number of performance metrics show that the augmented classifiers can achieve better performance than non-augmented classifiers. This paper also analyses how the enhanced performance is obtained.
2305.10110
Wenzhao Zhao
Wenzhao Zhao, Barbara D. Wichtmann, Steffen Albert, Angelika Maurer, Frank G. Z\"ollner, Ulrike Attenberger and J\"urgen Hesser
Adaptive aggregation of Monte Carlo augmented decomposed filters for efficient group-equivariant convolutional neural network
null
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Group-equivariant convolutional neural networks (G-CNN) heavily rely on parameter sharing to increase CNN's data efficiency and performance. However, the parameter-sharing strategy greatly increases the computational burden for each added parameter, which hampers its application to deep neural network models. In this paper, we address these problems by proposing a non-parameter-sharing approach for group equivariant neural networks. The proposed methods adaptively aggregate a diverse range of filters by a weighted sum of stochastically augmented decomposed filters. We give theoretical proof about how the continuous group convolution can be approximated by our methods. Our method applies to both continuous and discrete groups, where the augmentation is implemented using Monte Carlo sampling and bootstrap resampling, respectively. We demonstrate that our methods serve as an efficient extension of standard CNN. Experiments on group equivariance tests show how our methods can achieve superior performance to parameter-sharing group equivariant networks. Experiments on image classification and image denoising tasks show that in certain scenarios, with a suitable set of filter bases, our method helps improve the performance of standard CNNs and build efficient lightweight image denoising networks. The code will be available at https://github.com/ZhaoWenzhao/MCG_CNN.
[ { "created": "Wed, 17 May 2023 10:18:02 GMT", "version": "v1" }, { "created": "Sun, 4 Feb 2024 22:22:29 GMT", "version": "v2" }, { "created": "Wed, 1 May 2024 21:54:24 GMT", "version": "v3" } ]
2024-05-03
[ [ "Zhao", "Wenzhao", "" ], [ "Wichtmann", "Barbara D.", "" ], [ "Albert", "Steffen", "" ], [ "Maurer", "Angelika", "" ], [ "Zöllner", "Frank G.", "" ], [ "Attenberger", "Ulrike", "" ], [ "Hesser", "Jürgen", "" ] ]
Group-equivariant convolutional neural networks (G-CNN) heavily rely on parameter sharing to increase CNN's data efficiency and performance. However, the parameter-sharing strategy greatly increases the computational burden for each added parameter, which hampers its application to deep neural network models. In this paper, we address these problems by proposing a non-parameter-sharing approach for group equivariant neural networks. The proposed methods adaptively aggregate a diverse range of filters by a weighted sum of stochastically augmented decomposed filters. We give theoretical proof about how the continuous group convolution can be approximated by our methods. Our method applies to both continuous and discrete groups, where the augmentation is implemented using Monte Carlo sampling and bootstrap resampling, respectively. We demonstrate that our methods serve as an efficient extension of standard CNN. Experiments on group equivariance tests show how our methods can achieve superior performance to parameter-sharing group equivariant networks. Experiments on image classification and image denoising tasks show that in certain scenarios, with a suitable set of filter bases, our method helps improve the performance of standard CNNs and build efficient lightweight image denoising networks. The code will be available at https://github.com/ZhaoWenzhao/MCG_CNN.
2004.10878
Sujit Bhattacharya Professor
Sujit Bhattacharya and Shubham Singh
Visible Insights of the Invisible Pandemic: A Scientometric, Altmetric and Topic Trend Analysis
21 pages, 4 Figures and 4 tables
null
null
null
cs.DL cs.IR
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The recent SARS-COV-2 virus outbreak has created an unprecedented global health crisis! The disease is showing alarming trends with the number of people getting infected with this disease, new cases and death rate are all highlighting the need to control this disease at the earliest. The strategy now for the governments around the globe is how to limit the spread of the virus until the research community develops treatment/drug or vaccination against the virus. The outbreak of this disease has unsurprisingly led to huge volume of research within a short period of time surrounding this disease. It has also led to aggressive social media activity on twitter, Facebook, dedicated blogs, news reports and other online sites actively involved in discussing about the various aspects of and related to this disease. It becomes a useful and challenging exercise to draw from this huge volume of research, the key papers that form the research front, its influence in the research community, and other important research insights. Similarly, it becomes important to discern the key issues that influence the society concerning this disease. The paper is motivated by this. It attempts to distinguish which are the most influential papers, the key knowledge base and major topics surrounding the research covered by COVID-19. Further it attempts to capture the society's perception by discerning key topics that are trending online. The study concludes by highlighting the implications of this study.
[ { "created": "Wed, 22 Apr 2020 21:53:15 GMT", "version": "v1" } ]
2020-04-24
[ [ "Bhattacharya", "Sujit", "" ], [ "Singh", "Shubham", "" ] ]
The recent SARS-COV-2 virus outbreak has created an unprecedented global health crisis! The disease is showing alarming trends with the number of people getting infected with this disease, new cases and death rate are all highlighting the need to control this disease at the earliest. The strategy now for the governments around the globe is how to limit the spread of the virus until the research community develops treatment/drug or vaccination against the virus. The outbreak of this disease has unsurprisingly led to huge volume of research within a short period of time surrounding this disease. It has also led to aggressive social media activity on twitter, Facebook, dedicated blogs, news reports and other online sites actively involved in discussing about the various aspects of and related to this disease. It becomes a useful and challenging exercise to draw from this huge volume of research, the key papers that form the research front, its influence in the research community, and other important research insights. Similarly, it becomes important to discern the key issues that influence the society concerning this disease. The paper is motivated by this. It attempts to distinguish which are the most influential papers, the key knowledge base and major topics surrounding the research covered by COVID-19. Further it attempts to capture the society's perception by discerning key topics that are trending online. The study concludes by highlighting the implications of this study.
2111.06334
Sarthak Khanal
Sarthak Khanal, Maria Traskowsky, Doina Caragea
Identification of Fine-Grained Location Mentions in Crisis Tweets
null
null
null
null
cs.CL cs.LG
http://creativecommons.org/licenses/by/4.0/
Identification of fine-grained location mentions in crisis tweets is central in transforming situational awareness information extracted from social media into actionable information. Most prior works have focused on identifying generic locations, without considering their specific types. To facilitate progress on the fine-grained location identification task, we assemble two tweet crisis datasets and manually annotate them with specific location types. The first dataset contains tweets from a mixed set of crisis events, while the second dataset contains tweets from the global COVID-19 pandemic. We investigate the performance of state-of-the-art deep learning models for sequence tagging on these datasets, in both in-domain and cross-domain settings.
[ { "created": "Thu, 11 Nov 2021 17:48:03 GMT", "version": "v1" } ]
2021-11-12
[ [ "Khanal", "Sarthak", "" ], [ "Traskowsky", "Maria", "" ], [ "Caragea", "Doina", "" ] ]
Identification of fine-grained location mentions in crisis tweets is central in transforming situational awareness information extracted from social media into actionable information. Most prior works have focused on identifying generic locations, without considering their specific types. To facilitate progress on the fine-grained location identification task, we assemble two tweet crisis datasets and manually annotate them with specific location types. The first dataset contains tweets from a mixed set of crisis events, while the second dataset contains tweets from the global COVID-19 pandemic. We investigate the performance of state-of-the-art deep learning models for sequence tagging on these datasets, in both in-domain and cross-domain settings.
cs/0407036
David Eppstein
David Eppstein
All Maximal Independent Sets and Dynamic Dominance for Sparse Graphs
10 pages
ACM Trans. Algorithms 5(4):A38, 2009
10.1145/1597036.1597042
null
cs.DS
null
We describe algorithms, based on Avis and Fukuda's reverse search paradigm, for listing all maximal independent sets in a sparse graph in polynomial time and delay per output. For bounded degree graphs, our algorithms take constant time per set generated; for minor-closed graph families, the time is O(n) per set, and for more general sparse graph families we achieve subquadratic time per set. We also describe new data structures for maintaining a dynamic vertex set S in a sparse or minor-closed graph family, and querying the number of vertices not dominated by S; for minor-closed graph families the time per update is constant, while it is sublinear for any sparse graph family. We can also maintain a dynamic vertex set in an arbitrary m-edge graph and test the independence of the maintained set in time O(sqrt m) per update. We use the domination data structures as part of our enumeration algorithms.
[ { "created": "Thu, 15 Jul 2004 21:04:45 GMT", "version": "v1" } ]
2010-01-11
[ [ "Eppstein", "David", "" ] ]
We describe algorithms, based on Avis and Fukuda's reverse search paradigm, for listing all maximal independent sets in a sparse graph in polynomial time and delay per output. For bounded degree graphs, our algorithms take constant time per set generated; for minor-closed graph families, the time is O(n) per set, and for more general sparse graph families we achieve subquadratic time per set. We also describe new data structures for maintaining a dynamic vertex set S in a sparse or minor-closed graph family, and querying the number of vertices not dominated by S; for minor-closed graph families the time per update is constant, while it is sublinear for any sparse graph family. We can also maintain a dynamic vertex set in an arbitrary m-edge graph and test the independence of the maintained set in time O(sqrt m) per update. We use the domination data structures as part of our enumeration algorithms.
2104.11298
Siddhartha Jayanti
Siddhartha Jayanti
Nash Equilibria of The Multiplayer Colonel Blotto Game on Arbitrary Measure Spaces
19 pages
null
null
null
cs.GT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The Colonel Blotto Problem proposed by Borel in 1921 has served as a widely applicable model of budget-constrained simultaneous winner-take-all competitions in the social sciences. Applications include elections, advertising, R&D and more. However, the classic Blotto problem and variants limit the study to competitions over a finite set of discrete battlefields. In this paper, we extend the classical theory to study multiplayer Blotto games over arbitrary measurable battlegrounds, provide an algorithm to efficiently sample equilibria of symmetric "equipartionable" Generalized Blotto games, and characterize the symmetric fair equilibria of the Blotto game over the unit interval.
[ { "created": "Thu, 22 Apr 2021 19:52:47 GMT", "version": "v1" } ]
2021-04-26
[ [ "Jayanti", "Siddhartha", "" ] ]
The Colonel Blotto Problem proposed by Borel in 1921 has served as a widely applicable model of budget-constrained simultaneous winner-take-all competitions in the social sciences. Applications include elections, advertising, R&D and more. However, the classic Blotto problem and variants limit the study to competitions over a finite set of discrete battlefields. In this paper, we extend the classical theory to study multiplayer Blotto games over arbitrary measurable battlegrounds, provide an algorithm to efficiently sample equilibria of symmetric "equipartionable" Generalized Blotto games, and characterize the symmetric fair equilibria of the Blotto game over the unit interval.
2212.14129
Bryan Ford
Bryan Ford
Matchertext: Towards Verbatim Interlanguage Embedding
23 pages, 4 figures, 2 tables
null
null
null
cs.PL
http://creativecommons.org/licenses/by/4.0/
Embedding text in one language within text of another is commonplace for numerous purposes, but usually requires tedious and error-prone "escaping" transformations on the embedded string. We propose a simple cross-language syntactic discipline, matchertext, which enables the safe embedding a string in any compliant language into a string in any other language via simple "copy-and-paste" - in particular with no escaping, obfuscation, or expansion of embedded strings. We apply this syntactic discipline to several common and frequently-embedded language syntaxes such as URIs, HTML, and JavaScript, exploring the benefits, costs, and compatibility issues in adopting the proposed matchertext discipline. One early matchertext-based language is MinML, a concise but general alternative syntax for writing HTML or XML.
[ { "created": "Thu, 29 Dec 2022 00:10:31 GMT", "version": "v1" } ]
2023-01-02
[ [ "Ford", "Bryan", "" ] ]
Embedding text in one language within text of another is commonplace for numerous purposes, but usually requires tedious and error-prone "escaping" transformations on the embedded string. We propose a simple cross-language syntactic discipline, matchertext, which enables the safe embedding a string in any compliant language into a string in any other language via simple "copy-and-paste" - in particular with no escaping, obfuscation, or expansion of embedded strings. We apply this syntactic discipline to several common and frequently-embedded language syntaxes such as URIs, HTML, and JavaScript, exploring the benefits, costs, and compatibility issues in adopting the proposed matchertext discipline. One early matchertext-based language is MinML, a concise but general alternative syntax for writing HTML or XML.
2311.04076
Lindia Tjuatja
Lindia Tjuatja, Valerie Chen, Sherry Tongshuang Wu, Ameet Talwalkar, Graham Neubig
Do LLMs exhibit human-like response biases? A case study in survey design
null
null
null
null
cs.CL
http://creativecommons.org/licenses/by/4.0/
As large language models (LLMs) become more capable, there is growing excitement about the possibility of using LLMs as proxies for humans in real-world tasks where subjective labels are desired, such as in surveys and opinion polling. One widely-cited barrier to the adoption of LLMs as proxies for humans in subjective tasks is their sensitivity to prompt wording - but interestingly, humans also display sensitivities to instruction changes in the form of response biases. We investigate the extent to which LLMs reflect human response biases, if at all. We look to survey design, where human response biases caused by changes in the wordings of "prompts" have been extensively explored in social psychology literature. Drawing from these works, we design a dataset and framework to evaluate whether LLMs exhibit human-like response biases in survey questionnaires. Our comprehensive evaluation of nine models shows that popular open and commercial LLMs generally fail to reflect human-like behavior, particularly in models that have undergone RLHF. Furthermore, even if a model shows a significant change in the same direction as humans, we find that they are sensitive to perturbations that do not elicit significant changes in humans. These results highlight the pitfalls of using LLMs as human proxies, and underscore the need for finer-grained characterizations of model behavior. Our code, dataset, and collected samples are available at https://github.com/lindiatjuatja/BiasMonkey
[ { "created": "Tue, 7 Nov 2023 15:40:43 GMT", "version": "v1" }, { "created": "Wed, 29 Nov 2023 22:00:12 GMT", "version": "v2" }, { "created": "Mon, 15 Jan 2024 17:52:31 GMT", "version": "v3" }, { "created": "Mon, 5 Feb 2024 15:12:06 GMT", "version": "v4" }, { "created": "Tue, 6 Feb 2024 04:16:17 GMT", "version": "v5" } ]
2024-02-07
[ [ "Tjuatja", "Lindia", "" ], [ "Chen", "Valerie", "" ], [ "Wu", "Sherry Tongshuang", "" ], [ "Talwalkar", "Ameet", "" ], [ "Neubig", "Graham", "" ] ]
As large language models (LLMs) become more capable, there is growing excitement about the possibility of using LLMs as proxies for humans in real-world tasks where subjective labels are desired, such as in surveys and opinion polling. One widely-cited barrier to the adoption of LLMs as proxies for humans in subjective tasks is their sensitivity to prompt wording - but interestingly, humans also display sensitivities to instruction changes in the form of response biases. We investigate the extent to which LLMs reflect human response biases, if at all. We look to survey design, where human response biases caused by changes in the wordings of "prompts" have been extensively explored in social psychology literature. Drawing from these works, we design a dataset and framework to evaluate whether LLMs exhibit human-like response biases in survey questionnaires. Our comprehensive evaluation of nine models shows that popular open and commercial LLMs generally fail to reflect human-like behavior, particularly in models that have undergone RLHF. Furthermore, even if a model shows a significant change in the same direction as humans, we find that they are sensitive to perturbations that do not elicit significant changes in humans. These results highlight the pitfalls of using LLMs as human proxies, and underscore the need for finer-grained characterizations of model behavior. Our code, dataset, and collected samples are available at https://github.com/lindiatjuatja/BiasMonkey
cs/0606057
Fredrik Kuivinen
Fredrik Kuivinen
Approximability of Bounded Occurrence Max Ones
Accepted to MFCS 2006
null
null
null
cs.CC
null
We study the approximability of Max Ones when the number of variable occurrences is bounded by a constant. For conservative constraint languages (i.e., when the unary relations are included) we give a complete classification when the number of occurrences is three or more and a partial classification when the bound is two. For the non-conservative case we prove that it is either trivial or equivalent to the corresponding conservative problem under polynomial-time many-one reductions.
[ { "created": "Tue, 13 Jun 2006 06:44:21 GMT", "version": "v1" } ]
2007-05-23
[ [ "Kuivinen", "Fredrik", "" ] ]
We study the approximability of Max Ones when the number of variable occurrences is bounded by a constant. For conservative constraint languages (i.e., when the unary relations are included) we give a complete classification when the number of occurrences is three or more and a partial classification when the bound is two. For the non-conservative case we prove that it is either trivial or equivalent to the corresponding conservative problem under polynomial-time many-one reductions.
2103.11528
Son T. Luu
Son T. Luu, Kiet Van Nguyen and Ngan Luu-Thuy Nguyen
A Large-scale Dataset for Hate Speech Detection on Vietnamese Social Media Texts
IEA/AIE 2021: Advances and Trends in Artificial Intelligence. Artificial Intelligence Practices, pp 415-426
null
10.1007/978-3-030-79457-6_35
null
cs.CL
http://creativecommons.org/licenses/by-nc-nd/4.0/
In recent years, Vietnam witnesses the mass development of social network users on different social platforms such as Facebook, Youtube, Instagram, and Tiktok. On social medias, hate speech has become a critical problem for social network users. To solve this problem, we introduce the ViHSD - a human-annotated dataset for automatically detecting hate speech on the social network. This dataset contains over 30,000 comments, each comment in the dataset has one of three labels: CLEAN, OFFENSIVE, or HATE. Besides, we introduce the data creation process for annotating and evaluating the quality of the dataset. Finally, we evaluated the dataset by deep learning models and transformer models.
[ { "created": "Mon, 22 Mar 2021 00:55:47 GMT", "version": "v1" }, { "created": "Mon, 29 Mar 2021 02:46:47 GMT", "version": "v2" }, { "created": "Mon, 5 Apr 2021 09:29:18 GMT", "version": "v3" }, { "created": "Tue, 20 Jul 2021 06:22:08 GMT", "version": "v4" } ]
2021-07-21
[ [ "Luu", "Son T.", "" ], [ "Van Nguyen", "Kiet", "" ], [ "Nguyen", "Ngan Luu-Thuy", "" ] ]
In recent years, Vietnam witnesses the mass development of social network users on different social platforms such as Facebook, Youtube, Instagram, and Tiktok. On social medias, hate speech has become a critical problem for social network users. To solve this problem, we introduce the ViHSD - a human-annotated dataset for automatically detecting hate speech on the social network. This dataset contains over 30,000 comments, each comment in the dataset has one of three labels: CLEAN, OFFENSIVE, or HATE. Besides, we introduce the data creation process for annotating and evaluating the quality of the dataset. Finally, we evaluated the dataset by deep learning models and transformer models.
2210.05391
Ruoyu Guo
Chenxia Li, Ruoyu Guo, Jun Zhou, Mengtao An, Yuning Du, Lingfeng Zhu, Yi Liu, Xiaoguang Hu, Dianhai Yu
PP-StructureV2: A Stronger Document Analysis System
null
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
A large amount of document data exists in unstructured form such as raw images without any text information. Designing a practical document image analysis system is a meaningful but challenging task. In previous work, we proposed an intelligent document analysis system PP-Structure. In order to further upgrade the function and performance of PP-Structure, we propose PP-StructureV2 in this work, which contains two subsystems: Layout Information Extraction and Key Information Extraction. Firstly, we integrate Image Direction Correction module and Layout Restoration module to enhance the functionality of the system. Secondly, 8 practical strategies are utilized in PP-StructureV2 for better performance. For Layout Analysis model, we introduce ultra light-weight detector PP-PicoDet and knowledge distillation algorithm FGD for model lightweighting, which increased the inference speed by 11 times with comparable mAP. For Table Recognition model, we utilize PP-LCNet, CSP-PAN and SLAHead to optimize the backbone module, feature fusion module and decoding module, respectively, which improved the table structure accuracy by 6\% with comparable inference speed. For Key Information Extraction model, we introduce VI-LayoutXLM which is a visual-feature independent LayoutXLM architecture, TB-YX sorting algorithm and U-DML knowledge distillation algorithm, which brought 2.8\% and 9.1\% improvement respectively on the Hmean of Semantic Entity Recognition and Relation Extraction tasks. All the above mentioned models and code are open-sourced in the GitHub repository PaddleOCR.
[ { "created": "Tue, 11 Oct 2022 12:07:32 GMT", "version": "v1" }, { "created": "Thu, 13 Oct 2022 07:11:59 GMT", "version": "v2" } ]
2022-10-14
[ [ "Li", "Chenxia", "" ], [ "Guo", "Ruoyu", "" ], [ "Zhou", "Jun", "" ], [ "An", "Mengtao", "" ], [ "Du", "Yuning", "" ], [ "Zhu", "Lingfeng", "" ], [ "Liu", "Yi", "" ], [ "Hu", "Xiaoguang", "" ], [ "Yu", "Dianhai", "" ] ]
A large amount of document data exists in unstructured form such as raw images without any text information. Designing a practical document image analysis system is a meaningful but challenging task. In previous work, we proposed an intelligent document analysis system PP-Structure. In order to further upgrade the function and performance of PP-Structure, we propose PP-StructureV2 in this work, which contains two subsystems: Layout Information Extraction and Key Information Extraction. Firstly, we integrate Image Direction Correction module and Layout Restoration module to enhance the functionality of the system. Secondly, 8 practical strategies are utilized in PP-StructureV2 for better performance. For Layout Analysis model, we introduce ultra light-weight detector PP-PicoDet and knowledge distillation algorithm FGD for model lightweighting, which increased the inference speed by 11 times with comparable mAP. For Table Recognition model, we utilize PP-LCNet, CSP-PAN and SLAHead to optimize the backbone module, feature fusion module and decoding module, respectively, which improved the table structure accuracy by 6\% with comparable inference speed. For Key Information Extraction model, we introduce VI-LayoutXLM which is a visual-feature independent LayoutXLM architecture, TB-YX sorting algorithm and U-DML knowledge distillation algorithm, which brought 2.8\% and 9.1\% improvement respectively on the Hmean of Semantic Entity Recognition and Relation Extraction tasks. All the above mentioned models and code are open-sourced in the GitHub repository PaddleOCR.
1912.11576
Sheng Zhou
Yining Xu, Sheng Zhou
On the Coverage and Capacity of Ultra-Dense Networks with Directional Transmissions
5 pages, 4 figures, accepted by IEEE Wireless Commuincations Letters
null
null
null
cs.IT math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We investigate the performance of a downlink ultra-dense network (UDN) with directional transmissions via stochastic geometry. Considering the dual-slope path loss model and sectored beamforming pattern, we derive the expressions and asymptotic characteristics of the coverage probability and constrained area spectrum efficiency (ASE). Several special scenarios, namely the physically feasible path loss model and adjustable beam pattern, are also analyzed. Although signal-to-interference-plus-noise ratio collapsing still exists when the path loss exponent in the near-field is no larger than 2, using strategies like beam pattern adaption, can avoid the decrease of the coverage probability and constrained ASE even when the base station density approaches infinity.
[ { "created": "Wed, 25 Dec 2019 01:59:04 GMT", "version": "v1" } ]
2019-12-30
[ [ "Xu", "Yining", "" ], [ "Zhou", "Sheng", "" ] ]
We investigate the performance of a downlink ultra-dense network (UDN) with directional transmissions via stochastic geometry. Considering the dual-slope path loss model and sectored beamforming pattern, we derive the expressions and asymptotic characteristics of the coverage probability and constrained area spectrum efficiency (ASE). Several special scenarios, namely the physically feasible path loss model and adjustable beam pattern, are also analyzed. Although signal-to-interference-plus-noise ratio collapsing still exists when the path loss exponent in the near-field is no larger than 2, using strategies like beam pattern adaption, can avoid the decrease of the coverage probability and constrained ASE even when the base station density approaches infinity.
1203.1754
Marek Cygan
Marek Cygan and Marcin Pilipczuk and Micha{\l} Pilipczuk
Known algorithms for EDGE CLIQUE COVER are probably optimal
To appear in SODA 2013
null
null
null
cs.DS cs.CC cs.DM
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In the EDGE CLIQUE COVER (ECC) problem, given a graph G and an integer k, we ask whether the edges of G can be covered with k complete subgraphs of G or, equivalently, whether G admits an intersection model on k-element universe. Gramm et al. [JEA 2008] have shown a set of simple rules that reduce the number of vertices of G to 2^k, and no algorithm is known with significantly better running time bound than a brute-force search on this reduced instance. In this paper we show that the approach of Gramm et al. is essentially optimal: we present a polynomial time algorithm that reduces an arbitrary 3-CNF-SAT formula with n variables and m clauses to an equivalent ECC instance (G,k) with k = O(log n) and |V(G)| = O(n + m). Consequently, there is no 2^{2^{o(k)}}poly(n) time algorithm for the ECC problem, unless the Exponential Time Hypothesis fails. To the best of our knowledge, these are the first results for a natural, fixed-parameter tractable problem, and proving that a doubly-exponential dependency on the parameter is essentially necessary.
[ { "created": "Thu, 8 Mar 2012 11:19:09 GMT", "version": "v1" }, { "created": "Wed, 26 Sep 2012 08:51:46 GMT", "version": "v2" } ]
2012-09-27
[ [ "Cygan", "Marek", "" ], [ "Pilipczuk", "Marcin", "" ], [ "Pilipczuk", "Michał", "" ] ]
In the EDGE CLIQUE COVER (ECC) problem, given a graph G and an integer k, we ask whether the edges of G can be covered with k complete subgraphs of G or, equivalently, whether G admits an intersection model on k-element universe. Gramm et al. [JEA 2008] have shown a set of simple rules that reduce the number of vertices of G to 2^k, and no algorithm is known with significantly better running time bound than a brute-force search on this reduced instance. In this paper we show that the approach of Gramm et al. is essentially optimal: we present a polynomial time algorithm that reduces an arbitrary 3-CNF-SAT formula with n variables and m clauses to an equivalent ECC instance (G,k) with k = O(log n) and |V(G)| = O(n + m). Consequently, there is no 2^{2^{o(k)}}poly(n) time algorithm for the ECC problem, unless the Exponential Time Hypothesis fails. To the best of our knowledge, these are the first results for a natural, fixed-parameter tractable problem, and proving that a doubly-exponential dependency on the parameter is essentially necessary.
2304.07493
Cong Guo
Cong Guo, Jiaming Tang, Weiming Hu, Jingwen Leng, Chen Zhang, Fan Yang, Yunxin Liu, Minyi Guo, Yuhao Zhu
OliVe: Accelerating Large Language Models via Hardware-friendly Outlier-Victim Pair Quantization
ISCA 2023
null
10.1145/3579371.3589038
null
cs.AR
http://creativecommons.org/licenses/by/4.0/
Transformer-based large language models (LLMs) have achieved great success with the growing model size. LLMs' size grows by $240\times$ every two years, which outpaces the hardware progress and makes model inference increasingly costly. Model quantization is a promising approach to mitigate the widening gap between LLM size and hardware capacity. However, the existence of outliers, values with significant magnitudes, in LLMs makes existing quantization methods less effective. Prior outlier-aware quantization schemes adopt sparsity encoding techniques to separate outliers from normal values where the process requires global coordination (e.g., a global sparsity coordination list). This incurs complex encoding/decoding hardware logics and an extra orchestration controller for the computation between outlier and normal values. As such, it is not hardware-efficient and hence only achieves sub-optimal quantization benefits. We propose OliVe, an algorithm/architecture co-designed solution that adopts an outlier-victim pair (OVP) quantization and handles outlier values locally with low hardware overheads and high performance gains. The key insight of OliVe is that outliers are important while the normal values next to them are not. Thus those normal values (called victims) can be sacrificed to accommodate outliers. This enables a memory-aligned OVP encoding scheme, which can be efficiently integrated to the existing hardware accelerators like systolic array and tensor core. As a result, OliVe-based accelerator surpasses the existing outlier-aware accelerator, GOBO, by 4.5$\times$ speedup and 4.0$\times$ energy reduction, respectively, with a superior model accuracy.
[ { "created": "Sat, 15 Apr 2023 07:12:05 GMT", "version": "v1" } ]
2023-04-18
[ [ "Guo", "Cong", "" ], [ "Tang", "Jiaming", "" ], [ "Hu", "Weiming", "" ], [ "Leng", "Jingwen", "" ], [ "Zhang", "Chen", "" ], [ "Yang", "Fan", "" ], [ "Liu", "Yunxin", "" ], [ "Guo", "Minyi", "" ], [ "Zhu", "Yuhao", "" ] ]
Transformer-based large language models (LLMs) have achieved great success with the growing model size. LLMs' size grows by $240\times$ every two years, which outpaces the hardware progress and makes model inference increasingly costly. Model quantization is a promising approach to mitigate the widening gap between LLM size and hardware capacity. However, the existence of outliers, values with significant magnitudes, in LLMs makes existing quantization methods less effective. Prior outlier-aware quantization schemes adopt sparsity encoding techniques to separate outliers from normal values where the process requires global coordination (e.g., a global sparsity coordination list). This incurs complex encoding/decoding hardware logics and an extra orchestration controller for the computation between outlier and normal values. As such, it is not hardware-efficient and hence only achieves sub-optimal quantization benefits. We propose OliVe, an algorithm/architecture co-designed solution that adopts an outlier-victim pair (OVP) quantization and handles outlier values locally with low hardware overheads and high performance gains. The key insight of OliVe is that outliers are important while the normal values next to them are not. Thus those normal values (called victims) can be sacrificed to accommodate outliers. This enables a memory-aligned OVP encoding scheme, which can be efficiently integrated to the existing hardware accelerators like systolic array and tensor core. As a result, OliVe-based accelerator surpasses the existing outlier-aware accelerator, GOBO, by 4.5$\times$ speedup and 4.0$\times$ energy reduction, respectively, with a superior model accuracy.
2405.07946
Qiang Zou
Yaonaiming Zhao, Qiang Zou, Guoyue Luo, Jiayu Wu, Sifan Chen, Depeng Gao, Minghao Xuan, Fuyu Wang
TPMS2STEP: error-controlled and C2 continuity-preserving translation of TPMS models to STEP files based on constrained-PIA
null
null
null
null
cs.CG
http://creativecommons.org/publicdomain/zero/1.0/
Triply periodic minimal surface (TPMS) is emerging as an important way of designing microstructures. However, there has been limited use of commercial CAD/CAM/CAE software packages for TPMS design and manufacturing. This is mainly because TPMS is consistently described in the functional representation (F-rep) format, while modern CAD/CAM/CAE tools are built upon the boundary representation (B-rep) format. One possible solution to this gap is translating TPMS to STEP, which is the standard data exchange format of CAD/CAM/CAE. Following this direction, this paper proposes a new translation method with error-controlling and $C^2$ continuity-preserving features. It is based on an approximation error-driven TPMS sampling algorithm and a constrained-PIA algorithm. The sampling algorithm controls the deviation between the original and translated models. With it, an error bound of $2\epsilon$ on the deviation can be ensured if two conditions called $\epsilon$-density and $\epsilon$-approximation are satisfied. The constrained-PIA algorithm enforces $C^2$ continuity constraints during TPMS approximation, and meanwhile attaining high efficiency. A theoretical convergence proof of this algorithm is also given. The effectiveness of the translation method has been demonstrated by a series of examples and comparisons.
[ { "created": "Mon, 13 May 2024 17:22:44 GMT", "version": "v1" }, { "created": "Fri, 24 May 2024 02:36:26 GMT", "version": "v2" } ]
2024-05-27
[ [ "Zhao", "Yaonaiming", "" ], [ "Zou", "Qiang", "" ], [ "Luo", "Guoyue", "" ], [ "Wu", "Jiayu", "" ], [ "Chen", "Sifan", "" ], [ "Gao", "Depeng", "" ], [ "Xuan", "Minghao", "" ], [ "Wang", "Fuyu", "" ] ]
Triply periodic minimal surface (TPMS) is emerging as an important way of designing microstructures. However, there has been limited use of commercial CAD/CAM/CAE software packages for TPMS design and manufacturing. This is mainly because TPMS is consistently described in the functional representation (F-rep) format, while modern CAD/CAM/CAE tools are built upon the boundary representation (B-rep) format. One possible solution to this gap is translating TPMS to STEP, which is the standard data exchange format of CAD/CAM/CAE. Following this direction, this paper proposes a new translation method with error-controlling and $C^2$ continuity-preserving features. It is based on an approximation error-driven TPMS sampling algorithm and a constrained-PIA algorithm. The sampling algorithm controls the deviation between the original and translated models. With it, an error bound of $2\epsilon$ on the deviation can be ensured if two conditions called $\epsilon$-density and $\epsilon$-approximation are satisfied. The constrained-PIA algorithm enforces $C^2$ continuity constraints during TPMS approximation, and meanwhile attaining high efficiency. A theoretical convergence proof of this algorithm is also given. The effectiveness of the translation method has been demonstrated by a series of examples and comparisons.
2210.01400
Rui Yuan
Rui Yuan, Simon S. Du, Robert M. Gower, Alessandro Lazaric, Lin Xiao
Linear Convergence of Natural Policy Gradient Methods with Log-Linear Policies
This version adds a table of comparison for the literature review. The paper is published as a conference paper at ICLR 2023
null
null
null
cs.LG cs.AI math.OC
http://creativecommons.org/licenses/by-nc-nd/4.0/
We consider infinite-horizon discounted Markov decision processes and study the convergence rates of the natural policy gradient (NPG) and the Q-NPG methods with the log-linear policy class. Using the compatible function approximation framework, both methods with log-linear policies can be written as inexact versions of the policy mirror descent (PMD) method. We show that both methods attain linear convergence rates and $\tilde{\mathcal{O}}(1/\epsilon^2)$ sample complexities using a simple, non-adaptive geometrically increasing step size, without resorting to entropy or other strongly convex regularization. Lastly, as a byproduct, we obtain sublinear convergence rates for both methods with arbitrary constant step size.
[ { "created": "Tue, 4 Oct 2022 06:17:52 GMT", "version": "v1" }, { "created": "Mon, 21 Nov 2022 12:58:36 GMT", "version": "v2" }, { "created": "Tue, 21 Feb 2023 14:48:00 GMT", "version": "v3" } ]
2023-02-22
[ [ "Yuan", "Rui", "" ], [ "Du", "Simon S.", "" ], [ "Gower", "Robert M.", "" ], [ "Lazaric", "Alessandro", "" ], [ "Xiao", "Lin", "" ] ]
We consider infinite-horizon discounted Markov decision processes and study the convergence rates of the natural policy gradient (NPG) and the Q-NPG methods with the log-linear policy class. Using the compatible function approximation framework, both methods with log-linear policies can be written as inexact versions of the policy mirror descent (PMD) method. We show that both methods attain linear convergence rates and $\tilde{\mathcal{O}}(1/\epsilon^2)$ sample complexities using a simple, non-adaptive geometrically increasing step size, without resorting to entropy or other strongly convex regularization. Lastly, as a byproduct, we obtain sublinear convergence rates for both methods with arbitrary constant step size.
2110.05305
Pascal Koiran
Pascal Koiran and Subhayan Saha
Black Box Absolute Reconstruction for Sums of Powers of Linear Forms
null
null
null
null
cs.CC cs.DS
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We study the decomposition of multivariate polynomials as sums of powers of linear forms. We give a randomized algorithm for the following problem: If a homogeneous polynomial $f \in K[x_1 , . . . , x_n]$ (where $K \subseteq \mathbb{C}$) of degree $d$ is given as a blackbox, decide whether it can be written as a linear combination of $d$-th powers of linearly independent complex linear forms. The main novel features of the algorithm are: (1) For $d = 3$, we improve by a factor of $n$ on the running time from an algorithm by Koiran and Skomra. The price to be paid for this improvement though is that the algorithm now has two-sided error. (2) For $d > 3$, we provide the first randomized blackbox algorithm for this problem that runs in time polynomial in $n$ and $d$ (in an algebraic model where only arithmetic operations and equality tests are allowed). Previous algorithms for this problem as well as most of the existing reconstruction algorithms for other classes appeal to a polynomial factorization subroutine. This requires extraction of complex polynomial roots at unit cost and in standard models such as the unit-cost RAM or the Turing machine this approach does not yield polynomial time algorithms. (3) For $d > 3$, when $f$ has rational coefficients, the running time of the blackbox algorithm is polynomial in $n,d$ and the maximal bit size of any coefficient of $f$. This yields the first algorithm for this problem over $\mathbb{C}$ with polynomial running time in the bit model of computation.
[ { "created": "Mon, 11 Oct 2021 14:25:24 GMT", "version": "v1" } ]
2021-10-12
[ [ "Koiran", "Pascal", "" ], [ "Saha", "Subhayan", "" ] ]
We study the decomposition of multivariate polynomials as sums of powers of linear forms. We give a randomized algorithm for the following problem: If a homogeneous polynomial $f \in K[x_1 , . . . , x_n]$ (where $K \subseteq \mathbb{C}$) of degree $d$ is given as a blackbox, decide whether it can be written as a linear combination of $d$-th powers of linearly independent complex linear forms. The main novel features of the algorithm are: (1) For $d = 3$, we improve by a factor of $n$ on the running time from an algorithm by Koiran and Skomra. The price to be paid for this improvement though is that the algorithm now has two-sided error. (2) For $d > 3$, we provide the first randomized blackbox algorithm for this problem that runs in time polynomial in $n$ and $d$ (in an algebraic model where only arithmetic operations and equality tests are allowed). Previous algorithms for this problem as well as most of the existing reconstruction algorithms for other classes appeal to a polynomial factorization subroutine. This requires extraction of complex polynomial roots at unit cost and in standard models such as the unit-cost RAM or the Turing machine this approach does not yield polynomial time algorithms. (3) For $d > 3$, when $f$ has rational coefficients, the running time of the blackbox algorithm is polynomial in $n,d$ and the maximal bit size of any coefficient of $f$. This yields the first algorithm for this problem over $\mathbb{C}$ with polynomial running time in the bit model of computation.
2302.09813
Juexiao Zhou
Juexiao Zhou, Haoyang Li, Xingyu Liao, Bin Zhang, Wenjia He, Zhongxiao Li, Longxi Zhou, Xin Gao
Audit to Forget: A Unified Method to Revoke Patients' Private Data in Intelligent Healthcare
null
null
10.1038/s41467-023-41703-x
null
cs.LG cs.CR
http://creativecommons.org/licenses/by-nc-nd/4.0/
Revoking personal private data is one of the basic human rights, which has already been sheltered by several privacy-preserving laws in many countries. However, with the development of data science, machine learning and deep learning techniques, this right is usually neglected or violated as more and more patients' data are being collected and used for model training, especially in intelligent healthcare, thus making intelligent healthcare a sector where technology must meet the law, regulations, and privacy principles to ensure that the innovation is for the common good. In order to secure patients' right to be forgotten, we proposed a novel solution by using auditing to guide the forgetting process, where auditing means determining whether a dataset has been used to train the model and forgetting requires the information of a query dataset to be forgotten from the target model. We unified these two tasks by introducing a new approach called knowledge purification. To implement our solution, we developed AFS, a unified open-source software, which is able to evaluate and revoke patients' private data from pre-trained deep learning models. We demonstrated the generality of AFS by applying it to four tasks on different datasets with various data sizes and architectures of deep learning networks. The software is publicly available at \url{https://github.com/JoshuaChou2018/AFS}.
[ { "created": "Mon, 20 Feb 2023 07:29:22 GMT", "version": "v1" } ]
2024-01-12
[ [ "Zhou", "Juexiao", "" ], [ "Li", "Haoyang", "" ], [ "Liao", "Xingyu", "" ], [ "Zhang", "Bin", "" ], [ "He", "Wenjia", "" ], [ "Li", "Zhongxiao", "" ], [ "Zhou", "Longxi", "" ], [ "Gao", "Xin", "" ] ]
Revoking personal private data is one of the basic human rights, which has already been sheltered by several privacy-preserving laws in many countries. However, with the development of data science, machine learning and deep learning techniques, this right is usually neglected or violated as more and more patients' data are being collected and used for model training, especially in intelligent healthcare, thus making intelligent healthcare a sector where technology must meet the law, regulations, and privacy principles to ensure that the innovation is for the common good. In order to secure patients' right to be forgotten, we proposed a novel solution by using auditing to guide the forgetting process, where auditing means determining whether a dataset has been used to train the model and forgetting requires the information of a query dataset to be forgotten from the target model. We unified these two tasks by introducing a new approach called knowledge purification. To implement our solution, we developed AFS, a unified open-source software, which is able to evaluate and revoke patients' private data from pre-trained deep learning models. We demonstrated the generality of AFS by applying it to four tasks on different datasets with various data sizes and architectures of deep learning networks. The software is publicly available at \url{https://github.com/JoshuaChou2018/AFS}.
cs/0703030
Konstantin Rybnikov
Konstantin Rybnikov
An Efficient Local Approach to Convexity Testing of Piecewise-Linear Hypersurfaces
3 figures
null
null
null
cs.CG
null
We show that a closed piecewise-linear hypersurface immersed in $R^n$ ($n\ge 3$) is the boundary of a convex body if and only if every point in the interior of each $(n-3)$-face has a neighborhood that lies on the boundary of some convex body; no assumptions about the hypersurface's topology are needed. We derive this criterion from our generalization of Van Heijenoort's (1952) theorem on locally convex hypersurfaces in $R^n$ to spherical spaces. We also give an easy-to-implement convexity testing algorithm, which is based on our criterion. For $R^3$ the number of arithmetic operations used by the algorithm is at most linear in the number of vertices, while in general it is at most linear in the number of incidences between the $(n-2)$-faces and $(n-3)$-faces. When the dimension $n$ is not fixed and only ring arithmetic is allowed, the algorithm still remains polynomial. Our method works in more general situations than the convexity verification algorithms developed by Mehlhorn et al. (1996) and Devillers et al. (1998) -- for example, our method does not require the input surface to be orientable, nor it requires the input data to include normal vectors to the facets that are oriented "in a coherent way". For $R^3$ the complexity of our algorithm is the same as that of previous algorithms; for higher dimensions there seems to be no clear winner, but our approach is the only one that easily handles inputs in which the facet normals are not known to be coherently oriented or are not given at all. Furthermore, our method can be extended to piecewise-polynomial surfaces of small degree.
[ { "created": "Wed, 7 Mar 2007 07:33:02 GMT", "version": "v1" } ]
2007-05-23
[ [ "Rybnikov", "Konstantin", "" ] ]
We show that a closed piecewise-linear hypersurface immersed in $R^n$ ($n\ge 3$) is the boundary of a convex body if and only if every point in the interior of each $(n-3)$-face has a neighborhood that lies on the boundary of some convex body; no assumptions about the hypersurface's topology are needed. We derive this criterion from our generalization of Van Heijenoort's (1952) theorem on locally convex hypersurfaces in $R^n$ to spherical spaces. We also give an easy-to-implement convexity testing algorithm, which is based on our criterion. For $R^3$ the number of arithmetic operations used by the algorithm is at most linear in the number of vertices, while in general it is at most linear in the number of incidences between the $(n-2)$-faces and $(n-3)$-faces. When the dimension $n$ is not fixed and only ring arithmetic is allowed, the algorithm still remains polynomial. Our method works in more general situations than the convexity verification algorithms developed by Mehlhorn et al. (1996) and Devillers et al. (1998) -- for example, our method does not require the input surface to be orientable, nor it requires the input data to include normal vectors to the facets that are oriented "in a coherent way". For $R^3$ the complexity of our algorithm is the same as that of previous algorithms; for higher dimensions there seems to be no clear winner, but our approach is the only one that easily handles inputs in which the facet normals are not known to be coherently oriented or are not given at all. Furthermore, our method can be extended to piecewise-polynomial surfaces of small degree.
2207.09159
Ahmed El Kerim
Ahmed El Kerim and Pierre Gosselet and Frederic Magoules
Couplage Global-Local en asynchrone pour des probl\`emes lin\'eaires
in French language
null
null
null
cs.DC
http://creativecommons.org/licenses/by-nc-nd/4.0/
An asynchronous parallel version of the non-intrusive global-local coupling is implemented. The case of many patches, including those covering the entire structure, is studied. The asynchronism limits the dependency on communications, failures, and load imbalance. We detail the method and illustrate its performance in an academic case.
[ { "created": "Tue, 19 Jul 2022 09:59:12 GMT", "version": "v1" } ]
2022-07-20
[ [ "Kerim", "Ahmed El", "" ], [ "Gosselet", "Pierre", "" ], [ "Magoules", "Frederic", "" ] ]
An asynchronous parallel version of the non-intrusive global-local coupling is implemented. The case of many patches, including those covering the entire structure, is studied. The asynchronism limits the dependency on communications, failures, and load imbalance. We detail the method and illustrate its performance in an academic case.
1811.08225
Danilo Vasconcellos Vargas
Danilo Vasconcellos Vargas and Hirotaka Takano and Junichi Murata
Self Organizing Classifiers: First Steps in Structured Evolutionary Machine Learning
null
Evolutionary Intelligence 6 (2), 57-72 (2013)
null
null
cs.NE cs.AI cs.LG cs.MA
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Learning classifier systems (LCSs) are evolutionary machine learning algorithms, flexible enough to be applied to reinforcement, supervised and unsupervised learning problems with good performance. Recently, self organizing classifiers were proposed which are similar to LCSs but have the advantage that in its structured population no balance between niching and fitness pressure is necessary. However, more tests and analysis are required to verify its benefits. Here, a variation of the first algorithm is proposed which uses a parameterless self organizing map (SOM). This algorithm is applied in challenging problems such as big, noisy as well as dynamically changing continuous input-action mazes (growing and compressing mazes are included) with good performance. Moreover, a genetic operator is proposed which utilizes the topological information of the SOM's population structure, improving the results. Thus, the first steps in structured evolutionary machine learning are shown, nonetheless, the problems faced are more difficult than the state-of-art continuous input-action multi-step ones.
[ { "created": "Tue, 20 Nov 2018 13:00:51 GMT", "version": "v1" } ]
2018-11-21
[ [ "Vargas", "Danilo Vasconcellos", "" ], [ "Takano", "Hirotaka", "" ], [ "Murata", "Junichi", "" ] ]
Learning classifier systems (LCSs) are evolutionary machine learning algorithms, flexible enough to be applied to reinforcement, supervised and unsupervised learning problems with good performance. Recently, self organizing classifiers were proposed which are similar to LCSs but have the advantage that in its structured population no balance between niching and fitness pressure is necessary. However, more tests and analysis are required to verify its benefits. Here, a variation of the first algorithm is proposed which uses a parameterless self organizing map (SOM). This algorithm is applied in challenging problems such as big, noisy as well as dynamically changing continuous input-action mazes (growing and compressing mazes are included) with good performance. Moreover, a genetic operator is proposed which utilizes the topological information of the SOM's population structure, improving the results. Thus, the first steps in structured evolutionary machine learning are shown, nonetheless, the problems faced are more difficult than the state-of-art continuous input-action multi-step ones.
1210.6719
Jun Muramatsu
Jun Muramatsu and Shigeki Miyake
Construction of Multiple Access Channel Codes Based on Hash Property
This paper has been presented in part at Proc. 2011 IEEE Internal Symposium on Information Theory and submitted to IEEE Transactions on Information Theory. 39 pages
null
null
null
cs.IT math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The aim of this paper is to introduce the construction of codes for a general discrete stationary memoryless multiple access channel based on the the notion of the hash property. Since an ensemble of sparse matrices has a hash property, we can use sparse matrices for code construction. Our approach has a potential advantage compared to the conventional random coding because it is expected that we can use some approximation algorithms by using the sparse structure of codes.
[ { "created": "Thu, 25 Oct 2012 01:34:37 GMT", "version": "v1" } ]
2012-10-26
[ [ "Muramatsu", "Jun", "" ], [ "Miyake", "Shigeki", "" ] ]
The aim of this paper is to introduce the construction of codes for a general discrete stationary memoryless multiple access channel based on the the notion of the hash property. Since an ensemble of sparse matrices has a hash property, we can use sparse matrices for code construction. Our approach has a potential advantage compared to the conventional random coding because it is expected that we can use some approximation algorithms by using the sparse structure of codes.
2201.07048
Tianyu Fang
Tianyu Fang, Yijie Mao, Shanpu Shen, Zhencai Zhu, Bruno Clerckx
Fully Connected Reconfigurable Intelligent Surface Aided Rate-Splitting Multiple Access for Multi-User Multi-Antenna Transmission
6 pages, 5figures, conference
null
null
null
cs.IT math.IT
http://creativecommons.org/licenses/by/4.0/
Rate-splitting multiple access (RSMA) has been recognized as a promising and powerful multiple access (MA) scheme, non-orthogonal transmission framework and interference management strategy for 6G. Inspired by the appealing spectral efficiency gain achieved by RSMA over conventional MA schemes in multi-user multi-antenna transmission, in this paper we introduce RSMA to reconfigurable intelligent surface (RIS)-aided multiple-input single-out (MISO) broadcast channel (BC). To further enhance the spectral efficiency, a more generalized RIS architecture called fully connected RIS is considered. By jointly optimizing the scattering matrix of the fully connected RIS and the transmit beamformers to maximize the sum-rate, we show that the proposed fully connected RIS aided RSMA transmission scheme significantly improves the spectral efficiency compared with the conventional single connected RIS schemes and the schemes without RIS. It acts as a new benchmark for linearly precoded multi-user multi-antenna networks.
[ { "created": "Tue, 18 Jan 2022 15:20:02 GMT", "version": "v1" }, { "created": "Thu, 10 Mar 2022 05:19:47 GMT", "version": "v2" } ]
2022-03-11
[ [ "Fang", "Tianyu", "" ], [ "Mao", "Yijie", "" ], [ "Shen", "Shanpu", "" ], [ "Zhu", "Zhencai", "" ], [ "Clerckx", "Bruno", "" ] ]
Rate-splitting multiple access (RSMA) has been recognized as a promising and powerful multiple access (MA) scheme, non-orthogonal transmission framework and interference management strategy for 6G. Inspired by the appealing spectral efficiency gain achieved by RSMA over conventional MA schemes in multi-user multi-antenna transmission, in this paper we introduce RSMA to reconfigurable intelligent surface (RIS)-aided multiple-input single-out (MISO) broadcast channel (BC). To further enhance the spectral efficiency, a more generalized RIS architecture called fully connected RIS is considered. By jointly optimizing the scattering matrix of the fully connected RIS and the transmit beamformers to maximize the sum-rate, we show that the proposed fully connected RIS aided RSMA transmission scheme significantly improves the spectral efficiency compared with the conventional single connected RIS schemes and the schemes without RIS. It acts as a new benchmark for linearly precoded multi-user multi-antenna networks.
1510.02181
Alejandro Erickson
Alejandro Erickson and and Iain A. Stewart and Javier Navaridas and Abbas E. Kiasari
The Stellar Transformation: From Interconnection Networks to Datacenter Networks
Submitted to a journal
null
null
null
cs.DC
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The first dual-port server-centric datacenter network, FiConn, was introduced in 2009 and there are several others now in existence; however, the pool of topologies to choose from remains small. We propose a new generic construction, the stellar transformation, that dramatically increases the size of this pool by facilitating the transformation of well-studied topologies from interconnection networks, along with their networking properties and routing algorithms, into viable dual-port server-centric datacenter network topologies. We demonstrate that under our transformation, numerous interconnection networks yield datacenter network topologies with potentially good, and easily computable, baseline properties. We instantiate our construction so as to apply it to generalized hypercubes and obtain the datacenter networks GQ*. Our construction automatically yields routing algorithms for GQ* and we empirically compare GQ* (and its routing algorithms) with the established datacenter networks FiConn and DPillar (and their routing algorithms); this comparison is with respect to network throughput, latency, load balancing, fault-tolerance, and cost to build, and is with regard to all-to-all, many all-to-all, butterfly, and random traffic patterns. We find that GQ* outperforms both FiConn and DPillar (sometimes significantly so) and that there is substantial scope for our stellar transformation to yield new dual-port server-centric datacenter networks that are a considerable improvement on existing ones.
[ { "created": "Thu, 8 Oct 2015 02:09:29 GMT", "version": "v1" }, { "created": "Mon, 30 Nov 2015 01:16:36 GMT", "version": "v2" }, { "created": "Wed, 17 Feb 2016 01:01:34 GMT", "version": "v3" }, { "created": "Mon, 27 Jun 2016 21:41:20 GMT", "version": "v4" } ]
2016-06-29
[ [ "Erickson", "Alejandro", "" ], [ "Stewart", "and Iain A.", "" ], [ "Navaridas", "Javier", "" ], [ "Kiasari", "Abbas E.", "" ] ]
The first dual-port server-centric datacenter network, FiConn, was introduced in 2009 and there are several others now in existence; however, the pool of topologies to choose from remains small. We propose a new generic construction, the stellar transformation, that dramatically increases the size of this pool by facilitating the transformation of well-studied topologies from interconnection networks, along with their networking properties and routing algorithms, into viable dual-port server-centric datacenter network topologies. We demonstrate that under our transformation, numerous interconnection networks yield datacenter network topologies with potentially good, and easily computable, baseline properties. We instantiate our construction so as to apply it to generalized hypercubes and obtain the datacenter networks GQ*. Our construction automatically yields routing algorithms for GQ* and we empirically compare GQ* (and its routing algorithms) with the established datacenter networks FiConn and DPillar (and their routing algorithms); this comparison is with respect to network throughput, latency, load balancing, fault-tolerance, and cost to build, and is with regard to all-to-all, many all-to-all, butterfly, and random traffic patterns. We find that GQ* outperforms both FiConn and DPillar (sometimes significantly so) and that there is substantial scope for our stellar transformation to yield new dual-port server-centric datacenter networks that are a considerable improvement on existing ones.
2303.07710
Th\'eo Pierron
Nicolas Bousquet, Valentin Gledel, Jonathan Narboni, Th\'eo Pierron
A note on the flip distance between non-crossing spanning trees
null
null
null
null
cs.CG cs.DM math.CO
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We consider spanning trees of $n$ points in convex position whose edges are pairwise non-crossing. Applying a flip to such a tree consists in adding an edge and removing another so that the result is still a non-crossing spanning tree. Given two trees, we investigate the minimum number of flips required to transform one into the other. The naive $2n-\Omega(1)$ upper bound stood for 25 years until a recent breakthrough from Aichholzer et al. yielding a $2n-\Omega(\log n)$ bound. We improve their result with a $2n-\Omega(\sqrt{n})$ upper bound, and we strengthen and shorten the proofs of several of their results.
[ { "created": "Tue, 14 Mar 2023 08:52:36 GMT", "version": "v1" } ]
2023-03-15
[ [ "Bousquet", "Nicolas", "" ], [ "Gledel", "Valentin", "" ], [ "Narboni", "Jonathan", "" ], [ "Pierron", "Théo", "" ] ]
We consider spanning trees of $n$ points in convex position whose edges are pairwise non-crossing. Applying a flip to such a tree consists in adding an edge and removing another so that the result is still a non-crossing spanning tree. Given two trees, we investigate the minimum number of flips required to transform one into the other. The naive $2n-\Omega(1)$ upper bound stood for 25 years until a recent breakthrough from Aichholzer et al. yielding a $2n-\Omega(\log n)$ bound. We improve their result with a $2n-\Omega(\sqrt{n})$ upper bound, and we strengthen and shorten the proofs of several of their results.
1704.05091
Pedro Saleiro
Pedro Saleiro, Eduarda Mendes Rodrigues, Carlos Soares, Eug\'enio Oliveira
FEUP at SemEval-2017 Task 5: Predicting Sentiment Polarity and Intensity with Financial Word Embeddings
null
null
null
null
cs.CL cs.IR
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This paper presents the approach developed at the Faculty of Engineering of University of Porto, to participate in SemEval 2017, Task 5: Fine-grained Sentiment Analysis on Financial Microblogs and News. The task consisted in predicting a real continuous variable from -1.0 to +1.0 representing the polarity and intensity of sentiment concerning companies/stocks mentioned in short texts. We modeled the task as a regression analysis problem and combined traditional techniques such as pre-processing short texts, bag-of-words representations and lexical-based features with enhanced financial specific bag-of-embeddings. We used an external collection of tweets and news headlines mentioning companies/stocks from S\&P 500 to create financial word embeddings which are able to capture domain-specific syntactic and semantic similarities. The resulting approach obtained a cosine similarity score of 0.69 in sub-task 5.1 - Microblogs and 0.68 in sub-task 5.2 - News Headlines.
[ { "created": "Mon, 17 Apr 2017 18:48:00 GMT", "version": "v1" } ]
2017-04-19
[ [ "Saleiro", "Pedro", "" ], [ "Rodrigues", "Eduarda Mendes", "" ], [ "Soares", "Carlos", "" ], [ "Oliveira", "Eugénio", "" ] ]
This paper presents the approach developed at the Faculty of Engineering of University of Porto, to participate in SemEval 2017, Task 5: Fine-grained Sentiment Analysis on Financial Microblogs and News. The task consisted in predicting a real continuous variable from -1.0 to +1.0 representing the polarity and intensity of sentiment concerning companies/stocks mentioned in short texts. We modeled the task as a regression analysis problem and combined traditional techniques such as pre-processing short texts, bag-of-words representations and lexical-based features with enhanced financial specific bag-of-embeddings. We used an external collection of tweets and news headlines mentioning companies/stocks from S\&P 500 to create financial word embeddings which are able to capture domain-specific syntactic and semantic similarities. The resulting approach obtained a cosine similarity score of 0.69 in sub-task 5.1 - Microblogs and 0.68 in sub-task 5.2 - News Headlines.
2402.19076
Moritz Blum
Gennaro Nolano, Moritz Blum, Basil Ell, Philipp Cimiano
Pointing out the Shortcomings of Relation Extraction Models with Semantically Motivated Adversarials
null
null
null
null
cs.CL
http://creativecommons.org/licenses/by/4.0/
In recent years, large language models have achieved state-of-the-art performance across various NLP tasks. However, investigations have shown that these models tend to rely on shortcut features, leading to inaccurate predictions and causing the models to be unreliable at generalization to out-of-distribution (OOD) samples. For instance, in the context of relation extraction (RE), we would expect a model to identify the same relation independently of the entities involved in it. For example, consider the sentence "Leonardo da Vinci painted the Mona Lisa" expressing the created(Leonardo_da_Vinci, Mona_Lisa) relation. If we substiute "Leonardo da Vinci" with "Barack Obama", then the sentence still expresses the created relation. A robust model is supposed to detect the same relation in both cases. In this work, we describe several semantically-motivated strategies to generate adversarial examples by replacing entity mentions and investigate how state-of-the-art RE models perform under pressure. Our analyses show that the performance of these models significantly deteriorates on the modified datasets (avg. of -48.5% in F1), which indicates that these models rely to a great extent on shortcuts, such as surface forms (or patterns therein) of entities, without making full use of the information present in the sentences.
[ { "created": "Thu, 29 Feb 2024 12:01:46 GMT", "version": "v1" } ]
2024-03-01
[ [ "Nolano", "Gennaro", "" ], [ "Blum", "Moritz", "" ], [ "Ell", "Basil", "" ], [ "Cimiano", "Philipp", "" ] ]
In recent years, large language models have achieved state-of-the-art performance across various NLP tasks. However, investigations have shown that these models tend to rely on shortcut features, leading to inaccurate predictions and causing the models to be unreliable at generalization to out-of-distribution (OOD) samples. For instance, in the context of relation extraction (RE), we would expect a model to identify the same relation independently of the entities involved in it. For example, consider the sentence "Leonardo da Vinci painted the Mona Lisa" expressing the created(Leonardo_da_Vinci, Mona_Lisa) relation. If we substiute "Leonardo da Vinci" with "Barack Obama", then the sentence still expresses the created relation. A robust model is supposed to detect the same relation in both cases. In this work, we describe several semantically-motivated strategies to generate adversarial examples by replacing entity mentions and investigate how state-of-the-art RE models perform under pressure. Our analyses show that the performance of these models significantly deteriorates on the modified datasets (avg. of -48.5% in F1), which indicates that these models rely to a great extent on shortcuts, such as surface forms (or patterns therein) of entities, without making full use of the information present in the sentences.
2307.08074
Longyue Wang
Longyue Wang, Zefeng Du, Donghuai Liu, Deng Cai, Dian Yu, Haiyun Jiang, Yan Wang, Leyang Cui, Shuming Shi, Zhaopeng Tu
Disco-Bench: A Discourse-Aware Evaluation Benchmark for Language Modelling
Zhaopeng Tu is the corresponding author
null
null
null
cs.CL cs.AI
http://creativecommons.org/licenses/by/4.0/
Modeling discourse -- the linguistic phenomena that go beyond individual sentences, is a fundamental yet challenging aspect of natural language processing (NLP). However, existing evaluation benchmarks primarily focus on the evaluation of inter-sentence properties and overlook critical discourse phenomena that cross sentences. To bridge the gap, we propose Disco-Bench, a benchmark that can evaluate intra-sentence discourse properties across a diverse set of NLP tasks, covering understanding, translation, and generation. Disco-Bench consists of 9 document-level testsets in the literature domain, which contain rich discourse phenomena (e.g. cohesion and coherence) in Chinese and/or English. For linguistic analysis, we also design a diagnostic test suite that can examine whether the target models learn discourse knowledge. We totally evaluate 20 general-, in-domain and commercial models based on Transformer, advanced pretraining architectures and large language models (LLMs). Our results show (1) the challenge and necessity of our evaluation benchmark; (2) fine-grained pretraining based on literary document-level training data consistently improves the modeling of discourse information. We will release the datasets, pretrained models, and leaderboard, which we hope can significantly facilitate research in this field: https://github.com/longyuewangdcu/Disco-Bench.
[ { "created": "Sun, 16 Jul 2023 15:18:25 GMT", "version": "v1" }, { "created": "Sat, 22 Jul 2023 00:11:24 GMT", "version": "v2" } ]
2023-07-25
[ [ "Wang", "Longyue", "" ], [ "Du", "Zefeng", "" ], [ "Liu", "Donghuai", "" ], [ "Cai", "Deng", "" ], [ "Yu", "Dian", "" ], [ "Jiang", "Haiyun", "" ], [ "Wang", "Yan", "" ], [ "Cui", "Leyang", "" ], [ "Shi", "Shuming", "" ], [ "Tu", "Zhaopeng", "" ] ]
Modeling discourse -- the linguistic phenomena that go beyond individual sentences, is a fundamental yet challenging aspect of natural language processing (NLP). However, existing evaluation benchmarks primarily focus on the evaluation of inter-sentence properties and overlook critical discourse phenomena that cross sentences. To bridge the gap, we propose Disco-Bench, a benchmark that can evaluate intra-sentence discourse properties across a diverse set of NLP tasks, covering understanding, translation, and generation. Disco-Bench consists of 9 document-level testsets in the literature domain, which contain rich discourse phenomena (e.g. cohesion and coherence) in Chinese and/or English. For linguistic analysis, we also design a diagnostic test suite that can examine whether the target models learn discourse knowledge. We totally evaluate 20 general-, in-domain and commercial models based on Transformer, advanced pretraining architectures and large language models (LLMs). Our results show (1) the challenge and necessity of our evaluation benchmark; (2) fine-grained pretraining based on literary document-level training data consistently improves the modeling of discourse information. We will release the datasets, pretrained models, and leaderboard, which we hope can significantly facilitate research in this field: https://github.com/longyuewangdcu/Disco-Bench.
2212.08241
Inayat Ali
Sonia Sabir, Inayat Ali, Eraj Khan
H-LPS: a hybrid approach for user's location privacy in location-based services
null
null
null
null
cs.CR
http://creativecommons.org/publicdomain/zero/1.0/
Applications providing location-based services (LBS) have gained much attention and importance with the notion of the internet of things (IoT). Users are utilizing LBS by providing their location information to third-party service providers. However, location data is very sensitive that can reveal user's private life to adversaries. The passive and pervasive data collection in IoT upsurges serious issues of location privacy. Privacy-preserving location-based services are a hot research topic. Many anonymization and obfuscation techniques have been proposed to overcome location privacy issues. In this paper, we have proposed a hybrid location privacy scheme (H-LPS), a hybrid scheme mainly based on obfuscation and collaboration for protecting users' location privacy while using location-based services. Obfuscation naturally degrades the quality of service but provides more privacy as compared to anonymization. Our proposed scheme, H-LPS, provides a very high-level of privacy yet provides good accuracy for most of the users. The privacy level and service accuracy of H-LPS are compared with state-of-the-art location privacy schemes and it is shown that H-LPS could be a candidate solution for preserving user location privacy in location-based services.
[ { "created": "Fri, 16 Dec 2022 02:16:29 GMT", "version": "v1" } ]
2022-12-19
[ [ "Sabir", "Sonia", "" ], [ "Ali", "Inayat", "" ], [ "Khan", "Eraj", "" ] ]
Applications providing location-based services (LBS) have gained much attention and importance with the notion of the internet of things (IoT). Users are utilizing LBS by providing their location information to third-party service providers. However, location data is very sensitive that can reveal user's private life to adversaries. The passive and pervasive data collection in IoT upsurges serious issues of location privacy. Privacy-preserving location-based services are a hot research topic. Many anonymization and obfuscation techniques have been proposed to overcome location privacy issues. In this paper, we have proposed a hybrid location privacy scheme (H-LPS), a hybrid scheme mainly based on obfuscation and collaboration for protecting users' location privacy while using location-based services. Obfuscation naturally degrades the quality of service but provides more privacy as compared to anonymization. Our proposed scheme, H-LPS, provides a very high-level of privacy yet provides good accuracy for most of the users. The privacy level and service accuracy of H-LPS are compared with state-of-the-art location privacy schemes and it is shown that H-LPS could be a candidate solution for preserving user location privacy in location-based services.
2311.09217
Yinghao Xu
Yinghao Xu, Hao Tan, Fujun Luan, Sai Bi, Peng Wang, Jiahao Li, Zifan Shi, Kalyan Sunkavalli, Gordon Wetzstein, Zexiang Xu, Kai Zhang
DMV3D: Denoising Multi-View Diffusion using 3D Large Reconstruction Model
Project Page: https://justimyhxu.github.io/projects/dmv3d/
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We propose \textbf{DMV3D}, a novel 3D generation approach that uses a transformer-based 3D large reconstruction model to denoise multi-view diffusion. Our reconstruction model incorporates a triplane NeRF representation and can denoise noisy multi-view images via NeRF reconstruction and rendering, achieving single-stage 3D generation in $\sim$30s on single A100 GPU. We train \textbf{DMV3D} on large-scale multi-view image datasets of highly diverse objects using only image reconstruction losses, without accessing 3D assets. We demonstrate state-of-the-art results for the single-image reconstruction problem where probabilistic modeling of unseen object parts is required for generating diverse reconstructions with sharp textures. We also show high-quality text-to-3D generation results outperforming previous 3D diffusion models. Our project website is at: https://justimyhxu.github.io/projects/dmv3d/ .
[ { "created": "Wed, 15 Nov 2023 18:58:41 GMT", "version": "v1" } ]
2023-11-16
[ [ "Xu", "Yinghao", "" ], [ "Tan", "Hao", "" ], [ "Luan", "Fujun", "" ], [ "Bi", "Sai", "" ], [ "Wang", "Peng", "" ], [ "Li", "Jiahao", "" ], [ "Shi", "Zifan", "" ], [ "Sunkavalli", "Kalyan", "" ], [ "Wetzstein", "Gordon", "" ], [ "Xu", "Zexiang", "" ], [ "Zhang", "Kai", "" ] ]
We propose \textbf{DMV3D}, a novel 3D generation approach that uses a transformer-based 3D large reconstruction model to denoise multi-view diffusion. Our reconstruction model incorporates a triplane NeRF representation and can denoise noisy multi-view images via NeRF reconstruction and rendering, achieving single-stage 3D generation in $\sim$30s on single A100 GPU. We train \textbf{DMV3D} on large-scale multi-view image datasets of highly diverse objects using only image reconstruction losses, without accessing 3D assets. We demonstrate state-of-the-art results for the single-image reconstruction problem where probabilistic modeling of unseen object parts is required for generating diverse reconstructions with sharp textures. We also show high-quality text-to-3D generation results outperforming previous 3D diffusion models. Our project website is at: https://justimyhxu.github.io/projects/dmv3d/ .
2402.11217
Wenting Chen
Wenxuan Wang, Yihang Su, Jingyuan Huan, Jie Liu, Wenting Chen, Yudi Zhang, Cheng-Yi Li, Kao-Jung Chang, Xiaohan Xin, Linlin Shen, Michael R. Lyu
Asclepius: A Spectrum Evaluation Benchmark for Medical Multi-Modal Large Language Models
20 pages, 15 figures
null
null
null
cs.CL cs.CV
http://creativecommons.org/licenses/by-nc-sa/4.0/
The significant breakthroughs of Medical Multi-Modal Large Language Models (Med-MLLMs) renovate modern healthcare with robust information synthesis and medical decision support. However, these models are often evaluated on benchmarks that are unsuitable for the Med-MLLMs due to the intricate nature of the real-world diagnostic frameworks, which encompass diverse medical specialties and involve complex clinical decisions. Moreover, these benchmarks are susceptible to data leakage, since Med-MLLMs are trained on large assemblies of publicly available data. Thus, an isolated and clinically representative benchmark is highly desirable for credible Med-MLLMs evaluation. To this end, we introduce Asclepius, a novel Med-MLLM benchmark that rigorously and comprehensively assesses model capability in terms of: distinct medical specialties (cardiovascular, gastroenterology, etc.) and different diagnostic capacities (perception, disease analysis, etc.). Grounded in 3 proposed core principles, Asclepius ensures a comprehensive evaluation by encompassing 15 medical specialties, stratifying into 3 main categories and 8 sub-categories of clinical tasks, and exempting from train-validate contamination. We further provide an in-depth analysis of 6 Med-MLLMs and compare them with 5 human specialists, providing insights into their competencies and limitations in various medical contexts. Our work not only advances the understanding of Med-MLLMs' capabilities but also sets a precedent for future evaluations and the safe deployment of these models in clinical environments. We launch and maintain a leaderboard for community assessment of Med-MLLM capabilities (https://asclepius-med.github.io/).
[ { "created": "Sat, 17 Feb 2024 08:04:23 GMT", "version": "v1" } ]
2024-02-20
[ [ "Wang", "Wenxuan", "" ], [ "Su", "Yihang", "" ], [ "Huan", "Jingyuan", "" ], [ "Liu", "Jie", "" ], [ "Chen", "Wenting", "" ], [ "Zhang", "Yudi", "" ], [ "Li", "Cheng-Yi", "" ], [ "Chang", "Kao-Jung", "" ], [ "Xin", "Xiaohan", "" ], [ "Shen", "Linlin", "" ], [ "Lyu", "Michael R.", "" ] ]
The significant breakthroughs of Medical Multi-Modal Large Language Models (Med-MLLMs) renovate modern healthcare with robust information synthesis and medical decision support. However, these models are often evaluated on benchmarks that are unsuitable for the Med-MLLMs due to the intricate nature of the real-world diagnostic frameworks, which encompass diverse medical specialties and involve complex clinical decisions. Moreover, these benchmarks are susceptible to data leakage, since Med-MLLMs are trained on large assemblies of publicly available data. Thus, an isolated and clinically representative benchmark is highly desirable for credible Med-MLLMs evaluation. To this end, we introduce Asclepius, a novel Med-MLLM benchmark that rigorously and comprehensively assesses model capability in terms of: distinct medical specialties (cardiovascular, gastroenterology, etc.) and different diagnostic capacities (perception, disease analysis, etc.). Grounded in 3 proposed core principles, Asclepius ensures a comprehensive evaluation by encompassing 15 medical specialties, stratifying into 3 main categories and 8 sub-categories of clinical tasks, and exempting from train-validate contamination. We further provide an in-depth analysis of 6 Med-MLLMs and compare them with 5 human specialists, providing insights into their competencies and limitations in various medical contexts. Our work not only advances the understanding of Med-MLLMs' capabilities but also sets a precedent for future evaluations and the safe deployment of these models in clinical environments. We launch and maintain a leaderboard for community assessment of Med-MLLM capabilities (https://asclepius-med.github.io/).
2109.06710
Junlin Zhao
Mehmet Emre Ozfatura, Junlin Zhao, and Deniz G\"und\"uz
Fast Federated Edge Learning with Overlapped Communication and Computation and Channel-Aware Fair Client Scheduling
Accepted in IEEE SPAWC 2021
null
null
null
cs.IT cs.LG math.IT
http://creativecommons.org/licenses/by/4.0/
We consider federated edge learning (FEEL) over wireless fading channels taking into account the downlink and uplink channel latencies, and the random computation delays at the clients. We speed up the training process by overlapping the communication with computation. With fountain coded transmission of the global model update, clients receive the global model asynchronously, and start performing local computations right away. Then, we propose a dynamic client scheduling policy, called MRTP, for uploading local model updates to the parameter server (PS), which, at any time, schedules the client with the minimum remaining upload time. However, MRTP can lead to biased participation of clients in the update process, resulting in performance degradation in non-iid data scenarios. To overcome this, we propose two alternative schemes with fairness considerations, termed as age-aware MRTP (A-MRTP), and opportunistically fair MRTP (OF-MRTP). In A-MRTP, the remaining clients are scheduled according to the ratio between their remaining transmission time and the update age, while in OF-MRTP, the selection mechanism utilizes the long term average channel rate of the clients to further reduce the latency while ensuring fair participation of the clients. It is shown through numerical simulations that OF-MRTP provides significant reduction in latency without sacrificing test accuracy.
[ { "created": "Tue, 14 Sep 2021 14:16:01 GMT", "version": "v1" } ]
2021-09-15
[ [ "Ozfatura", "Mehmet Emre", "" ], [ "Zhao", "Junlin", "" ], [ "Gündüz", "Deniz", "" ] ]
We consider federated edge learning (FEEL) over wireless fading channels taking into account the downlink and uplink channel latencies, and the random computation delays at the clients. We speed up the training process by overlapping the communication with computation. With fountain coded transmission of the global model update, clients receive the global model asynchronously, and start performing local computations right away. Then, we propose a dynamic client scheduling policy, called MRTP, for uploading local model updates to the parameter server (PS), which, at any time, schedules the client with the minimum remaining upload time. However, MRTP can lead to biased participation of clients in the update process, resulting in performance degradation in non-iid data scenarios. To overcome this, we propose two alternative schemes with fairness considerations, termed as age-aware MRTP (A-MRTP), and opportunistically fair MRTP (OF-MRTP). In A-MRTP, the remaining clients are scheduled according to the ratio between their remaining transmission time and the update age, while in OF-MRTP, the selection mechanism utilizes the long term average channel rate of the clients to further reduce the latency while ensuring fair participation of the clients. It is shown through numerical simulations that OF-MRTP provides significant reduction in latency without sacrificing test accuracy.
2407.00265
Shayan Khorassany
Shayan Khorassany, Eric B. Dew, Mohammad Rahim Sobhani, Roger J. Zemp
Radiation Impedance of Rectangular CMUTs
18 pages, 10 figures, submitted to Sensors
null
null
null
cs.CE
http://creativecommons.org/licenses/by/4.0/
Recently, capacitive micromachined ultrasound transducers (CMUTs) with long rectangular membranes have demonstrated performance advantages over conventional piezoelectric transducers; however, modeling these CMUT geometries has been limited to computationally burdensome numerical methods. Improved fast modeling methods such as equivalent circuit models could help achieve designs with even better performance. The primary obstacle in developing such methods is the lack of tractable methods for computing the radiation impedance of clamped rectangular radiators. This paper presents a method which approximates the velocity profile using a polynomial shape model to rapidly and accurately estimate radiation impedance. The validity of the approximate velocity profile and corresponding radiation impedance calculation was assessed using finite element simulations for a variety of membrane aspect ratios and bias voltages. Our method was evaluated for rectangular radiators with width:length ratios from 1:1 up to 1:25. At all aspect ratios, the radiation resistance was closely modeled. However, when calculating the radiation reactance, our initial approach was only accurate for low aspect ratios. This motivated us to consider an alternative shape model for high aspect ratios, which was more accurate when compared with FEM. To facilitate development of future rectangular CMUTs, we provide a MATLAB script which quickly calculates radiation impedance using both methods.
[ { "created": "Fri, 28 Jun 2024 23:55:28 GMT", "version": "v1" } ]
2024-07-02
[ [ "Khorassany", "Shayan", "" ], [ "Dew", "Eric B.", "" ], [ "Sobhani", "Mohammad Rahim", "" ], [ "Zemp", "Roger J.", "" ] ]
Recently, capacitive micromachined ultrasound transducers (CMUTs) with long rectangular membranes have demonstrated performance advantages over conventional piezoelectric transducers; however, modeling these CMUT geometries has been limited to computationally burdensome numerical methods. Improved fast modeling methods such as equivalent circuit models could help achieve designs with even better performance. The primary obstacle in developing such methods is the lack of tractable methods for computing the radiation impedance of clamped rectangular radiators. This paper presents a method which approximates the velocity profile using a polynomial shape model to rapidly and accurately estimate radiation impedance. The validity of the approximate velocity profile and corresponding radiation impedance calculation was assessed using finite element simulations for a variety of membrane aspect ratios and bias voltages. Our method was evaluated for rectangular radiators with width:length ratios from 1:1 up to 1:25. At all aspect ratios, the radiation resistance was closely modeled. However, when calculating the radiation reactance, our initial approach was only accurate for low aspect ratios. This motivated us to consider an alternative shape model for high aspect ratios, which was more accurate when compared with FEM. To facilitate development of future rectangular CMUTs, we provide a MATLAB script which quickly calculates radiation impedance using both methods.
2309.01240
Visweswaran Baskaran
Akshaya C S, Karthik Soma, Visweswaran B, Aditya Ravichander and Venkata Nagarjun PM
Decentralized shape formation and force-based interactive formation control in robot swarms
6 pages, 10 figures
null
null
null
cs.MA cs.RO
http://creativecommons.org/licenses/by/4.0/
Swarm robotic systems utilize collective behaviour to achieve goals that might be too complex for a lone entity, but become attainable with localized communication and collective decision making. In this paper, a behaviour-based distributed approach to shape formation is proposed. Flocking into strategic formations is observed in migratory birds and fish to avoid predators and also for energy conservation. The formation is maintained throughout long periods without collapsing and is advantageous for communicating within the flock. Similar behaviour can be deployed in multi-agent systems to enhance coordination within the swarm. Existing methods for formation control are either dependent on the size and geometry of the formation or rely on maintaining the formation with a single reference in the swarm (the leader). These methods are not resilient to failure and involve a high degree of deformation upon obstacle encounter before the shape is recovered again. To improve the performance, artificial force-based interaction amongst the entities of the swarm to maintain shape integrity while encountering obstacles is elucidated.
[ { "created": "Sun, 3 Sep 2023 18:46:39 GMT", "version": "v1" } ]
2023-09-06
[ [ "S", "Akshaya C", "" ], [ "Soma", "Karthik", "" ], [ "B", "Visweswaran", "" ], [ "Ravichander", "Aditya", "" ], [ "PM", "Venkata Nagarjun", "" ] ]
Swarm robotic systems utilize collective behaviour to achieve goals that might be too complex for a lone entity, but become attainable with localized communication and collective decision making. In this paper, a behaviour-based distributed approach to shape formation is proposed. Flocking into strategic formations is observed in migratory birds and fish to avoid predators and also for energy conservation. The formation is maintained throughout long periods without collapsing and is advantageous for communicating within the flock. Similar behaviour can be deployed in multi-agent systems to enhance coordination within the swarm. Existing methods for formation control are either dependent on the size and geometry of the formation or rely on maintaining the formation with a single reference in the swarm (the leader). These methods are not resilient to failure and involve a high degree of deformation upon obstacle encounter before the shape is recovered again. To improve the performance, artificial force-based interaction amongst the entities of the swarm to maintain shape integrity while encountering obstacles is elucidated.
2308.11204
Zihang Liu
Zihang Liu, Le Yu, Tongyu Zhu, Leiei Sun
A Simple Framework for Multi-mode Spatial-Temporal Data Modeling
null
null
null
null
cs.LG
http://creativecommons.org/licenses/by/4.0/
Spatial-temporal data modeling aims to mine the underlying spatial relationships and temporal dependencies of objects in a system. However, most existing methods focus on the modeling of spatial-temporal data in a single mode, lacking the understanding of multiple modes. Though very few methods have been presented to learn the multi-mode relationships recently, they are built on complicated components with higher model complexities. In this paper, we propose a simple framework for multi-mode spatial-temporal data modeling to bring both effectiveness and efficiency together. Specifically, we design a general cross-mode spatial relationships learning component to adaptively establish connections between multiple modes and propagate information along the learned connections. Moreover, we employ multi-layer perceptrons to capture the temporal dependencies and channel correlations, which are conceptually and technically succinct. Experiments on three real-world datasets show that our model can consistently outperform the baselines with lower space and time complexity, opening up a promising direction for modeling spatial-temporal data. The generalizability of the cross-mode spatial relationships learning module is also validated.
[ { "created": "Tue, 22 Aug 2023 05:41:20 GMT", "version": "v1" } ]
2023-08-23
[ [ "Liu", "Zihang", "" ], [ "Yu", "Le", "" ], [ "Zhu", "Tongyu", "" ], [ "Sun", "Leiei", "" ] ]
Spatial-temporal data modeling aims to mine the underlying spatial relationships and temporal dependencies of objects in a system. However, most existing methods focus on the modeling of spatial-temporal data in a single mode, lacking the understanding of multiple modes. Though very few methods have been presented to learn the multi-mode relationships recently, they are built on complicated components with higher model complexities. In this paper, we propose a simple framework for multi-mode spatial-temporal data modeling to bring both effectiveness and efficiency together. Specifically, we design a general cross-mode spatial relationships learning component to adaptively establish connections between multiple modes and propagate information along the learned connections. Moreover, we employ multi-layer perceptrons to capture the temporal dependencies and channel correlations, which are conceptually and technically succinct. Experiments on three real-world datasets show that our model can consistently outperform the baselines with lower space and time complexity, opening up a promising direction for modeling spatial-temporal data. The generalizability of the cross-mode spatial relationships learning module is also validated.
2203.15706
Alec Linot
Alec J. Linot, Joshua W. Burby, Qi Tang, Prasanna Balaprakash, Michael D. Graham, Romit Maulik
Stabilized Neural Ordinary Differential Equations for Long-Time Forecasting of Dynamical Systems
null
null
10.1016/j.jcp.2022.111838
null
cs.LG
http://creativecommons.org/licenses/by/4.0/
In data-driven modeling of spatiotemporal phenomena careful consideration often needs to be made in capturing the dynamics of the high wavenumbers. This problem becomes especially challenging when the system of interest exhibits shocks or chaotic dynamics. We present a data-driven modeling method that accurately captures shocks and chaotic dynamics by proposing a novel architecture, stabilized neural ordinary differential equation (ODE). In our proposed architecture, we learn the right-hand-side (RHS) of an ODE by adding the outputs of two NN together where one learns a linear term and the other a nonlinear term. Specifically, we implement this by training a sparse linear convolutional NN to learn the linear term and a dense fully-connected nonlinear NN to learn the nonlinear term. This is in contrast with the standard neural ODE which involves training only a single NN for learning the RHS. We apply this setup to the viscous Burgers equation, which exhibits shocked behavior, and show better short-time tracking and prediction of the energy spectrum at high wavenumbers than a standard neural ODE. We also find that the stabilized neural ODE models are much more robust to noisy initial conditions than the standard neural ODE approach. We also apply this method to chaotic trajectories of the Kuramoto-Sivashinsky equation. In this case, stabilized neural ODEs keep long-time trajectories on the attractor, and are highly robust to noisy initial conditions, while standard neural ODEs fail at achieving either of these results. We conclude by demonstrating how stabilizing neural ODEs provide a natural extension for use in reduced-order modeling by projecting the dynamics onto the eigenvectors of the learned linear term.
[ { "created": "Tue, 29 Mar 2022 16:10:34 GMT", "version": "v1" }, { "created": "Tue, 4 Oct 2022 00:03:30 GMT", "version": "v2" } ]
2022-12-28
[ [ "Linot", "Alec J.", "" ], [ "Burby", "Joshua W.", "" ], [ "Tang", "Qi", "" ], [ "Balaprakash", "Prasanna", "" ], [ "Graham", "Michael D.", "" ], [ "Maulik", "Romit", "" ] ]
In data-driven modeling of spatiotemporal phenomena careful consideration often needs to be made in capturing the dynamics of the high wavenumbers. This problem becomes especially challenging when the system of interest exhibits shocks or chaotic dynamics. We present a data-driven modeling method that accurately captures shocks and chaotic dynamics by proposing a novel architecture, stabilized neural ordinary differential equation (ODE). In our proposed architecture, we learn the right-hand-side (RHS) of an ODE by adding the outputs of two NN together where one learns a linear term and the other a nonlinear term. Specifically, we implement this by training a sparse linear convolutional NN to learn the linear term and a dense fully-connected nonlinear NN to learn the nonlinear term. This is in contrast with the standard neural ODE which involves training only a single NN for learning the RHS. We apply this setup to the viscous Burgers equation, which exhibits shocked behavior, and show better short-time tracking and prediction of the energy spectrum at high wavenumbers than a standard neural ODE. We also find that the stabilized neural ODE models are much more robust to noisy initial conditions than the standard neural ODE approach. We also apply this method to chaotic trajectories of the Kuramoto-Sivashinsky equation. In this case, stabilized neural ODEs keep long-time trajectories on the attractor, and are highly robust to noisy initial conditions, while standard neural ODEs fail at achieving either of these results. We conclude by demonstrating how stabilizing neural ODEs provide a natural extension for use in reduced-order modeling by projecting the dynamics onto the eigenvectors of the learned linear term.
2306.02190
Sofia Serrano
Sofia Serrano, Jesse Dodge, Noah A. Smith
Stubborn Lexical Bias in Data and Models
ACL Findings 2023
null
null
null
cs.CL
http://creativecommons.org/licenses/by/4.0/
In NLP, recent work has seen increased focus on spurious correlations between various features and labels in training data, and how these influence model behavior. However, the presence and effect of such correlations are typically examined feature by feature. We investigate the cumulative impact on a model of many such intersecting features. Using a new statistical method, we examine whether such spurious patterns in data appear in models trained on the data. We select two tasks -- natural language inference and duplicate-question detection -- for which any unigram feature on its own should ideally be uninformative, which gives us a large pool of automatically extracted features with which to experiment. The large size of this pool allows us to investigate the intersection of features spuriously associated with (potentially different) labels. We then apply an optimization approach to *reweight* the training data, reducing thousands of spurious correlations, and examine how doing so affects models trained on the reweighted data. Surprisingly, though this method can successfully reduce lexical biases in the training data, we still find strong evidence of corresponding bias in the trained models, including worsened bias for slightly more complex features (bigrams). We close with discussion about the implications of our results on what it means to "debias" training data, and how issues of data quality can affect model bias.
[ { "created": "Sat, 3 Jun 2023 20:12:27 GMT", "version": "v1" } ]
2023-06-06
[ [ "Serrano", "Sofia", "" ], [ "Dodge", "Jesse", "" ], [ "Smith", "Noah A.", "" ] ]
In NLP, recent work has seen increased focus on spurious correlations between various features and labels in training data, and how these influence model behavior. However, the presence and effect of such correlations are typically examined feature by feature. We investigate the cumulative impact on a model of many such intersecting features. Using a new statistical method, we examine whether such spurious patterns in data appear in models trained on the data. We select two tasks -- natural language inference and duplicate-question detection -- for which any unigram feature on its own should ideally be uninformative, which gives us a large pool of automatically extracted features with which to experiment. The large size of this pool allows us to investigate the intersection of features spuriously associated with (potentially different) labels. We then apply an optimization approach to *reweight* the training data, reducing thousands of spurious correlations, and examine how doing so affects models trained on the reweighted data. Surprisingly, though this method can successfully reduce lexical biases in the training data, we still find strong evidence of corresponding bias in the trained models, including worsened bias for slightly more complex features (bigrams). We close with discussion about the implications of our results on what it means to "debias" training data, and how issues of data quality can affect model bias.
1501.07422
Kohta Ishikawa
Kohta Ishikawa, Ikuro Sato, Mitsuru Ambai
Pairwise Rotation Hashing for High-dimensional Features
16 pages, 8 figures, wrote at Mar 2014
null
null
null
cs.CV stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Binary Hashing is widely used for effective approximate nearest neighbors search. Even though various binary hashing methods have been proposed, very few methods are feasible for extremely high-dimensional features often used in visual tasks today. We propose a novel highly sparse linear hashing method based on pairwise rotations. The encoding cost of the proposed algorithm is $\mathrm{O}(n \log n)$ for n-dimensional features, whereas that of the existing state-of-the-art method is typically $\mathrm{O}(n^2)$. The proposed method is also remarkably faster in the learning phase. Along with the efficiency, the retrieval accuracy is comparable to or slightly outperforming the state-of-the-art. Pairwise rotations used in our method are formulated from an analytical study of the trade-off relationship between quantization error and entropy of binary codes. Although these hashing criteria are widely used in previous researches, its analytical behavior is rarely studied. All building blocks of our algorithm are based on the analytical solution, and it thus provides a fairly simple and efficient procedure.
[ { "created": "Thu, 29 Jan 2015 11:50:33 GMT", "version": "v1" } ]
2015-01-30
[ [ "Ishikawa", "Kohta", "" ], [ "Sato", "Ikuro", "" ], [ "Ambai", "Mitsuru", "" ] ]
Binary Hashing is widely used for effective approximate nearest neighbors search. Even though various binary hashing methods have been proposed, very few methods are feasible for extremely high-dimensional features often used in visual tasks today. We propose a novel highly sparse linear hashing method based on pairwise rotations. The encoding cost of the proposed algorithm is $\mathrm{O}(n \log n)$ for n-dimensional features, whereas that of the existing state-of-the-art method is typically $\mathrm{O}(n^2)$. The proposed method is also remarkably faster in the learning phase. Along with the efficiency, the retrieval accuracy is comparable to or slightly outperforming the state-of-the-art. Pairwise rotations used in our method are formulated from an analytical study of the trade-off relationship between quantization error and entropy of binary codes. Although these hashing criteria are widely used in previous researches, its analytical behavior is rarely studied. All building blocks of our algorithm are based on the analytical solution, and it thus provides a fairly simple and efficient procedure.
cs/0612081
Manas Tungare
Manas Tungare, Pardha S. Pyla, Manuel P\'erez-Qui\~nones, and Steve Harrison
Personal Information Ecosystems and Implications for Design
null
null
null
null
cs.HC
null
Today, people use multiple devices to fulfill their information needs. However, designers design each device individually, without accounting for the other devices that users may also use. In many cases, the applications on all these devices are designed to be functional replicates of each other. We argue that this results in an over-reliance on data synchronization across devices, version control nightmares, and increased burden of file management. In this paper, we present the idea of a \textit{personal information ecosystem}, an analogy to biological ecosystems, which allows us to discuss the inter-relationships among these devices to fulfill the information needs of the user. There is a need for designers to design devices as part of a complete ecosystem, not as independent devices that simply share data replicated across them. To help us understand this domain and to facilitate the dialogue and study of such systems, we present the terminology, classifications of the interdependencies among different devices, and resulting implications for design.
[ { "created": "Mon, 18 Dec 2006 07:53:34 GMT", "version": "v1" } ]
2007-05-23
[ [ "Tungare", "Manas", "" ], [ "Pyla", "Pardha S.", "" ], [ "Pérez-Quiñones", "Manuel", "" ], [ "Harrison", "Steve", "" ] ]
Today, people use multiple devices to fulfill their information needs. However, designers design each device individually, without accounting for the other devices that users may also use. In many cases, the applications on all these devices are designed to be functional replicates of each other. We argue that this results in an over-reliance on data synchronization across devices, version control nightmares, and increased burden of file management. In this paper, we present the idea of a \textit{personal information ecosystem}, an analogy to biological ecosystems, which allows us to discuss the inter-relationships among these devices to fulfill the information needs of the user. There is a need for designers to design devices as part of a complete ecosystem, not as independent devices that simply share data replicated across them. To help us understand this domain and to facilitate the dialogue and study of such systems, we present the terminology, classifications of the interdependencies among different devices, and resulting implications for design.
2305.09527
Dominik Muhle
Dominik Muhle, Lukas Koestler, Krishna Murthy Jatavallabhula, Daniel Cremers
Learning Correspondence Uncertainty via Differentiable Nonlinear Least Squares
null
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We propose a differentiable nonlinear least squares framework to account for uncertainty in relative pose estimation from feature correspondences. Specifically, we introduce a symmetric version of the probabilistic normal epipolar constraint, and an approach to estimate the covariance of feature positions by differentiating through the camera pose estimation procedure. We evaluate our approach on synthetic, as well as the KITTI and EuRoC real-world datasets. On the synthetic dataset, we confirm that our learned covariances accurately approximate the true noise distribution. In real world experiments, we find that our approach consistently outperforms state-of-the-art non-probabilistic and probabilistic approaches, regardless of the feature extraction algorithm of choice.
[ { "created": "Tue, 16 May 2023 15:21:09 GMT", "version": "v1" }, { "created": "Thu, 18 May 2023 18:35:23 GMT", "version": "v2" } ]
2023-05-22
[ [ "Muhle", "Dominik", "" ], [ "Koestler", "Lukas", "" ], [ "Jatavallabhula", "Krishna Murthy", "" ], [ "Cremers", "Daniel", "" ] ]
We propose a differentiable nonlinear least squares framework to account for uncertainty in relative pose estimation from feature correspondences. Specifically, we introduce a symmetric version of the probabilistic normal epipolar constraint, and an approach to estimate the covariance of feature positions by differentiating through the camera pose estimation procedure. We evaluate our approach on synthetic, as well as the KITTI and EuRoC real-world datasets. On the synthetic dataset, we confirm that our learned covariances accurately approximate the true noise distribution. In real world experiments, we find that our approach consistently outperforms state-of-the-art non-probabilistic and probabilistic approaches, regardless of the feature extraction algorithm of choice.
2206.00169
Giannis Daras
Giannis Daras and Alexandros G. Dimakis
Discovering the Hidden Vocabulary of DALLE-2
6 pages, 4 figures
null
null
null
cs.LG cs.CL cs.CR cs.CV
http://creativecommons.org/licenses/by/4.0/
We discover that DALLE-2 seems to have a hidden vocabulary that can be used to generate images with absurd prompts. For example, it seems that \texttt{Apoploe vesrreaitais} means birds and \texttt{Contarra ccetnxniams luryca tanniounons} (sometimes) means bugs or pests. We find that these prompts are often consistent in isolation but also sometimes in combinations. We present our black-box method to discover words that seem random but have some correspondence to visual concepts. This creates important security and interpretability challenges.
[ { "created": "Wed, 1 Jun 2022 01:14:48 GMT", "version": "v1" } ]
2022-06-02
[ [ "Daras", "Giannis", "" ], [ "Dimakis", "Alexandros G.", "" ] ]
We discover that DALLE-2 seems to have a hidden vocabulary that can be used to generate images with absurd prompts. For example, it seems that \texttt{Apoploe vesrreaitais} means birds and \texttt{Contarra ccetnxniams luryca tanniounons} (sometimes) means bugs or pests. We find that these prompts are often consistent in isolation but also sometimes in combinations. We present our black-box method to discover words that seem random but have some correspondence to visual concepts. This creates important security and interpretability challenges.
2005.05455
Ron Roth
Ron M. Roth, Paul H. Siegel
Variable-Length Constrained Coding and Kraft Conditions: The Parity-Preserving Case
Title has been changed, along with minor modification in text
null
null
null
cs.IT math.CO math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Previous work by the authors on parity-preserving fixed-length constrained encoders is extended to the variable-length case. Parity-preserving variable-length encoders are formally defined, and, to this end, Kraft conditions are developed for the parity-preserving variable-length setting. Then, a necessary and sufficient condition is presented for the existence of deterministic parity-preserving variable-length encoders for a given constraint. Examples are provided that show that there are coding ratios where parity-preserving variable-length encoders exist, while fixed-length encoders do not.
[ { "created": "Mon, 11 May 2020 21:53:22 GMT", "version": "v1" }, { "created": "Thu, 1 Jul 2021 20:53:48 GMT", "version": "v2" } ]
2021-07-05
[ [ "Roth", "Ron M.", "" ], [ "Siegel", "Paul H.", "" ] ]
Previous work by the authors on parity-preserving fixed-length constrained encoders is extended to the variable-length case. Parity-preserving variable-length encoders are formally defined, and, to this end, Kraft conditions are developed for the parity-preserving variable-length setting. Then, a necessary and sufficient condition is presented for the existence of deterministic parity-preserving variable-length encoders for a given constraint. Examples are provided that show that there are coding ratios where parity-preserving variable-length encoders exist, while fixed-length encoders do not.
2305.18859
Jan Mrkos
David Fiedler and Jan Mrkos
Large-scale Ridesharing DARP Instances Based on Real Travel Demand
8 pages, 9 figures. Submitted to 26th IEEE International Conference on Intelligent Transportation Systems ITSC 2023. For the published associated dataset and source codes, see the repository https://github.com/aicenter/Ridesharing_DARP_instances
null
null
null
cs.AI math.OC
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Accurately predicting the real-life performance of algorithms solving the Dial-a-Ride Problem (DARP) in the context of Mobility on Demand (MoD) systems with ridesharing requires evaluating them on representative instances. However, the benchmarking of state-of-the-art DARP solution methods has been limited to small, artificial instances or outdated non-public instances, hindering direct comparisons. With the rise of large MoD systems and the availability of open travel demand datasets for many US cities, there is now an opportunity to evaluate these algorithms on standardized, realistic, and representative instances. Despite the significant challenges involved in processing obfuscated and diverse datasets, we have developed a methodology using which we have created a comprehensive set of large-scale demand instances based on real-world data. These instances cover diverse use cases, one of which is demonstrated in an evaluation of two established DARP methods: the insertion heuristic and optimal vehicle-group assignment method. We publish the full results of both methods in a standardized format. The results show significant differences between areas in all measured quantities, emphasizing the importance of evaluating methods across different cities.
[ { "created": "Tue, 30 May 2023 08:51:11 GMT", "version": "v1" } ]
2023-05-31
[ [ "Fiedler", "David", "" ], [ "Mrkos", "Jan", "" ] ]
Accurately predicting the real-life performance of algorithms solving the Dial-a-Ride Problem (DARP) in the context of Mobility on Demand (MoD) systems with ridesharing requires evaluating them on representative instances. However, the benchmarking of state-of-the-art DARP solution methods has been limited to small, artificial instances or outdated non-public instances, hindering direct comparisons. With the rise of large MoD systems and the availability of open travel demand datasets for many US cities, there is now an opportunity to evaluate these algorithms on standardized, realistic, and representative instances. Despite the significant challenges involved in processing obfuscated and diverse datasets, we have developed a methodology using which we have created a comprehensive set of large-scale demand instances based on real-world data. These instances cover diverse use cases, one of which is demonstrated in an evaluation of two established DARP methods: the insertion heuristic and optimal vehicle-group assignment method. We publish the full results of both methods in a standardized format. The results show significant differences between areas in all measured quantities, emphasizing the importance of evaluating methods across different cities.
1809.08198
Huda Nassar
Huda Nassar, Georgios Kollias, Ananth Grama, David F. Gleich
Low rank methods for multiple network alignment
17 pages, 10 figures
null
null
null
cs.SI cs.LG physics.soc-ph
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Multiple network alignment is the problem of identifying similar and related regions in a given set of networks. While there are a large number of effective techniques for pairwise problems with two networks that scale in terms of edges, these cannot be readily extended to align multiple networks as the computational complexity will tend to grow exponentially with the number of networks.In this paper we introduce a new multiple network alignment algorithm and framework that is effective at aligning thousands of networks with thousands of nodes. The key enabling technique of our algorithm is identifying an exact and easy to compute low-rank tensor structure inside of a principled heuristic procedure for pairwise network alignment called IsoRank. This can be combined with a new algorithm for $k$-dimensional matching problems on low-rank tensors to produce the alignment. We demonstrate results on synthetic and real-world problems that show our technique (i) is as good or better in terms of quality as existing methods, when they work on small problems, while running considerably faster and (ii) is able to scale to aligning a number of networks unreachable by current methods. We show in this paper that our method is the realistic choice for aligning multiple networks when no prior information is present.
[ { "created": "Fri, 21 Sep 2018 16:38:36 GMT", "version": "v1" } ]
2018-09-24
[ [ "Nassar", "Huda", "" ], [ "Kollias", "Georgios", "" ], [ "Grama", "Ananth", "" ], [ "Gleich", "David F.", "" ] ]
Multiple network alignment is the problem of identifying similar and related regions in a given set of networks. While there are a large number of effective techniques for pairwise problems with two networks that scale in terms of edges, these cannot be readily extended to align multiple networks as the computational complexity will tend to grow exponentially with the number of networks.In this paper we introduce a new multiple network alignment algorithm and framework that is effective at aligning thousands of networks with thousands of nodes. The key enabling technique of our algorithm is identifying an exact and easy to compute low-rank tensor structure inside of a principled heuristic procedure for pairwise network alignment called IsoRank. This can be combined with a new algorithm for $k$-dimensional matching problems on low-rank tensors to produce the alignment. We demonstrate results on synthetic and real-world problems that show our technique (i) is as good or better in terms of quality as existing methods, when they work on small problems, while running considerably faster and (ii) is able to scale to aligning a number of networks unreachable by current methods. We show in this paper that our method is the realistic choice for aligning multiple networks when no prior information is present.
1806.00428
Aditya Vora
Aditya Vora
A Classification approach towards Unsupervised Learning of Visual Representations
null
null
null
null
cs.CV cs.LG stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In this paper, we present a technique for unsupervised learning of visual representations. Specifically, we train a model for foreground and background classification task, in the process of which it learns visual representations. Foreground and background patches for training come af- ter mining for such patches from hundreds and thousands of unlabelled videos available on the web which we ex- tract using a proposed patch extraction algorithm. With- out using any supervision, with just using 150, 000 unla- belled videos and the PASCAL VOC 2007 dataset, we train a object recognition model that achieves 45.3 mAP which is close to the best performing unsupervised feature learn- ing technique whereas better than many other proposed al- gorithms. The code for patch extraction is implemented in Matlab and available open source at the following link .
[ { "created": "Fri, 1 Jun 2018 16:35:08 GMT", "version": "v1" } ]
2018-06-04
[ [ "Vora", "Aditya", "" ] ]
In this paper, we present a technique for unsupervised learning of visual representations. Specifically, we train a model for foreground and background classification task, in the process of which it learns visual representations. Foreground and background patches for training come af- ter mining for such patches from hundreds and thousands of unlabelled videos available on the web which we ex- tract using a proposed patch extraction algorithm. With- out using any supervision, with just using 150, 000 unla- belled videos and the PASCAL VOC 2007 dataset, we train a object recognition model that achieves 45.3 mAP which is close to the best performing unsupervised feature learn- ing technique whereas better than many other proposed al- gorithms. The code for patch extraction is implemented in Matlab and available open source at the following link .
2401.09456
Denis Shchepakin
Denis Shchepakin, Sreecharan Sankaranarayanan, Dawn Zimmaro
Parametric Constraints for Bayesian Knowledge Tracing from First Principles
null
null
null
null
cs.CY cs.LG stat.ML
http://creativecommons.org/licenses/by-nc-nd/4.0/
Bayesian Knowledge Tracing (BKT) is a probabilistic model of a learner's state of mastery corresponding to a knowledge component. It considers the learner's state of mastery as a "hidden" or latent binary variable and updates this state based on the observed correctness of the learner's response using parameters that represent transition probabilities between states. BKT is often represented as a Hidden Markov Model and the Expectation-Maximization (EM) algorithm is used to infer these parameters. However, this algorithm can suffer from several issues including producing multiple viable sets of parameters, settling into a local minima, producing degenerate parameter values, and a high computational cost during fitting. This paper takes a "from first principles" approach to deriving constraints that can be imposed on the BKT parameter space. Starting from the basic mathematical truths of probability and building up to the behaviors expected of the BKT parameters in real systems, this paper presents a mathematical derivation that results in succinct constraints that can be imposed on the BKT parameter space. Since these constraints are necessary conditions, they can be applied prior to fitting in order to reduce computational cost and the likelihood of issues that can emerge from the EM procedure. In order to see that promise through, the paper further introduces a novel algorithm for estimating BKT parameters subject to the newly defined constraints. While the issue of degenerate parameter values has been reported previously, this paper is the first, to our best knowledge, to derive the constrains from first principles while also presenting an algorithm that respects those constraints.
[ { "created": "Sat, 23 Dec 2023 03:58:41 GMT", "version": "v1" } ]
2024-01-19
[ [ "Shchepakin", "Denis", "" ], [ "Sankaranarayanan", "Sreecharan", "" ], [ "Zimmaro", "Dawn", "" ] ]
Bayesian Knowledge Tracing (BKT) is a probabilistic model of a learner's state of mastery corresponding to a knowledge component. It considers the learner's state of mastery as a "hidden" or latent binary variable and updates this state based on the observed correctness of the learner's response using parameters that represent transition probabilities between states. BKT is often represented as a Hidden Markov Model and the Expectation-Maximization (EM) algorithm is used to infer these parameters. However, this algorithm can suffer from several issues including producing multiple viable sets of parameters, settling into a local minima, producing degenerate parameter values, and a high computational cost during fitting. This paper takes a "from first principles" approach to deriving constraints that can be imposed on the BKT parameter space. Starting from the basic mathematical truths of probability and building up to the behaviors expected of the BKT parameters in real systems, this paper presents a mathematical derivation that results in succinct constraints that can be imposed on the BKT parameter space. Since these constraints are necessary conditions, they can be applied prior to fitting in order to reduce computational cost and the likelihood of issues that can emerge from the EM procedure. In order to see that promise through, the paper further introduces a novel algorithm for estimating BKT parameters subject to the newly defined constraints. While the issue of degenerate parameter values has been reported previously, this paper is the first, to our best knowledge, to derive the constrains from first principles while also presenting an algorithm that respects those constraints.
1511.09360
Faisal Abu-Khzam
Faisal N. Abu-Khzam
On the Complexity of Multi-Parameterized Cluster Editing
null
null
null
null
cs.DS cs.CC
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The Cluster Editing problem seeks a transformation of a given undirected graph into a disjoint union of cliques via a minimum number of edge additions or deletions. A multi-parameterized version of the problem is studied, featuring a number of input parameters that bound the amount of both edge-additions and deletions per single vertex, as well as the size of a clique-cluster. We show that the problem remains NP-hard even when only one edge can be deleted and at most two edges can be added per vertex. However, the new formulation allows us to solve Cluster Editing (exactly) in polynomial time when the number of edge-edit operations per vertex is smaller than half the minimum cluster size. In other words, Correlation Clustering can be solved efficiently when the number of false positives/negatives per single data element is expected to be small compared to the minimum cluster size. As a byproduct, we obtain a kernelization algorithm that delivers linear-size kernels when the two edge-edit bounds are small constants.
[ { "created": "Mon, 30 Nov 2015 15:56:47 GMT", "version": "v1" } ]
2015-12-01
[ [ "Abu-Khzam", "Faisal N.", "" ] ]
The Cluster Editing problem seeks a transformation of a given undirected graph into a disjoint union of cliques via a minimum number of edge additions or deletions. A multi-parameterized version of the problem is studied, featuring a number of input parameters that bound the amount of both edge-additions and deletions per single vertex, as well as the size of a clique-cluster. We show that the problem remains NP-hard even when only one edge can be deleted and at most two edges can be added per vertex. However, the new formulation allows us to solve Cluster Editing (exactly) in polynomial time when the number of edge-edit operations per vertex is smaller than half the minimum cluster size. In other words, Correlation Clustering can be solved efficiently when the number of false positives/negatives per single data element is expected to be small compared to the minimum cluster size. As a byproduct, we obtain a kernelization algorithm that delivers linear-size kernels when the two edge-edit bounds are small constants.
1512.06922
Yinxiao Li
Yinxiao Li and Yonghao Yue and Danfei Xu and Eitan Grinspun and Peter Allen
Folding Deformable Objects using Predictive Simulation and Trajectory Optimization
8 pages, 9 figures, Proceedings of IROS 2015
null
null
null
cs.RO
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Robotic manipulation of deformable objects remains a challenging task. One such task is folding a garment autonomously. Given start and end folding positions, what is an optimal trajectory to move the robotic arm to fold a garment? Certain trajectories will cause the garment to move, creating wrinkles, and gaps, other trajectories will fail altogether. We present a novel solution to find an optimal trajectory that avoids such problematic scenarios. The trajectory is optimized by minimizing a quadratic objective function in an off-line simulator, which includes material properties of the garment and frictional force on the table. The function measures the dissimilarity between a user folded shape and the folded garment in simulation, which is then used as an error measurement to create an optimal trajectory. We demonstrate that our two-arm robot can follow the optimized trajectories, achieving accurate and efficient manipulations of deformable objects.
[ { "created": "Tue, 22 Dec 2015 00:46:47 GMT", "version": "v1" } ]
2015-12-23
[ [ "Li", "Yinxiao", "" ], [ "Yue", "Yonghao", "" ], [ "Xu", "Danfei", "" ], [ "Grinspun", "Eitan", "" ], [ "Allen", "Peter", "" ] ]
Robotic manipulation of deformable objects remains a challenging task. One such task is folding a garment autonomously. Given start and end folding positions, what is an optimal trajectory to move the robotic arm to fold a garment? Certain trajectories will cause the garment to move, creating wrinkles, and gaps, other trajectories will fail altogether. We present a novel solution to find an optimal trajectory that avoids such problematic scenarios. The trajectory is optimized by minimizing a quadratic objective function in an off-line simulator, which includes material properties of the garment and frictional force on the table. The function measures the dissimilarity between a user folded shape and the folded garment in simulation, which is then used as an error measurement to create an optimal trajectory. We demonstrate that our two-arm robot can follow the optimized trajectories, achieving accurate and efficient manipulations of deformable objects.
2312.09369
Dmitriy Serdyuk
Avner May, Dmitriy Serdyuk, Ankit Parag Shah, Otavio Braga, Olivier Siohan
Audio-visual fine-tuning of audio-only ASR models
null
null
null
null
cs.SD cs.AI eess.AS
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Audio-visual automatic speech recognition (AV-ASR) models are very effective at reducing word error rates on noisy speech, but require large amounts of transcribed AV training data. Recently, audio-visual self-supervised learning (SSL) approaches have been developed to reduce this dependence on transcribed AV data, but these methods are quite complex and computationally expensive. In this work, we propose replacing these expensive AV-SSL methods with a simple and fast \textit{audio-only} SSL method, and then performing AV supervised fine-tuning. We show that this approach is competitive with state-of-the-art (SOTA) AV-SSL methods on the LRS3-TED benchmark task (within 0.5% absolute WER), while being dramatically simpler and more efficient (12-30x faster to pre-train). Furthermore, we show we can extend this approach to convert a SOTA audio-only ASR model into an AV model. By doing so, we match SOTA AV-SSL results, even though no AV data was used during pre-training.
[ { "created": "Thu, 14 Dec 2023 22:05:15 GMT", "version": "v1" } ]
2023-12-18
[ [ "May", "Avner", "" ], [ "Serdyuk", "Dmitriy", "" ], [ "Shah", "Ankit Parag", "" ], [ "Braga", "Otavio", "" ], [ "Siohan", "Olivier", "" ] ]
Audio-visual automatic speech recognition (AV-ASR) models are very effective at reducing word error rates on noisy speech, but require large amounts of transcribed AV training data. Recently, audio-visual self-supervised learning (SSL) approaches have been developed to reduce this dependence on transcribed AV data, but these methods are quite complex and computationally expensive. In this work, we propose replacing these expensive AV-SSL methods with a simple and fast \textit{audio-only} SSL method, and then performing AV supervised fine-tuning. We show that this approach is competitive with state-of-the-art (SOTA) AV-SSL methods on the LRS3-TED benchmark task (within 0.5% absolute WER), while being dramatically simpler and more efficient (12-30x faster to pre-train). Furthermore, we show we can extend this approach to convert a SOTA audio-only ASR model into an AV model. By doing so, we match SOTA AV-SSL results, even though no AV data was used during pre-training.
2406.08754
HengRui Xing
Bangxin Li and Hengrui Xing and Chao Huang and Jin Qian and Huangqing Xiao and Linfeng Feng and Cong Tian
Exploiting Uncommon Text-Encoded Structures for Automated Jailbreaks in LLMs
12 pages, 4 figures
null
null
null
cs.CL cs.CR
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Large Language Models (LLMs) are widely used in natural language processing but face the risk of jailbreak attacks that maliciously induce them to generate harmful content. Existing jailbreak attacks, including character-level and context-level attacks, mainly focus on the prompt of the plain text without specifically exploring the significant influence of its structure. In this paper, we focus on studying how prompt structure contributes to the jailbreak attack. We introduce a novel structure-level attack method based on tail structures that are rarely used during LLM training, which we refer to as Uncommon Text-Encoded Structure (UTES). We extensively study 12 UTESs templates and 6 obfuscation methods to build an effective automated jailbreak tool named StructuralSleight that contains three escalating attack strategies: Structural Attack, Structural and Character/Context Obfuscation Attack, and Fully Obfuscated Structural Attack. Extensive experiments on existing LLMs show that StructuralSleight significantly outperforms baseline methods. In particular, the attack success rate reaches 94.62\% on GPT-4o, which has not been addressed by state-of-the-art techniques.
[ { "created": "Thu, 13 Jun 2024 02:24:08 GMT", "version": "v1" }, { "created": "Fri, 19 Jul 2024 08:23:38 GMT", "version": "v2" } ]
2024-07-22
[ [ "Li", "Bangxin", "" ], [ "Xing", "Hengrui", "" ], [ "Huang", "Chao", "" ], [ "Qian", "Jin", "" ], [ "Xiao", "Huangqing", "" ], [ "Feng", "Linfeng", "" ], [ "Tian", "Cong", "" ] ]
Large Language Models (LLMs) are widely used in natural language processing but face the risk of jailbreak attacks that maliciously induce them to generate harmful content. Existing jailbreak attacks, including character-level and context-level attacks, mainly focus on the prompt of the plain text without specifically exploring the significant influence of its structure. In this paper, we focus on studying how prompt structure contributes to the jailbreak attack. We introduce a novel structure-level attack method based on tail structures that are rarely used during LLM training, which we refer to as Uncommon Text-Encoded Structure (UTES). We extensively study 12 UTESs templates and 6 obfuscation methods to build an effective automated jailbreak tool named StructuralSleight that contains three escalating attack strategies: Structural Attack, Structural and Character/Context Obfuscation Attack, and Fully Obfuscated Structural Attack. Extensive experiments on existing LLMs show that StructuralSleight significantly outperforms baseline methods. In particular, the attack success rate reaches 94.62\% on GPT-4o, which has not been addressed by state-of-the-art techniques.
2311.10732
Daniel Leiker
Daniel Leiker
White Paper: The Generative Education (GenEd) Framework
null
null
null
null
cs.CY
http://creativecommons.org/licenses/by-nc-sa/4.0/
The Generative Education (GenEd) Framework explores the transition from Large Language Models (LLMs) to Large Multimodal Models (LMMs) in education, envisioning a harmonious relationship between AI and educators to enhance learning experiences. This paper delves into the potential of LMMs to create personalized, interactive, and emotionally-aware learning environments. Through addressing the Two-Sigma problem and the introduction of a conceptual product named Harmony, the narrative emphasizes educator development, adapting policy frameworks, and fostering cross-sector collaboration to realize the envisioned AI-enhanced education landscape. The discussion underscores the urgency for proactive adaptation amidst AI's evolution, offering a pragmatic roadmap to navigate the technical, ethical, and policy intricacies of integrating AI in education.
[ { "created": "Mon, 16 Oct 2023 23:30:42 GMT", "version": "v1" }, { "created": "Wed, 22 Nov 2023 16:07:26 GMT", "version": "v2" } ]
2023-11-23
[ [ "Leiker", "Daniel", "" ] ]
The Generative Education (GenEd) Framework explores the transition from Large Language Models (LLMs) to Large Multimodal Models (LMMs) in education, envisioning a harmonious relationship between AI and educators to enhance learning experiences. This paper delves into the potential of LMMs to create personalized, interactive, and emotionally-aware learning environments. Through addressing the Two-Sigma problem and the introduction of a conceptual product named Harmony, the narrative emphasizes educator development, adapting policy frameworks, and fostering cross-sector collaboration to realize the envisioned AI-enhanced education landscape. The discussion underscores the urgency for proactive adaptation amidst AI's evolution, offering a pragmatic roadmap to navigate the technical, ethical, and policy intricacies of integrating AI in education.
1810.01351
Elena Guti\'errez Viedma
Pierre Ganty and Elena Guti\'errez
The Parikh Property for Weighted Context-Free Grammars
29 pages, 2 figures, long version of FSTTCS'18 paper
null
10.4230/LIPIcs.FSTTCS.2018
null
cs.FL
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Parikh's Theorem states that every context-free grammar (CFG) is equivalent to some regular CFG when the ordering of symbols in the words is ignored. The same is not true for the so-called weighted CFGs, which additionally assign a weight to each grammar rule. If the result holds for a given weighted CFG $G$, we say that $G$ satisfies the Parikh property. We prove constructively that the Parikh property holds for every weighted nonexpansive CFG. We also give a decision procedure for the property when the weights are over the rationals.
[ { "created": "Tue, 2 Oct 2018 16:22:27 GMT", "version": "v1" }, { "created": "Thu, 27 Dec 2018 14:49:19 GMT", "version": "v2" }, { "created": "Wed, 19 Jun 2019 01:24:47 GMT", "version": "v3" } ]
2019-06-20
[ [ "Ganty", "Pierre", "" ], [ "Gutiérrez", "Elena", "" ] ]
Parikh's Theorem states that every context-free grammar (CFG) is equivalent to some regular CFG when the ordering of symbols in the words is ignored. The same is not true for the so-called weighted CFGs, which additionally assign a weight to each grammar rule. If the result holds for a given weighted CFG $G$, we say that $G$ satisfies the Parikh property. We prove constructively that the Parikh property holds for every weighted nonexpansive CFG. We also give a decision procedure for the property when the weights are over the rationals.
1705.01823
Michael Vanden Boom
Michael Benedikt, Pierre Bourhis, Michael Vanden Boom
Definability and Interpolation within Decidable Fixpoint Logics
null
Logical Methods in Computer Science, Volume 15, Issue 3 (September 10, 2019) lmcs:4729
10.23638/LMCS-15(3:29)2019
null
cs.LO
http://creativecommons.org/licenses/by/4.0/
We look at characterizing which formulas are expressible in rich decidable logics such as guarded fixpoint logic, unary negation fixpoint logic, and guarded negation fixpoint logic. We consider semantic characterizations of definability, as well as effective characterizations. Our algorithms revolve around a finer analysis of the tree-model property and a refinement of the method of moving back and forth between relational logics and logics over trees.
[ { "created": "Thu, 4 May 2017 12:59:24 GMT", "version": "v1" }, { "created": "Tue, 31 Jul 2018 14:28:42 GMT", "version": "v2" }, { "created": "Sun, 2 Jun 2019 07:39:10 GMT", "version": "v3" }, { "created": "Mon, 9 Sep 2019 15:00:17 GMT", "version": "v4" } ]
2023-06-22
[ [ "Benedikt", "Michael", "" ], [ "Bourhis", "Pierre", "" ], [ "Boom", "Michael Vanden", "" ] ]
We look at characterizing which formulas are expressible in rich decidable logics such as guarded fixpoint logic, unary negation fixpoint logic, and guarded negation fixpoint logic. We consider semantic characterizations of definability, as well as effective characterizations. Our algorithms revolve around a finer analysis of the tree-model property and a refinement of the method of moving back and forth between relational logics and logics over trees.