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
2405.18357
Jundong Xu
Jundong Xu, Hao Fei, Liangming Pan, Qian Liu, Mong-Li Lee, Wynne Hsu
Faithful Logical Reasoning via Symbolic Chain-of-Thought
Accepted by ACL 2024 (main proceeding)
null
null
null
cs.CL
http://creativecommons.org/licenses/by-nc-nd/4.0/
While the recent Chain-of-Thought (CoT) technique enhances the reasoning ability of large language models (LLMs) with the theory of mind, it might still struggle in handling logical reasoning that relies much on symbolic expressions and rigid deducing rules. To strengthen the logical reasoning capability of LLMs, we propose a novel Symbolic Chain-of-Thought, namely SymbCoT, a fully LLM-based framework that integrates symbolic expressions and logic rules with CoT prompting. Technically, building upon an LLM, SymbCoT 1) first translates the natural language context into the symbolic format, and then 2) derives a step-by-step plan to solve the problem with symbolic logical rules, 3) followed by a verifier to check the translation and reasoning chain. Via thorough evaluations on 5 standard datasets with both First-Order Logic and Constraint Optimization symbolic expressions, SymbCoT shows striking improvements over the CoT method consistently, meanwhile refreshing the current state-of-the-art performances. We further demonstrate that our system advances in more faithful, flexible, and explainable logical reasoning. To our knowledge, this is the first to combine symbolic expressions and rules into CoT for logical reasoning with LLMs. Code is open at https://github.com/Aiden0526/SymbCoT.
[ { "created": "Tue, 28 May 2024 16:55:33 GMT", "version": "v1" }, { "created": "Tue, 11 Jun 2024 07:41:03 GMT", "version": "v2" } ]
2024-06-12
[ [ "Xu", "Jundong", "" ], [ "Fei", "Hao", "" ], [ "Pan", "Liangming", "" ], [ "Liu", "Qian", "" ], [ "Lee", "Mong-Li", "" ], [ "Hsu", "Wynne", "" ] ]
While the recent Chain-of-Thought (CoT) technique enhances the reasoning ability of large language models (LLMs) with the theory of mind, it might still struggle in handling logical reasoning that relies much on symbolic expressions and rigid deducing rules. To strengthen the logical reasoning capability of LLMs, we propose a novel Symbolic Chain-of-Thought, namely SymbCoT, a fully LLM-based framework that integrates symbolic expressions and logic rules with CoT prompting. Technically, building upon an LLM, SymbCoT 1) first translates the natural language context into the symbolic format, and then 2) derives a step-by-step plan to solve the problem with symbolic logical rules, 3) followed by a verifier to check the translation and reasoning chain. Via thorough evaluations on 5 standard datasets with both First-Order Logic and Constraint Optimization symbolic expressions, SymbCoT shows striking improvements over the CoT method consistently, meanwhile refreshing the current state-of-the-art performances. We further demonstrate that our system advances in more faithful, flexible, and explainable logical reasoning. To our knowledge, this is the first to combine symbolic expressions and rules into CoT for logical reasoning with LLMs. Code is open at https://github.com/Aiden0526/SymbCoT.
1605.06020
Zeineb Guizani
Zeineb Guizani and Noureddine Hamdi
Spectrum Resource Management and Interference Mitigation for D2D Communications with Awareness of BER Constraint in mmWave 5G Underlay Network
Accepted in IEEE Symposium on Computers and Communications June, 2016
null
null
null
cs.NI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The work presented in this paper deals with the issue of massive demands for higher capacity. For that matter, we investigate the spectrum resource management in outdoor mmWave cell for the uplink of cellular and D2D communications. Indeed, we provide a first insight how to optimize the system performance in terms of achievable throughput while realizing a compromise between the large number of admitted devices and the generated interference constraint. We propose a mathematical formulation of the optimization objective which falls in the mixed integer-real optimization scheme. To overcome its complexity, we apply a heuristic algorithm and test its efficiency through simulation results with a particular regard to the BER impact in the QoS.
[ { "created": "Thu, 19 May 2016 15:53:52 GMT", "version": "v1" } ]
2016-05-20
[ [ "Guizani", "Zeineb", "" ], [ "Hamdi", "Noureddine", "" ] ]
The work presented in this paper deals with the issue of massive demands for higher capacity. For that matter, we investigate the spectrum resource management in outdoor mmWave cell for the uplink of cellular and D2D communications. Indeed, we provide a first insight how to optimize the system performance in terms of achievable throughput while realizing a compromise between the large number of admitted devices and the generated interference constraint. We propose a mathematical formulation of the optimization objective which falls in the mixed integer-real optimization scheme. To overcome its complexity, we apply a heuristic algorithm and test its efficiency through simulation results with a particular regard to the BER impact in the QoS.
2312.04838
Suhas Srinath
Suhas Srinath, Shankhanil Mitra, Shika Rao and Rajiv Soundararajan
Learning Generalizable Perceptual Representations for Data-Efficient No-Reference Image Quality Assessment
Accepted to IEEE/CVF WACV 2024
null
null
null
cs.CV
http://creativecommons.org/licenses/by/4.0/
No-reference (NR) image quality assessment (IQA) is an important tool in enhancing the user experience in diverse visual applications. A major drawback of state-of-the-art NR-IQA techniques is their reliance on a large number of human annotations to train models for a target IQA application. To mitigate this requirement, there is a need for unsupervised learning of generalizable quality representations that capture diverse distortions. We enable the learning of low-level quality features agnostic to distortion types by introducing a novel quality-aware contrastive loss. Further, we leverage the generalizability of vision-language models by fine-tuning one such model to extract high-level image quality information through relevant text prompts. The two sets of features are combined to effectively predict quality by training a simple regressor with very few samples on a target dataset. Additionally, we design zero-shot quality predictions from both pathways in a completely blind setting. Our experiments on diverse datasets encompassing various distortions show the generalizability of the features and their superior performance in the data-efficient and zero-shot settings. Code will be made available at https://github.com/suhas-srinath/GRepQ.
[ { "created": "Fri, 8 Dec 2023 05:24:21 GMT", "version": "v1" } ]
2023-12-11
[ [ "Srinath", "Suhas", "" ], [ "Mitra", "Shankhanil", "" ], [ "Rao", "Shika", "" ], [ "Soundararajan", "Rajiv", "" ] ]
No-reference (NR) image quality assessment (IQA) is an important tool in enhancing the user experience in diverse visual applications. A major drawback of state-of-the-art NR-IQA techniques is their reliance on a large number of human annotations to train models for a target IQA application. To mitigate this requirement, there is a need for unsupervised learning of generalizable quality representations that capture diverse distortions. We enable the learning of low-level quality features agnostic to distortion types by introducing a novel quality-aware contrastive loss. Further, we leverage the generalizability of vision-language models by fine-tuning one such model to extract high-level image quality information through relevant text prompts. The two sets of features are combined to effectively predict quality by training a simple regressor with very few samples on a target dataset. Additionally, we design zero-shot quality predictions from both pathways in a completely blind setting. Our experiments on diverse datasets encompassing various distortions show the generalizability of the features and their superior performance in the data-efficient and zero-shot settings. Code will be made available at https://github.com/suhas-srinath/GRepQ.
1606.01292
Kaisheng Yao
Kaisheng Yao and Baolin Peng and Geoffrey Zweig and Kam-Fai Wong
An Attentional Neural Conversation Model with Improved Specificity
null
null
null
null
cs.CL cs.HC
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In this paper we propose a neural conversation model for conducting dialogues. We demonstrate the use of this model to generate help desk responses, where users are asking questions about PC applications. Our model is distinguished by two characteristics. First, it models intention across turns with a recurrent network, and incorporates an attention model that is conditioned on the representation of intention. Secondly, it avoids generating non-specific responses by incorporating an IDF term in the objective function. The model is evaluated both as a pure generation model in which a help-desk response is generated from scratch, and as a retrieval model with performance measured using recall rates of the correct response. Experimental results indicate that the model outperforms previously proposed neural conversation architectures, and that using specificity in the objective function significantly improves performances for both generation and retrieval.
[ { "created": "Fri, 3 Jun 2016 22:26:01 GMT", "version": "v1" } ]
2016-06-07
[ [ "Yao", "Kaisheng", "" ], [ "Peng", "Baolin", "" ], [ "Zweig", "Geoffrey", "" ], [ "Wong", "Kam-Fai", "" ] ]
In this paper we propose a neural conversation model for conducting dialogues. We demonstrate the use of this model to generate help desk responses, where users are asking questions about PC applications. Our model is distinguished by two characteristics. First, it models intention across turns with a recurrent network, and incorporates an attention model that is conditioned on the representation of intention. Secondly, it avoids generating non-specific responses by incorporating an IDF term in the objective function. The model is evaluated both as a pure generation model in which a help-desk response is generated from scratch, and as a retrieval model with performance measured using recall rates of the correct response. Experimental results indicate that the model outperforms previously proposed neural conversation architectures, and that using specificity in the objective function significantly improves performances for both generation and retrieval.
2201.09208
I-Hsi Kao
I-Hsi Kao, Ya-Zhu Yian, Jian-An Su, Yi-Horng Lai, Jau-Woei Perng, Tung-Li Hsieh, Yi-Shueh Tsai, and Min-Shiu Hsieh
Design of Sensor Fusion Driver Assistance System for Active Pedestrian Safety
The 14th International Conference on Automation Technology (Automation 2017), December 8-10, 2017, Kaohsiung, Taiwan
null
null
null
cs.CV eess.SP
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In this paper, we present a parallel architecture for a sensor fusion detection system that combines a camera and 1D light detection and ranging (lidar) sensor for object detection. The system contains two object detection methods, one based on an optical flow, and the other using lidar. The two sensors can effectively complement the defects of the other. The accurate longitudinal accuracy of the object's location and its lateral movement information can be achieved simultaneously. Using a spatio-temporal alignment and a policy of sensor fusion, we completed the development of a fusion detection system with high reliability at distances of up to 20 m. Test results show that the proposed system achieves a high level of accuracy for pedestrian or object detection in front of a vehicle, and has high robustness to special environments.
[ { "created": "Sun, 23 Jan 2022 08:52:32 GMT", "version": "v1" } ]
2022-01-25
[ [ "Kao", "I-Hsi", "" ], [ "Yian", "Ya-Zhu", "" ], [ "Su", "Jian-An", "" ], [ "Lai", "Yi-Horng", "" ], [ "Perng", "Jau-Woei", "" ], [ "Hsieh", "Tung-Li", "" ], [ "Tsai", "Yi-Shueh", "" ], [ "Hsieh", "Min-Shiu", "" ] ]
In this paper, we present a parallel architecture for a sensor fusion detection system that combines a camera and 1D light detection and ranging (lidar) sensor for object detection. The system contains two object detection methods, one based on an optical flow, and the other using lidar. The two sensors can effectively complement the defects of the other. The accurate longitudinal accuracy of the object's location and its lateral movement information can be achieved simultaneously. Using a spatio-temporal alignment and a policy of sensor fusion, we completed the development of a fusion detection system with high reliability at distances of up to 20 m. Test results show that the proposed system achieves a high level of accuracy for pedestrian or object detection in front of a vehicle, and has high robustness to special environments.
2402.09500
Matthew Fox
Matthew Fox
On Formally Undecidable Traits of Intelligent Machines
34 pages
null
null
null
cs.AI cs.LO
http://creativecommons.org/licenses/by/4.0/
Building on work by Alfonseca et al. (2021), we study the conditions necessary for it to be logically possible to prove that an arbitrary artificially intelligent machine will exhibit certain behavior. To do this, we develop a formalism like -- but mathematically distinct from -- the theory of formal languages and their properties. Our formalism affords a precise means for not only talking about the traits we desire of machines (such as them being intelligent, contained, moral, and so forth), but also for detailing the conditions necessary for it to be logically possible to decide whether a given arbitrary machine possesses such a trait or not. Contrary to Alfonseca et al.'s (2021) results, we find that Rice's theorem from computability theory cannot in general be used to determine whether an arbitrary machine possesses a given trait or not. Therefore, it is not necessarily the case that deciding whether an arbitrary machine is intelligent, contained, moral, and so forth is logically impossible.
[ { "created": "Wed, 14 Feb 2024 18:59:37 GMT", "version": "v1" } ]
2024-02-16
[ [ "Fox", "Matthew", "" ] ]
Building on work by Alfonseca et al. (2021), we study the conditions necessary for it to be logically possible to prove that an arbitrary artificially intelligent machine will exhibit certain behavior. To do this, we develop a formalism like -- but mathematically distinct from -- the theory of formal languages and their properties. Our formalism affords a precise means for not only talking about the traits we desire of machines (such as them being intelligent, contained, moral, and so forth), but also for detailing the conditions necessary for it to be logically possible to decide whether a given arbitrary machine possesses such a trait or not. Contrary to Alfonseca et al.'s (2021) results, we find that Rice's theorem from computability theory cannot in general be used to determine whether an arbitrary machine possesses a given trait or not. Therefore, it is not necessarily the case that deciding whether an arbitrary machine is intelligent, contained, moral, and so forth is logically impossible.
2401.15720
Anat Hashavit
Anat Hashavit, Tamar Stern, Hongning Wang, Sarit Kraus
The Impact of Snippet Reliability on Misinformation in Online Health Search
null
null
null
null
cs.IR
http://creativecommons.org/licenses/by-nc-sa/4.0/
Search result snippets are crucial in modern search engines, providing users with a quick overview of a website's content. Snippets help users determine the relevance of a document to their information needs, and in certain scenarios even enable them to satisfy those needs without visiting web documents. Hence, it is crucial for snippets to reliably represent the content of their corresponding documents. While this may be a straightforward requirement for some queries, it can become challenging in the complex domain of healthcare, and can lead to misinformation. This paper aims to examine snippets' reliability in representing their corresponding documents, specifically in the health domain. To achieve this, we conduct a series of user studies using Google's search results, where participants are asked to infer viewpoints of search results pertaining to queries about the effectiveness of a medical intervention for a medical condition, based solely on their titles and snippets. Our findings reveal that a considerable portion of Google's snippets (28%) failed to present any viewpoint on the intervention's effectiveness, and that 35% were interpreted by participants as having a different viewpoint compared to their corresponding documents. To address this issue, we propose a snippet extraction solution tailored directly to users' information needs, i.e., extracting snippets that summarize documents' viewpoints regarding the intervention and condition that appear in the query. User study demonstrates that our information need-focused solution outperforms the mainstream query-based approach. With only 19.67% of snippets generated by our solution reported as not presenting a viewpoint and a mere 20.33% misinterpreted by participants. These results strongly suggest that an information need-focused approach can significantly improve the reliability of extracted snippets in online health search.
[ { "created": "Sun, 28 Jan 2024 17:59:55 GMT", "version": "v1" } ]
2024-01-30
[ [ "Hashavit", "Anat", "" ], [ "Stern", "Tamar", "" ], [ "Wang", "Hongning", "" ], [ "Kraus", "Sarit", "" ] ]
Search result snippets are crucial in modern search engines, providing users with a quick overview of a website's content. Snippets help users determine the relevance of a document to their information needs, and in certain scenarios even enable them to satisfy those needs without visiting web documents. Hence, it is crucial for snippets to reliably represent the content of their corresponding documents. While this may be a straightforward requirement for some queries, it can become challenging in the complex domain of healthcare, and can lead to misinformation. This paper aims to examine snippets' reliability in representing their corresponding documents, specifically in the health domain. To achieve this, we conduct a series of user studies using Google's search results, where participants are asked to infer viewpoints of search results pertaining to queries about the effectiveness of a medical intervention for a medical condition, based solely on their titles and snippets. Our findings reveal that a considerable portion of Google's snippets (28%) failed to present any viewpoint on the intervention's effectiveness, and that 35% were interpreted by participants as having a different viewpoint compared to their corresponding documents. To address this issue, we propose a snippet extraction solution tailored directly to users' information needs, i.e., extracting snippets that summarize documents' viewpoints regarding the intervention and condition that appear in the query. User study demonstrates that our information need-focused solution outperforms the mainstream query-based approach. With only 19.67% of snippets generated by our solution reported as not presenting a viewpoint and a mere 20.33% misinterpreted by participants. These results strongly suggest that an information need-focused approach can significantly improve the reliability of extracted snippets in online health search.
1109.3437
Jia Zeng
Jia Zeng and William K. Cheung and Jiming Liu
Learning Topic Models by Belief Propagation
14 pages, 17 figures
IEEE Transactions on Pattern Analysis and Machine Intelligence, Volume 33, Number 5, Pages 1121-1134, 2013
10.1109/TPAMI.2012.185
null
cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Latent Dirichlet allocation (LDA) is an important hierarchical Bayesian model for probabilistic topic modeling, which attracts worldwide interests and touches on many important applications in text mining, computer vision and computational biology. This paper represents LDA as a factor graph within the Markov random field (MRF) framework, which enables the classic loopy belief propagation (BP) algorithm for approximate inference and parameter estimation. Although two commonly-used approximate inference methods, such as variational Bayes (VB) and collapsed Gibbs sampling (GS), have gained great successes in learning LDA, the proposed BP is competitive in both speed and accuracy as validated by encouraging experimental results on four large-scale document data sets. Furthermore, the BP algorithm has the potential to become a generic learning scheme for variants of LDA-based topic models. To this end, we show how to learn two typical variants of LDA-based topic models, such as author-topic models (ATM) and relational topic models (RTM), using BP based on the factor graph representation.
[ { "created": "Thu, 15 Sep 2011 19:20:48 GMT", "version": "v1" }, { "created": "Sun, 25 Sep 2011 21:17:41 GMT", "version": "v2" }, { "created": "Mon, 3 Oct 2011 03:17:44 GMT", "version": "v3" }, { "created": "Sat, 24 Mar 2012 12:47:02 GMT", "version": "v4" } ]
2015-03-19
[ [ "Zeng", "Jia", "" ], [ "Cheung", "William K.", "" ], [ "Liu", "Jiming", "" ] ]
Latent Dirichlet allocation (LDA) is an important hierarchical Bayesian model for probabilistic topic modeling, which attracts worldwide interests and touches on many important applications in text mining, computer vision and computational biology. This paper represents LDA as a factor graph within the Markov random field (MRF) framework, which enables the classic loopy belief propagation (BP) algorithm for approximate inference and parameter estimation. Although two commonly-used approximate inference methods, such as variational Bayes (VB) and collapsed Gibbs sampling (GS), have gained great successes in learning LDA, the proposed BP is competitive in both speed and accuracy as validated by encouraging experimental results on four large-scale document data sets. Furthermore, the BP algorithm has the potential to become a generic learning scheme for variants of LDA-based topic models. To this end, we show how to learn two typical variants of LDA-based topic models, such as author-topic models (ATM) and relational topic models (RTM), using BP based on the factor graph representation.
2108.09105
Pierre-Louis Guhur
Pierre-Louis Guhur, Makarand Tapaswi, Shizhe Chen, Ivan Laptev, Cordelia Schmid
Airbert: In-domain Pretraining for Vision-and-Language Navigation
To be published on ICCV 2021. Webpage is at https://airbert-vln.github.io/ linking to our dataset, codes and models
null
null
null
cs.CV cs.AI cs.CL cs.HC cs.LG
http://creativecommons.org/licenses/by/4.0/
Vision-and-language navigation (VLN) aims to enable embodied agents to navigate in realistic environments using natural language instructions. Given the scarcity of domain-specific training data and the high diversity of image and language inputs, the generalization of VLN agents to unseen environments remains challenging. Recent methods explore pretraining to improve generalization, however, the use of generic image-caption datasets or existing small-scale VLN environments is suboptimal and results in limited improvements. In this work, we introduce BnB, a large-scale and diverse in-domain VLN dataset. We first collect image-caption (IC) pairs from hundreds of thousands of listings from online rental marketplaces. Using IC pairs we next propose automatic strategies to generate millions of VLN path-instruction (PI) pairs. We further propose a shuffling loss that improves the learning of temporal order inside PI pairs. We use BnB pretrain our Airbert model that can be adapted to discriminative and generative settings and show that it outperforms state of the art for Room-to-Room (R2R) navigation and Remote Referring Expression (REVERIE) benchmarks. Moreover, our in-domain pretraining significantly increases performance on a challenging few-shot VLN evaluation, where we train the model only on VLN instructions from a few houses.
[ { "created": "Fri, 20 Aug 2021 10:58:09 GMT", "version": "v1" } ]
2021-08-23
[ [ "Guhur", "Pierre-Louis", "" ], [ "Tapaswi", "Makarand", "" ], [ "Chen", "Shizhe", "" ], [ "Laptev", "Ivan", "" ], [ "Schmid", "Cordelia", "" ] ]
Vision-and-language navigation (VLN) aims to enable embodied agents to navigate in realistic environments using natural language instructions. Given the scarcity of domain-specific training data and the high diversity of image and language inputs, the generalization of VLN agents to unseen environments remains challenging. Recent methods explore pretraining to improve generalization, however, the use of generic image-caption datasets or existing small-scale VLN environments is suboptimal and results in limited improvements. In this work, we introduce BnB, a large-scale and diverse in-domain VLN dataset. We first collect image-caption (IC) pairs from hundreds of thousands of listings from online rental marketplaces. Using IC pairs we next propose automatic strategies to generate millions of VLN path-instruction (PI) pairs. We further propose a shuffling loss that improves the learning of temporal order inside PI pairs. We use BnB pretrain our Airbert model that can be adapted to discriminative and generative settings and show that it outperforms state of the art for Room-to-Room (R2R) navigation and Remote Referring Expression (REVERIE) benchmarks. Moreover, our in-domain pretraining significantly increases performance on a challenging few-shot VLN evaluation, where we train the model only on VLN instructions from a few houses.
2405.02782
David Wood
David A. Wood, Emily Guilhem, Sina Kafiabadi, Ayisha Al Busaidi, Kishan Dissanayake, Ahmed Hammam, Nina Mansoor, Matthew Townend, Siddharth Agarwal, Yiran Wei, Asif Mazumder, Gareth J. Barker, Peter Sasieni, Sebastien Ourselin, James H. Cole, Thomas C. Booth
A self-supervised text-vision framework for automated brain abnormality detection
Under Review
null
null
null
cs.CV
http://creativecommons.org/licenses/by/4.0/
Artificial neural networks trained on large, expert-labelled datasets are considered state-of-the-art for a range of medical image recognition tasks. However, categorically labelled datasets are time-consuming to generate and constrain classification to a pre-defined, fixed set of classes. For neuroradiological applications in particular, this represents a barrier to clinical adoption. To address these challenges, we present a self-supervised text-vision framework that learns to detect clinically relevant abnormalities in brain MRI scans by directly leveraging the rich information contained in accompanying free-text neuroradiology reports. Our training approach consisted of two-steps. First, a dedicated neuroradiological language model - NeuroBERT - was trained to generate fixed-dimensional vector representations of neuroradiology reports (N = 50,523) via domain-specific self-supervised learning tasks. Next, convolutional neural networks (one per MRI sequence) learnt to map individual brain scans to their corresponding text vector representations by optimising a mean square error loss. Once trained, our text-vision framework can be used to detect abnormalities in unreported brain MRI examinations by scoring scans against suitable query sentences (e.g., 'there is an acute stroke', 'there is hydrocephalus' etc.), enabling a range of classification-based applications including automated triage. Potentially, our framework could also serve as a clinical decision support tool, not only by suggesting findings to radiologists and detecting errors in provisional reports, but also by retrieving and displaying examples of pathologies from historical examinations that could be relevant to the current case based on textual descriptors.
[ { "created": "Sun, 5 May 2024 01:51:58 GMT", "version": "v1" }, { "created": "Wed, 12 Jun 2024 01:01:51 GMT", "version": "v2" } ]
2024-06-13
[ [ "Wood", "David A.", "" ], [ "Guilhem", "Emily", "" ], [ "Kafiabadi", "Sina", "" ], [ "Busaidi", "Ayisha Al", "" ], [ "Dissanayake", "Kishan", "" ], [ "Hammam", "Ahmed", "" ], [ "Mansoor", "Nina", "" ], [ "Townend", "Matthew", "" ], [ "Agarwal", "Siddharth", "" ], [ "Wei", "Yiran", "" ], [ "Mazumder", "Asif", "" ], [ "Barker", "Gareth J.", "" ], [ "Sasieni", "Peter", "" ], [ "Ourselin", "Sebastien", "" ], [ "Cole", "James H.", "" ], [ "Booth", "Thomas C.", "" ] ]
Artificial neural networks trained on large, expert-labelled datasets are considered state-of-the-art for a range of medical image recognition tasks. However, categorically labelled datasets are time-consuming to generate and constrain classification to a pre-defined, fixed set of classes. For neuroradiological applications in particular, this represents a barrier to clinical adoption. To address these challenges, we present a self-supervised text-vision framework that learns to detect clinically relevant abnormalities in brain MRI scans by directly leveraging the rich information contained in accompanying free-text neuroradiology reports. Our training approach consisted of two-steps. First, a dedicated neuroradiological language model - NeuroBERT - was trained to generate fixed-dimensional vector representations of neuroradiology reports (N = 50,523) via domain-specific self-supervised learning tasks. Next, convolutional neural networks (one per MRI sequence) learnt to map individual brain scans to their corresponding text vector representations by optimising a mean square error loss. Once trained, our text-vision framework can be used to detect abnormalities in unreported brain MRI examinations by scoring scans against suitable query sentences (e.g., 'there is an acute stroke', 'there is hydrocephalus' etc.), enabling a range of classification-based applications including automated triage. Potentially, our framework could also serve as a clinical decision support tool, not only by suggesting findings to radiologists and detecting errors in provisional reports, but also by retrieving and displaying examples of pathologies from historical examinations that could be relevant to the current case based on textual descriptors.
2207.08536
Xi Li
Zequn Qin, Jingyu Chen, Chao Chen, Xiaozhi Chen, Xi Li
UniFusion: Unified Multi-view Fusion Transformer for Spatial-Temporal Representation in Bird's-Eye-View
null
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Bird's eye view (BEV) representation is a new perception formulation for autonomous driving, which is based on spatial fusion. Further, temporal fusion is also introduced in BEV representation and gains great success. In this work, we propose a new method that unifies both spatial and temporal fusion and merges them into a unified mathematical formulation. The unified fusion could not only provide a new perspective on BEV fusion but also brings new capabilities. With the proposed unified spatial-temporal fusion, our method could support long-range fusion, which is hard to achieve in conventional BEV methods. Moreover, the BEV fusion in our work is temporal-adaptive and the weights of temporal fusion are learnable. In contrast, conventional methods mainly use fixed and equal weights for temporal fusion. Besides, the proposed unified fusion could avoid information lost in conventional BEV fusion methods and make full use of features. Extensive experiments and ablation studies on the NuScenes dataset show the effectiveness of the proposed method and our method gains the state-of-the-art performance in the map segmentation task.
[ { "created": "Mon, 18 Jul 2022 11:59:10 GMT", "version": "v1" }, { "created": "Mon, 20 Mar 2023 03:12:09 GMT", "version": "v2" } ]
2023-03-21
[ [ "Qin", "Zequn", "" ], [ "Chen", "Jingyu", "" ], [ "Chen", "Chao", "" ], [ "Chen", "Xiaozhi", "" ], [ "Li", "Xi", "" ] ]
Bird's eye view (BEV) representation is a new perception formulation for autonomous driving, which is based on spatial fusion. Further, temporal fusion is also introduced in BEV representation and gains great success. In this work, we propose a new method that unifies both spatial and temporal fusion and merges them into a unified mathematical formulation. The unified fusion could not only provide a new perspective on BEV fusion but also brings new capabilities. With the proposed unified spatial-temporal fusion, our method could support long-range fusion, which is hard to achieve in conventional BEV methods. Moreover, the BEV fusion in our work is temporal-adaptive and the weights of temporal fusion are learnable. In contrast, conventional methods mainly use fixed and equal weights for temporal fusion. Besides, the proposed unified fusion could avoid information lost in conventional BEV fusion methods and make full use of features. Extensive experiments and ablation studies on the NuScenes dataset show the effectiveness of the proposed method and our method gains the state-of-the-art performance in the map segmentation task.
2309.13808
EPTCS
Dafina Trufa\c{s} (University of Bucharest), Ioan Teodorescu (University of Bucharest), Denisa Diaconescu (University of Bucharest), Traian \c{S}erb\u{a}nu\c{t}\u{a} (University of Bucharest), Vlad Zamfir (independent researcher)
Asynchronous Muddy Children Puzzle (work in progress)
In Proceedings FROM 2023, arXiv:2309.12959
EPTCS 389, 2023, pp. 152-166
10.4204/EPTCS.389.13
null
cs.LO cs.MA
http://creativecommons.org/licenses/by/4.0/
In this work-in-progress paper we explore using the recently introduced VLSM formalism to define and reason about the dynamics of agent-based systems. To this aim we use VLSMs to formally present several possible approaches to modeling the interactions in the Muddy Children Puzzle as protocols that reach consensus asynchronously.
[ { "created": "Mon, 25 Sep 2023 01:24:21 GMT", "version": "v1" } ]
2023-09-26
[ [ "Trufaş", "Dafina", "", "University of Bucharest" ], [ "Teodorescu", "Ioan", "", "University of Bucharest" ], [ "Diaconescu", "Denisa", "", "University of Bucharest" ], [ "Şerbănuţă", "Traian", "", "University of Bucharest" ], [ "Zamfir", "Vlad", "", "independent researcher" ] ]
In this work-in-progress paper we explore using the recently introduced VLSM formalism to define and reason about the dynamics of agent-based systems. To this aim we use VLSMs to formally present several possible approaches to modeling the interactions in the Muddy Children Puzzle as protocols that reach consensus asynchronously.
1211.6940
Keehang Kwon
Keehang Kwon and Daeseong Kang
Choice Disjunctive Queries in Logic Programming
IEICE transaction on Information and Systems (to appear)
IEICE transaction on Information and Systems vol.E106-D,No.3, 2023
null
null
cs.LO cs.PL
http://creativecommons.org/licenses/by/4.0/
One of the long-standing research problems on logic programming is to treat the cut predicate in a logical, high-level way. We argue that this problem can be solved by adopting linear logic and choice-disjunctive goal formulas of the form $G_0 \add G_1$ where $G_0, G_1$ are goals. These goals have the following intended semantics: $choose$ the true disjunct $G_i$ and execute $G_i$ where $i (= 0\ {\rm or}\ 1)$, while $discarding$ the unchosen disjunct. Note that only one goal can remain alive during execution. These goals thus allow us to specify mutually exclusive tasks in a high-level way.
[ { "created": "Thu, 29 Nov 2012 15:04:31 GMT", "version": "v1" }, { "created": "Mon, 22 Sep 2014 03:30:04 GMT", "version": "v2" }, { "created": "Fri, 4 Sep 2015 06:27:20 GMT", "version": "v3" }, { "created": "Tue, 11 Oct 2022 04:28:33 GMT", "version": "v4" } ]
2023-01-31
[ [ "Kwon", "Keehang", "" ], [ "Kang", "Daeseong", "" ] ]
One of the long-standing research problems on logic programming is to treat the cut predicate in a logical, high-level way. We argue that this problem can be solved by adopting linear logic and choice-disjunctive goal formulas of the form $G_0 \add G_1$ where $G_0, G_1$ are goals. These goals have the following intended semantics: $choose$ the true disjunct $G_i$ and execute $G_i$ where $i (= 0\ {\rm or}\ 1)$, while $discarding$ the unchosen disjunct. Note that only one goal can remain alive during execution. These goals thus allow us to specify mutually exclusive tasks in a high-level way.
1508.02439
Di Wang
Di Wang, Satish Rao, Michael W. Mahoney
Unified Acceleration Method for Packing and Covering Problems via Diameter Reduction
Fixed typo in packing LP formulation (page 1), and wrong citation in the discussion of earlier works on page 2
null
null
null
cs.DS cs.NA
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The linear coupling method was introduced recently by Allen-Zhu and Orecchia for solving convex optimization problems with first order methods, and it provides a conceptually simple way to integrate a gradient descent step and mirror descent step in each iteration. The high-level approach of the linear coupling method is very flexible, and it has shown initial promise by providing improved algorithms for packing and covering linear programs. Somewhat surprisingly, however, while the dependence of the convergence rate on the error parameter $\epsilon$ for packing problems was improved to $O(1/\epsilon)$, which corresponds to what accelerated gradient methods are designed to achieve, the dependence for covering problems was only improved to $O(1/\epsilon^{1.5})$, and even that required a different more complicated algorithm. Given the close connections between packing and covering problems and since previous algorithms for these very related problems have led to the same $\epsilon$ dependence, this discrepancy is surprising, and it leaves open the question of the exact role that the linear coupling is playing in coordinating the complementary gradient and mirror descent step of the algorithm. In this paper, we clarify these issues for linear coupling algorithms for packing and covering linear programs, illustrating that the linear coupling method can lead to improved $O(1/\epsilon)$ dependence for both packing and covering problems in a unified manner, i.e., with the same algorithm and almost identical analysis. Our main technical result is a novel diameter reduction method for covering problems that is of independent interest and that may be useful in applying the accelerated linear coupling method to other combinatorial problems.
[ { "created": "Mon, 10 Aug 2015 21:56:20 GMT", "version": "v1" }, { "created": "Tue, 6 Oct 2015 06:41:38 GMT", "version": "v2" } ]
2015-10-07
[ [ "Wang", "Di", "" ], [ "Rao", "Satish", "" ], [ "Mahoney", "Michael W.", "" ] ]
The linear coupling method was introduced recently by Allen-Zhu and Orecchia for solving convex optimization problems with first order methods, and it provides a conceptually simple way to integrate a gradient descent step and mirror descent step in each iteration. The high-level approach of the linear coupling method is very flexible, and it has shown initial promise by providing improved algorithms for packing and covering linear programs. Somewhat surprisingly, however, while the dependence of the convergence rate on the error parameter $\epsilon$ for packing problems was improved to $O(1/\epsilon)$, which corresponds to what accelerated gradient methods are designed to achieve, the dependence for covering problems was only improved to $O(1/\epsilon^{1.5})$, and even that required a different more complicated algorithm. Given the close connections between packing and covering problems and since previous algorithms for these very related problems have led to the same $\epsilon$ dependence, this discrepancy is surprising, and it leaves open the question of the exact role that the linear coupling is playing in coordinating the complementary gradient and mirror descent step of the algorithm. In this paper, we clarify these issues for linear coupling algorithms for packing and covering linear programs, illustrating that the linear coupling method can lead to improved $O(1/\epsilon)$ dependence for both packing and covering problems in a unified manner, i.e., with the same algorithm and almost identical analysis. Our main technical result is a novel diameter reduction method for covering problems that is of independent interest and that may be useful in applying the accelerated linear coupling method to other combinatorial problems.
1812.10037
Jianpeng Cheng J
Jianpeng Cheng and Siva Reddy and Mirella Lapata
Building a Neural Semantic Parser from a Domain Ontology
null
null
null
null
cs.CL
http://creativecommons.org/licenses/by/4.0/
Semantic parsing is the task of converting natural language utterances into machine interpretable meaning representations which can be executed against a real-world environment such as a database. Scaling semantic parsing to arbitrary domains faces two interrelated challenges: obtaining broad coverage training data effectively and cheaply; and developing a model that generalizes to compositional utterances and complex intentions. We address these challenges with a framework which allows to elicit training data from a domain ontology and bootstrap a neural parser which recursively builds derivations of logical forms. In our framework meaning representations are described by sequences of natural language templates, where each template corresponds to a decomposed fragment of the underlying meaning representation. Although artificial, templates can be understood and paraphrased by humans to create natural utterances, resulting in parallel triples of utterances, meaning representations, and their decompositions. These allow us to train a neural semantic parser which learns to compose rules in deriving meaning representations. We crowdsource training data on six domains, covering both single-turn utterances which exhibit rich compositionality, and sequential utterances where a complex task is procedurally performed in steps. We then develop neural semantic parsers which perform such compositional tasks. In general, our approach allows to deploy neural semantic parsers quickly and cheaply from a given domain ontology.
[ { "created": "Tue, 25 Dec 2018 05:30:18 GMT", "version": "v1" } ]
2018-12-27
[ [ "Cheng", "Jianpeng", "" ], [ "Reddy", "Siva", "" ], [ "Lapata", "Mirella", "" ] ]
Semantic parsing is the task of converting natural language utterances into machine interpretable meaning representations which can be executed against a real-world environment such as a database. Scaling semantic parsing to arbitrary domains faces two interrelated challenges: obtaining broad coverage training data effectively and cheaply; and developing a model that generalizes to compositional utterances and complex intentions. We address these challenges with a framework which allows to elicit training data from a domain ontology and bootstrap a neural parser which recursively builds derivations of logical forms. In our framework meaning representations are described by sequences of natural language templates, where each template corresponds to a decomposed fragment of the underlying meaning representation. Although artificial, templates can be understood and paraphrased by humans to create natural utterances, resulting in parallel triples of utterances, meaning representations, and their decompositions. These allow us to train a neural semantic parser which learns to compose rules in deriving meaning representations. We crowdsource training data on six domains, covering both single-turn utterances which exhibit rich compositionality, and sequential utterances where a complex task is procedurally performed in steps. We then develop neural semantic parsers which perform such compositional tasks. In general, our approach allows to deploy neural semantic parsers quickly and cheaply from a given domain ontology.
2404.08259
Udo Kruschwitz
Wan-Hua Her and Udo Kruschwitz
Investigating Neural Machine Translation for Low-Resource Languages: Using Bavarian as a Case Study
Preprint accepted at the 3rd Annual Meeting of the Special Interest Group on Under-resourced Languages (SIGUL 2024)
null
null
null
cs.CL
http://creativecommons.org/licenses/by/4.0/
Machine Translation has made impressive progress in recent years offering close to human-level performance on many languages, but studies have primarily focused on high-resource languages with broad online presence and resources. With the help of growing Large Language Models, more and more low-resource languages achieve better results through the presence of other languages. However, studies have shown that not all low-resource languages can benefit from multilingual systems, especially those with insufficient training and evaluation data. In this paper, we revisit state-of-the-art Neural Machine Translation techniques to develop automatic translation systems between German and Bavarian. We investigate conditions of low-resource languages such as data scarcity and parameter sensitivity and focus on refined solutions that combat low-resource difficulties and creative solutions such as harnessing language similarity. Our experiment entails applying Back-translation and Transfer Learning to automatically generate more training data and achieve higher translation performance. We demonstrate noisiness in the data and present our approach to carry out text preprocessing extensively. Evaluation was conducted using combined metrics: BLEU, chrF and TER. Statistical significance results with Bonferroni correction show surprisingly high baseline systems, and that Back-translation leads to significant improvement. Furthermore, we present a qualitative analysis of translation errors and system limitations.
[ { "created": "Fri, 12 Apr 2024 06:16:26 GMT", "version": "v1" } ]
2024-04-15
[ [ "Her", "Wan-Hua", "" ], [ "Kruschwitz", "Udo", "" ] ]
Machine Translation has made impressive progress in recent years offering close to human-level performance on many languages, but studies have primarily focused on high-resource languages with broad online presence and resources. With the help of growing Large Language Models, more and more low-resource languages achieve better results through the presence of other languages. However, studies have shown that not all low-resource languages can benefit from multilingual systems, especially those with insufficient training and evaluation data. In this paper, we revisit state-of-the-art Neural Machine Translation techniques to develop automatic translation systems between German and Bavarian. We investigate conditions of low-resource languages such as data scarcity and parameter sensitivity and focus on refined solutions that combat low-resource difficulties and creative solutions such as harnessing language similarity. Our experiment entails applying Back-translation and Transfer Learning to automatically generate more training data and achieve higher translation performance. We demonstrate noisiness in the data and present our approach to carry out text preprocessing extensively. Evaluation was conducted using combined metrics: BLEU, chrF and TER. Statistical significance results with Bonferroni correction show surprisingly high baseline systems, and that Back-translation leads to significant improvement. Furthermore, we present a qualitative analysis of translation errors and system limitations.
1303.2211
Nilanjan Dey
Nilanjan Dey, Suvojit Acharjee, Debalina Biswas, Achintya Das, Sheli Sinha Chaudhuri
Medical Information Embedding in Compressed Watermarked Intravascular Ultrasound Video
Pages-7 Fig.-15 Tables-2
Scientific Bulletin of the Politehnica University of Timisoara - Transactions on Electronics and Communications p-ISSN 1583-3380 , vol. 57(71), no. 2, 2012
null
null
cs.MM cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In medical field, intravascular ultrasound (IVUS) is a tomographic imaging modality, which can identify the boundaries of different layers of blood vessels. IVUS can detect myocardial infarction (heart attack) that remains ignored and unattended when only angioplasty is done. During the past decade, it became easier for some individuals or groups to copy and transmits digital information without the permission of the owner. For increasing authentication and security of copyrights, digital watermarking, an information hiding technique, was introduced. Achieving watermarking technique with lesser amount of distortion in biomedical data is a challenging task. Watermark can be embedded into an image or in a video. As video data is a huge amount of information, therefore a large storage area is needed which is not feasible. In this case motion vector based video compression is done to reduce size. In this present paper, an Electronic Patient Record (EPR) is embedded as watermark within an IVUS video and then motion vector is calculated. This proposed method proves robustness as the extracted watermark has good PSNR value and less MSE.
[ { "created": "Sat, 9 Mar 2013 14:08:23 GMT", "version": "v1" } ]
2013-03-12
[ [ "Dey", "Nilanjan", "" ], [ "Acharjee", "Suvojit", "" ], [ "Biswas", "Debalina", "" ], [ "Das", "Achintya", "" ], [ "Chaudhuri", "Sheli Sinha", "" ] ]
In medical field, intravascular ultrasound (IVUS) is a tomographic imaging modality, which can identify the boundaries of different layers of blood vessels. IVUS can detect myocardial infarction (heart attack) that remains ignored and unattended when only angioplasty is done. During the past decade, it became easier for some individuals or groups to copy and transmits digital information without the permission of the owner. For increasing authentication and security of copyrights, digital watermarking, an information hiding technique, was introduced. Achieving watermarking technique with lesser amount of distortion in biomedical data is a challenging task. Watermark can be embedded into an image or in a video. As video data is a huge amount of information, therefore a large storage area is needed which is not feasible. In this case motion vector based video compression is done to reduce size. In this present paper, an Electronic Patient Record (EPR) is embedded as watermark within an IVUS video and then motion vector is calculated. This proposed method proves robustness as the extracted watermark has good PSNR value and less MSE.
1911.07292
Hufei Zhu
Hufei Zhu
Two Efficient Ridge Solutions for the Incremental Broad Learning System on Added Inputs
arXiv admin note: text overlap with arXiv:1911.04872
null
null
null
cs.LG stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This paper proposes the recursive and square-root BLS algorithms to improve the original BLS for new added inputs, which utilize the inverse and inverse Cholesky factor of the Hermitian matrix in the ridge inverse, respectively, to update the ridge solution. The recursive BLS updates the inverse by the matrix inversion lemma, while the square-root BLS updates the upper-triangular inverse Cholesky factor by multiplying it with an upper-triangular intermediate matrix. When the added p training samples are more than the total k nodes in the network, i.e., p>k, the inverse of a sum of matrices is applied to take a smaller matrix inversion or inverse Cholesky factorization. For the distributed BLS with data-parallelism, we introduce the parallel implementation of the square-root BLS, which is deduced from the parallel implementation of the inverse Cholesky factorization. The original BLS based on the generalized inverse with the ridge regression assumes the ridge parameter lamda->0 in the ridge inverse. When lambda->0 is not satisfied, the numerical experiments on the MNIST and NORB datasets show that both the proposed ridge solutions improve the testing accuracy of the original BLS, and the improvement becomes more significant as lambda is bigger. On the other hand, compared to the original BLS, both the proposed BLS algorithms theoretically require less complexities, and are significantly faster in the simulations on the MNIST dataset. The speedups in total training time of the recursive and square-root BLS algorithms over the original BLS are 4.41 and 6.92 respectively when p > k, and are 2.80 and 1.59 respectively when p < k.
[ { "created": "Tue, 12 Nov 2019 14:19:52 GMT", "version": "v1" }, { "created": "Tue, 13 Apr 2021 04:36:01 GMT", "version": "v2" }, { "created": "Fri, 16 Apr 2021 06:34:18 GMT", "version": "v3" }, { "created": "Mon, 22 Nov 2021 14:12:51 GMT", "version": "v4" }, { "created": "Wed, 25 Jan 2023 02:35:55 GMT", "version": "v5" } ]
2023-01-26
[ [ "Zhu", "Hufei", "" ] ]
This paper proposes the recursive and square-root BLS algorithms to improve the original BLS for new added inputs, which utilize the inverse and inverse Cholesky factor of the Hermitian matrix in the ridge inverse, respectively, to update the ridge solution. The recursive BLS updates the inverse by the matrix inversion lemma, while the square-root BLS updates the upper-triangular inverse Cholesky factor by multiplying it with an upper-triangular intermediate matrix. When the added p training samples are more than the total k nodes in the network, i.e., p>k, the inverse of a sum of matrices is applied to take a smaller matrix inversion or inverse Cholesky factorization. For the distributed BLS with data-parallelism, we introduce the parallel implementation of the square-root BLS, which is deduced from the parallel implementation of the inverse Cholesky factorization. The original BLS based on the generalized inverse with the ridge regression assumes the ridge parameter lamda->0 in the ridge inverse. When lambda->0 is not satisfied, the numerical experiments on the MNIST and NORB datasets show that both the proposed ridge solutions improve the testing accuracy of the original BLS, and the improvement becomes more significant as lambda is bigger. On the other hand, compared to the original BLS, both the proposed BLS algorithms theoretically require less complexities, and are significantly faster in the simulations on the MNIST dataset. The speedups in total training time of the recursive and square-root BLS algorithms over the original BLS are 4.41 and 6.92 respectively when p > k, and are 2.80 and 1.59 respectively when p < k.
2103.07658
Mallikarjun Byrasandra Ramalinga Reddy
Mallikarjun B R, Ayush Tewari, Abdallah Dib, Tim Weyrich, Bernd Bickel, Hans-Peter Seidel, Hanspeter Pfister, Wojciech Matusik, Louis Chevallier, Mohamed Elgharib, Christian Theobalt
PhotoApp: Photorealistic Appearance Editing of Head Portraits
http://gvv.mpi-inf.mpg.de/projects/PhotoApp/
null
null
null
cs.CV cs.GR cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Photorealistic editing of portraits is a challenging task as humans are very sensitive to inconsistencies in faces. We present an approach for high-quality intuitive editing of the camera viewpoint and scene illumination in a portrait image. This requires our method to capture and control the full reflectance field of the person in the image. Most editing approaches rely on supervised learning using training data captured with setups such as light and camera stages. Such datasets are expensive to acquire, not readily available and do not capture all the rich variations of in-the-wild portrait images. In addition, most supervised approaches only focus on relighting, and do not allow camera viewpoint editing. Thus, they only capture and control a subset of the reflectance field. Recently, portrait editing has been demonstrated by operating in the generative model space of StyleGAN. While such approaches do not require direct supervision, there is a significant loss of quality when compared to the supervised approaches. In this paper, we present a method which learns from limited supervised training data. The training images only include people in a fixed neutral expression with eyes closed, without much hair or background variations. Each person is captured under 150 one-light-at-a-time conditions and under 8 camera poses. Instead of training directly in the image space, we design a supervised problem which learns transformations in the latent space of StyleGAN. This combines the best of supervised learning and generative adversarial modeling. We show that the StyleGAN prior allows for generalisation to different expressions, hairstyles and backgrounds. This produces high-quality photorealistic results for in-the-wild images and significantly outperforms existing methods. Our approach can edit the illumination and pose simultaneously, and runs at interactive rates.
[ { "created": "Sat, 13 Mar 2021 08:59:49 GMT", "version": "v1" }, { "created": "Thu, 13 May 2021 17:59:43 GMT", "version": "v2" } ]
2021-05-14
[ [ "R", "Mallikarjun B", "" ], [ "Tewari", "Ayush", "" ], [ "Dib", "Abdallah", "" ], [ "Weyrich", "Tim", "" ], [ "Bickel", "Bernd", "" ], [ "Seidel", "Hans-Peter", "" ], [ "Pfister", "Hanspeter", "" ], [ "Matusik", "Wojciech", "" ], [ "Chevallier", "Louis", "" ], [ "Elgharib", "Mohamed", "" ], [ "Theobalt", "Christian", "" ] ]
Photorealistic editing of portraits is a challenging task as humans are very sensitive to inconsistencies in faces. We present an approach for high-quality intuitive editing of the camera viewpoint and scene illumination in a portrait image. This requires our method to capture and control the full reflectance field of the person in the image. Most editing approaches rely on supervised learning using training data captured with setups such as light and camera stages. Such datasets are expensive to acquire, not readily available and do not capture all the rich variations of in-the-wild portrait images. In addition, most supervised approaches only focus on relighting, and do not allow camera viewpoint editing. Thus, they only capture and control a subset of the reflectance field. Recently, portrait editing has been demonstrated by operating in the generative model space of StyleGAN. While such approaches do not require direct supervision, there is a significant loss of quality when compared to the supervised approaches. In this paper, we present a method which learns from limited supervised training data. The training images only include people in a fixed neutral expression with eyes closed, without much hair or background variations. Each person is captured under 150 one-light-at-a-time conditions and under 8 camera poses. Instead of training directly in the image space, we design a supervised problem which learns transformations in the latent space of StyleGAN. This combines the best of supervised learning and generative adversarial modeling. We show that the StyleGAN prior allows for generalisation to different expressions, hairstyles and backgrounds. This produces high-quality photorealistic results for in-the-wild images and significantly outperforms existing methods. Our approach can edit the illumination and pose simultaneously, and runs at interactive rates.
2305.15805
Sotiris Anagnostidis
Sotiris Anagnostidis, Dario Pavllo, Luca Biggio, Lorenzo Noci, Aurelien Lucchi, Thomas Hofmann
Dynamic Context Pruning for Efficient and Interpretable Autoregressive Transformers
null
null
null
null
cs.CL cs.LG
http://creativecommons.org/licenses/by/4.0/
Autoregressive Transformers adopted in Large Language Models (LLMs) are hard to scale to long sequences. Despite several works trying to reduce their computational cost, most of LLMs still adopt attention layers between all pairs of tokens in the sequence, thus incurring a quadratic cost. In this study, we present a novel approach that dynamically prunes contextual information while preserving the model's expressiveness, resulting in reduced memory and computational requirements during inference. Our method employs a learnable mechanism that determines which uninformative tokens can be dropped from the context at any point across the generation process. By doing so, our approach not only addresses performance concerns but also enhances interpretability, providing valuable insight into the model's decision-making process. Our technique can be applied to existing pre-trained models through a straightforward fine-tuning process, and the pruning strength can be specified by a sparsity parameter. Notably, our empirical findings demonstrate that we can effectively prune up to 80\% of the context without significant performance degradation on downstream tasks, offering a valuable tool for mitigating inference costs. Our reference implementation achieves up to $2\times$ increase in inference throughput and even greater memory savings.
[ { "created": "Thu, 25 May 2023 07:39:41 GMT", "version": "v1" }, { "created": "Sun, 28 May 2023 12:11:11 GMT", "version": "v2" }, { "created": "Fri, 31 May 2024 14:02:24 GMT", "version": "v3" } ]
2024-06-03
[ [ "Anagnostidis", "Sotiris", "" ], [ "Pavllo", "Dario", "" ], [ "Biggio", "Luca", "" ], [ "Noci", "Lorenzo", "" ], [ "Lucchi", "Aurelien", "" ], [ "Hofmann", "Thomas", "" ] ]
Autoregressive Transformers adopted in Large Language Models (LLMs) are hard to scale to long sequences. Despite several works trying to reduce their computational cost, most of LLMs still adopt attention layers between all pairs of tokens in the sequence, thus incurring a quadratic cost. In this study, we present a novel approach that dynamically prunes contextual information while preserving the model's expressiveness, resulting in reduced memory and computational requirements during inference. Our method employs a learnable mechanism that determines which uninformative tokens can be dropped from the context at any point across the generation process. By doing so, our approach not only addresses performance concerns but also enhances interpretability, providing valuable insight into the model's decision-making process. Our technique can be applied to existing pre-trained models through a straightforward fine-tuning process, and the pruning strength can be specified by a sparsity parameter. Notably, our empirical findings demonstrate that we can effectively prune up to 80\% of the context without significant performance degradation on downstream tasks, offering a valuable tool for mitigating inference costs. Our reference implementation achieves up to $2\times$ increase in inference throughput and even greater memory savings.
0908.0980
R Doomun
Syed S. Rizvi, Khaled M. Elleithy, Aasia Riasat
Deterministic Formulization of SNR for Wireless Multiuser DS-CDMA Networks
9 pages IEEE format, International Journal of Computer Science and Information Security, IJCSIS July 2009, ISSN 1947 5500, Impact Factor 0.423
International Journal of Computer Science and Information Security, IJCSIS, Vol. 3, No. 1, July 2009, USA
null
null
cs.NI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Wireless Multiuser receivers suffer from their relatively higher computational complexity that prevents widespread use of this technique. In addition, one of the main characteristics of multi-channel communications that can severely degrade the performance is the inconsistent and low values of SNR that result in high BER and poor channel capacity. It has been shown that the computational complexity of a multiuser receiver can be reduced by using the transformation matrix (TM) algorithm [4]. In this paper, we provide quantification of SNR based on the computational complexity of TM algorithm. We show that the reduction of complexity results high and consistent values of SNR that can consequently be used to achieve a desirable BER performance. In addition, our simulation results suggest that the high and consistent values of SNR can be achieved for a desirable BER performance. The performance measure adopted in this paper is the consistent values of SNR.
[ { "created": "Sun, 9 Aug 2009 06:57:39 GMT", "version": "v1" } ]
2009-08-10
[ [ "Rizvi", "Syed S.", "" ], [ "Elleithy", "Khaled M.", "" ], [ "Riasat", "Aasia", "" ] ]
Wireless Multiuser receivers suffer from their relatively higher computational complexity that prevents widespread use of this technique. In addition, one of the main characteristics of multi-channel communications that can severely degrade the performance is the inconsistent and low values of SNR that result in high BER and poor channel capacity. It has been shown that the computational complexity of a multiuser receiver can be reduced by using the transformation matrix (TM) algorithm [4]. In this paper, we provide quantification of SNR based on the computational complexity of TM algorithm. We show that the reduction of complexity results high and consistent values of SNR that can consequently be used to achieve a desirable BER performance. In addition, our simulation results suggest that the high and consistent values of SNR can be achieved for a desirable BER performance. The performance measure adopted in this paper is the consistent values of SNR.
2210.16083
JunKyu Lee
JunKyu Lee, Blesson Varghese, Hans Vandierendonck
ROMA: Run-Time Object Detection To Maximize Real-Time Accuracy
Accepted at the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) 2023
null
10.1109/WACV56688.2023.00634
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This paper analyzes the effects of dynamically varying video contents and detection latency on the real-time detection accuracy of a detector and proposes a new run-time accuracy variation model, ROMA, based on the findings from the analysis. ROMA is designed to select an optimal detector out of a set of detectors in real time without label information to maximize real-time object detection accuracy. ROMA utilizing four YOLOv4 detectors on an NVIDIA Jetson Nano shows real-time accuracy improvements by 4 to 37% for a scenario of dynamically varying video contents and detection latency consisting of MOT17Det and MOT20Det datasets, compared to individual YOLOv4 detectors and two state-of-the-art runtime techniques.
[ { "created": "Fri, 28 Oct 2022 12:06:29 GMT", "version": "v1" } ]
2024-04-30
[ [ "Lee", "JunKyu", "" ], [ "Varghese", "Blesson", "" ], [ "Vandierendonck", "Hans", "" ] ]
This paper analyzes the effects of dynamically varying video contents and detection latency on the real-time detection accuracy of a detector and proposes a new run-time accuracy variation model, ROMA, based on the findings from the analysis. ROMA is designed to select an optimal detector out of a set of detectors in real time without label information to maximize real-time object detection accuracy. ROMA utilizing four YOLOv4 detectors on an NVIDIA Jetson Nano shows real-time accuracy improvements by 4 to 37% for a scenario of dynamically varying video contents and detection latency consisting of MOT17Det and MOT20Det datasets, compared to individual YOLOv4 detectors and two state-of-the-art runtime techniques.
2406.15111
T\'eo Guichoux
Teo Guichoux, Laure Soulier, Nicolas Obin, Catherine Pelachaud
Investigating the impact of 2D gesture representation on co-speech gesture generation
8 pages. Paper accepted at WACAI 2024
null
null
null
cs.AI cs.CL cs.CV
http://creativecommons.org/publicdomain/zero/1.0/
Co-speech gestures play a crucial role in the interactions between humans and embodied conversational agents (ECA). Recent deep learning methods enable the generation of realistic, natural co-speech gestures synchronized with speech, but such approaches require large amounts of training data. "In-the-wild" datasets, which compile videos from sources such as YouTube through human pose detection models, offer a solution by providing 2D skeleton sequences that are paired with speech. Concurrently, innovative lifting models have emerged, capable of transforming these 2D pose sequences into their 3D counterparts, leading to large and diverse datasets of 3D gestures. However, the derived 3D pose estimation is essentially a pseudo-ground truth, with the actual ground truth being the 2D motion data. This distinction raises questions about the impact of gesture representation dimensionality on the quality of generated motions, a topic that, to our knowledge, remains largely unexplored. In this work, we evaluate the impact of the dimensionality of the training data, 2D or 3D joint coordinates, on the performance of a multimodal speech-to-gesture deep generative model. We use a lifting model to convert 2D-generated sequences of body pose to 3D. Then, we compare the sequence of gestures generated directly in 3D to the gestures generated in 2D and lifted to 3D as post-processing.
[ { "created": "Fri, 21 Jun 2024 12:59:20 GMT", "version": "v1" }, { "created": "Mon, 24 Jun 2024 08:19:00 GMT", "version": "v2" } ]
2024-06-25
[ [ "Guichoux", "Teo", "" ], [ "Soulier", "Laure", "" ], [ "Obin", "Nicolas", "" ], [ "Pelachaud", "Catherine", "" ] ]
Co-speech gestures play a crucial role in the interactions between humans and embodied conversational agents (ECA). Recent deep learning methods enable the generation of realistic, natural co-speech gestures synchronized with speech, but such approaches require large amounts of training data. "In-the-wild" datasets, which compile videos from sources such as YouTube through human pose detection models, offer a solution by providing 2D skeleton sequences that are paired with speech. Concurrently, innovative lifting models have emerged, capable of transforming these 2D pose sequences into their 3D counterparts, leading to large and diverse datasets of 3D gestures. However, the derived 3D pose estimation is essentially a pseudo-ground truth, with the actual ground truth being the 2D motion data. This distinction raises questions about the impact of gesture representation dimensionality on the quality of generated motions, a topic that, to our knowledge, remains largely unexplored. In this work, we evaluate the impact of the dimensionality of the training data, 2D or 3D joint coordinates, on the performance of a multimodal speech-to-gesture deep generative model. We use a lifting model to convert 2D-generated sequences of body pose to 3D. Then, we compare the sequence of gestures generated directly in 3D to the gestures generated in 2D and lifted to 3D as post-processing.
2107.14110
Juan C. P\'erez
Juan C. P\'erez, Motasem Alfarra, Guillaume Jeanneret, Laura Rueda, Ali Thabet, Bernard Ghanem, Pablo Arbel\'aez
Enhancing Adversarial Robustness via Test-time Transformation Ensembling
null
null
null
null
cs.LG cs.CR cs.CV
http://creativecommons.org/licenses/by/4.0/
Deep learning models are prone to being fooled by imperceptible perturbations known as adversarial attacks. In this work, we study how equipping models with Test-time Transformation Ensembling (TTE) can work as a reliable defense against such attacks. While transforming the input data, both at train and test times, is known to enhance model performance, its effects on adversarial robustness have not been studied. Here, we present a comprehensive empirical study of the impact of TTE, in the form of widely-used image transforms, on adversarial robustness. We show that TTE consistently improves model robustness against a variety of powerful attacks without any need for re-training, and that this improvement comes at virtually no trade-off with accuracy on clean samples. Finally, we show that the benefits of TTE transfer even to the certified robustness domain, in which TTE provides sizable and consistent improvements.
[ { "created": "Thu, 29 Jul 2021 15:32:35 GMT", "version": "v1" } ]
2021-07-30
[ [ "Pérez", "Juan C.", "" ], [ "Alfarra", "Motasem", "" ], [ "Jeanneret", "Guillaume", "" ], [ "Rueda", "Laura", "" ], [ "Thabet", "Ali", "" ], [ "Ghanem", "Bernard", "" ], [ "Arbeláez", "Pablo", "" ] ]
Deep learning models are prone to being fooled by imperceptible perturbations known as adversarial attacks. In this work, we study how equipping models with Test-time Transformation Ensembling (TTE) can work as a reliable defense against such attacks. While transforming the input data, both at train and test times, is known to enhance model performance, its effects on adversarial robustness have not been studied. Here, we present a comprehensive empirical study of the impact of TTE, in the form of widely-used image transforms, on adversarial robustness. We show that TTE consistently improves model robustness against a variety of powerful attacks without any need for re-training, and that this improvement comes at virtually no trade-off with accuracy on clean samples. Finally, we show that the benefits of TTE transfer even to the certified robustness domain, in which TTE provides sizable and consistent improvements.
2405.15640
Sungwoo Oh
Sungwoo Oh and Donggyu Kim
GECKO: Generative Language Model for English, Code and Korean
null
null
null
null
cs.CL cs.AI
http://creativecommons.org/licenses/by/4.0/
We introduce GECKO, a bilingual large language model (LLM) optimized for Korean and English, along with programming languages. GECKO is pretrained on the balanced, high-quality corpus of Korean and English employing LLaMA architecture. In this report, we share the experiences of several efforts to build a better data pipeline for the corpus and to train our model. GECKO shows great efficiency in token generations for both Korean and English, despite its small size of vocabulary. We measure the performance on the representative benchmarks in terms of Korean, English and Code, and it exhibits great performance on KMMLU (Korean MMLU) and modest performance in English and Code, even with its smaller number of trained tokens compared to English-focused LLMs. GECKO is available to the open-source community under a permissive license. We hope our work offers a research baseline and practical insights for Korean LLM research. The model can be found at: https://huggingface.co/kifai/GECKO-7B
[ { "created": "Fri, 24 May 2024 15:30:41 GMT", "version": "v1" } ]
2024-05-27
[ [ "Oh", "Sungwoo", "" ], [ "Kim", "Donggyu", "" ] ]
We introduce GECKO, a bilingual large language model (LLM) optimized for Korean and English, along with programming languages. GECKO is pretrained on the balanced, high-quality corpus of Korean and English employing LLaMA architecture. In this report, we share the experiences of several efforts to build a better data pipeline for the corpus and to train our model. GECKO shows great efficiency in token generations for both Korean and English, despite its small size of vocabulary. We measure the performance on the representative benchmarks in terms of Korean, English and Code, and it exhibits great performance on KMMLU (Korean MMLU) and modest performance in English and Code, even with its smaller number of trained tokens compared to English-focused LLMs. GECKO is available to the open-source community under a permissive license. We hope our work offers a research baseline and practical insights for Korean LLM research. The model can be found at: https://huggingface.co/kifai/GECKO-7B
2312.17641
Pan Liao
Yang Feng, Liao Pan, Wu Di, Liu Bo, Zhang Xingle
Motion State: A New Benchmark Multiple Object Tracking
null
null
null
null
cs.CV
http://creativecommons.org/licenses/by/4.0/
In the realm of video analysis, the field of multiple object tracking (MOT) assumes paramount importance, with the motion state of objects-whether static or dynamic relative to the ground-holding practical significance across diverse scenarios. However, the extant literature exhibits a notable dearth in the exploration of this aspect. Deep learning methodologies encounter challenges in accurately discerning object motion states, while conventional approaches reliant on comprehensive mathematical modeling may yield suboptimal tracking accuracy. To address these challenges, we introduce a Model-Data-Driven Motion State Judgment Object Tracking Method (MoD2T). This innovative architecture adeptly amalgamates traditional mathematical modeling with deep learning-based multi-object tracking frameworks. The integration of mathematical modeling and deep learning within MoD2T enhances the precision of object motion state determination, thereby elevating tracking accuracy. Our empirical investigations comprehensively validate the efficacy of MoD2T across varied scenarios, encompassing unmanned aerial vehicle surveillance and street-level tracking. Furthermore, to gauge the method's adeptness in discerning object motion states, we introduce the Motion State Validation F1 (MVF1) metric. This novel performance metric aims to quantitatively assess the accuracy of motion state classification, furnishing a comprehensive evaluation of MoD2T's performance. Elaborate experimental validations corroborate the rationality of MVF1. In order to holistically appraise MoD2T's performance, we meticulously annotate several renowned datasets and subject MoD2T to stringent testing. Remarkably, under conditions characterized by minimal or moderate camera motion, the achieved MVF1 values are particularly noteworthy, with exemplars including 0.774 for the KITTI dataset, 0.521 for MOT17, and 0.827 for UAVDT.
[ { "created": "Fri, 29 Dec 2023 15:08:06 GMT", "version": "v1" }, { "created": "Tue, 7 May 2024 13:42:52 GMT", "version": "v2" } ]
2024-05-08
[ [ "Feng", "Yang", "" ], [ "Pan", "Liao", "" ], [ "Di", "Wu", "" ], [ "Bo", "Liu", "" ], [ "Xingle", "Zhang", "" ] ]
In the realm of video analysis, the field of multiple object tracking (MOT) assumes paramount importance, with the motion state of objects-whether static or dynamic relative to the ground-holding practical significance across diverse scenarios. However, the extant literature exhibits a notable dearth in the exploration of this aspect. Deep learning methodologies encounter challenges in accurately discerning object motion states, while conventional approaches reliant on comprehensive mathematical modeling may yield suboptimal tracking accuracy. To address these challenges, we introduce a Model-Data-Driven Motion State Judgment Object Tracking Method (MoD2T). This innovative architecture adeptly amalgamates traditional mathematical modeling with deep learning-based multi-object tracking frameworks. The integration of mathematical modeling and deep learning within MoD2T enhances the precision of object motion state determination, thereby elevating tracking accuracy. Our empirical investigations comprehensively validate the efficacy of MoD2T across varied scenarios, encompassing unmanned aerial vehicle surveillance and street-level tracking. Furthermore, to gauge the method's adeptness in discerning object motion states, we introduce the Motion State Validation F1 (MVF1) metric. This novel performance metric aims to quantitatively assess the accuracy of motion state classification, furnishing a comprehensive evaluation of MoD2T's performance. Elaborate experimental validations corroborate the rationality of MVF1. In order to holistically appraise MoD2T's performance, we meticulously annotate several renowned datasets and subject MoD2T to stringent testing. Remarkably, under conditions characterized by minimal or moderate camera motion, the achieved MVF1 values are particularly noteworthy, with exemplars including 0.774 for the KITTI dataset, 0.521 for MOT17, and 0.827 for UAVDT.
2208.01352
Milad Ganjalizadeh
Milad Ganjalizadeh, Hossein S. Ghadikolaei, Johan Haraldson, Marina Petrova
Interplay between Distributed AI Workflow and URLLC
Accepted in 2022 IEEE Global Communications Conference (GLOBECOM)
null
null
null
cs.NI eess.SP
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Distributed artificial intelligence (AI) has recently accomplished tremendous breakthroughs in various communication services, ranging from fault-tolerant factory automation to smart cities. When distributed learning is run over a set of wireless connected devices, random channel fluctuations, and the incumbent services simultaneously running on the same network affect the performance of distributed learning. In this paper, we investigate the interplay between distributed AI workflow and ultra-reliable low latency communication (URLLC) services running concurrently over a network. Using 3GPP compliant simulations in a factory automation use case, we show the impact of various distributed AI settings (e.g., model size and the number of participating devices) on the convergence time of distributed AI and the application layer performance of URLLC. Unless we leverage the existing 5G-NR quality of service handling mechanisms to separate the traffic from the two services, our simulation results show that the impact of distributed AI on the availability of the URLLC devices is significant. Moreover, with proper setting of distributed AI (e.g., proper user selection), we can substantially reduce network resource utilization, leading to lower latency for distributed AI and higher availability for the URLLC users. Our results provide important insights for future 6G and AI standardization.
[ { "created": "Tue, 2 Aug 2022 10:46:50 GMT", "version": "v1" } ]
2022-08-03
[ [ "Ganjalizadeh", "Milad", "" ], [ "Ghadikolaei", "Hossein S.", "" ], [ "Haraldson", "Johan", "" ], [ "Petrova", "Marina", "" ] ]
Distributed artificial intelligence (AI) has recently accomplished tremendous breakthroughs in various communication services, ranging from fault-tolerant factory automation to smart cities. When distributed learning is run over a set of wireless connected devices, random channel fluctuations, and the incumbent services simultaneously running on the same network affect the performance of distributed learning. In this paper, we investigate the interplay between distributed AI workflow and ultra-reliable low latency communication (URLLC) services running concurrently over a network. Using 3GPP compliant simulations in a factory automation use case, we show the impact of various distributed AI settings (e.g., model size and the number of participating devices) on the convergence time of distributed AI and the application layer performance of URLLC. Unless we leverage the existing 5G-NR quality of service handling mechanisms to separate the traffic from the two services, our simulation results show that the impact of distributed AI on the availability of the URLLC devices is significant. Moreover, with proper setting of distributed AI (e.g., proper user selection), we can substantially reduce network resource utilization, leading to lower latency for distributed AI and higher availability for the URLLC users. Our results provide important insights for future 6G and AI standardization.
2108.09858
Martin Baigorria Alonso
Mart\'in Baigorria Alonso
Data Augmentation Using Many-To-Many RNNs for Session-Aware Recommender Systems
null
Proceedings of the ACM WSDM Workshop on Web Tourism (WSDM Webtour 2021)
null
null
cs.IR cs.LG
http://creativecommons.org/licenses/by/4.0/
The ACM WSDM WebTour 2021 Challenge organized by Booking.com focuses on applying Session-Aware recommender systems in the travel domain. Given a sequence of travel bookings in a user trip, we look to recommend the user's next destination. To handle the large dimensionality of the output's space, we propose a many-to-many RNN model, predicting the next destination chosen by the user at every sequence step as opposed to only the final one. We show how this is a computationally efficient alternative to doing data augmentation in a many-to-one RNN, where we consider every subsequence of a session starting from the first element. Our solution achieved 4th place in the final leaderboard, with an accuracy@4 of 0.5566.
[ { "created": "Sun, 22 Aug 2021 22:12:25 GMT", "version": "v1" } ]
2021-08-27
[ [ "Alonso", "Martín Baigorria", "" ] ]
The ACM WSDM WebTour 2021 Challenge organized by Booking.com focuses on applying Session-Aware recommender systems in the travel domain. Given a sequence of travel bookings in a user trip, we look to recommend the user's next destination. To handle the large dimensionality of the output's space, we propose a many-to-many RNN model, predicting the next destination chosen by the user at every sequence step as opposed to only the final one. We show how this is a computationally efficient alternative to doing data augmentation in a many-to-one RNN, where we consider every subsequence of a session starting from the first element. Our solution achieved 4th place in the final leaderboard, with an accuracy@4 of 0.5566.
2307.07240
Bin-Cheng Yang
Bincheng Yang and Gangshan Wu
MaxSR: Image Super-Resolution Using Improved MaxViT
null
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
While transformer models have been demonstrated to be effective for natural language processing tasks and high-level vision tasks, only a few attempts have been made to use powerful transformer models for single image super-resolution. Because transformer models have powerful representation capacity and the in-built self-attention mechanisms in transformer models help to leverage self-similarity prior in input low-resolution image to improve performance for single image super-resolution, we present a single image super-resolution model based on recent hybrid vision transformer of MaxViT, named as MaxSR. MaxSR consists of four parts, a shallow feature extraction block, multiple cascaded adaptive MaxViT blocks to extract deep hierarchical features and model global self-similarity from low-level features efficiently, a hierarchical feature fusion block, and finally a reconstruction block. The key component of MaxSR, i.e., adaptive MaxViT block, is based on MaxViT block which mixes MBConv with squeeze-and-excitation, block attention and grid attention. In order to achieve better global modelling of self-similarity in input low-resolution image, we improve block attention and grid attention in MaxViT block to adaptive block attention and adaptive grid attention which do self-attention inside each window across all grids and each grid across all windows respectively in the most efficient way. We instantiate proposed model for classical single image super-resolution (MaxSR) and lightweight single image super-resolution (MaxSR-light). Experiments show that our MaxSR and MaxSR-light establish new state-of-the-art performance efficiently.
[ { "created": "Fri, 14 Jul 2023 09:26:47 GMT", "version": "v1" } ]
2023-07-17
[ [ "Yang", "Bincheng", "" ], [ "Wu", "Gangshan", "" ] ]
While transformer models have been demonstrated to be effective for natural language processing tasks and high-level vision tasks, only a few attempts have been made to use powerful transformer models for single image super-resolution. Because transformer models have powerful representation capacity and the in-built self-attention mechanisms in transformer models help to leverage self-similarity prior in input low-resolution image to improve performance for single image super-resolution, we present a single image super-resolution model based on recent hybrid vision transformer of MaxViT, named as MaxSR. MaxSR consists of four parts, a shallow feature extraction block, multiple cascaded adaptive MaxViT blocks to extract deep hierarchical features and model global self-similarity from low-level features efficiently, a hierarchical feature fusion block, and finally a reconstruction block. The key component of MaxSR, i.e., adaptive MaxViT block, is based on MaxViT block which mixes MBConv with squeeze-and-excitation, block attention and grid attention. In order to achieve better global modelling of self-similarity in input low-resolution image, we improve block attention and grid attention in MaxViT block to adaptive block attention and adaptive grid attention which do self-attention inside each window across all grids and each grid across all windows respectively in the most efficient way. We instantiate proposed model for classical single image super-resolution (MaxSR) and lightweight single image super-resolution (MaxSR-light). Experiments show that our MaxSR and MaxSR-light establish new state-of-the-art performance efficiently.
0811.4170
Alain Barrat
Alain Barrat, Ciro Cattuto, Vittoria Colizza, Jean-Francois Pinton, Wouter Van den Broeck, Alessandro Vespignani
High resolution dynamical mapping of social interactions with active RFID
null
PLoS ONE 5(7): e11596 (2010)
10.1371/journal.pone.0011596
null
cs.CY cs.HC physics.soc-ph
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In this paper we present an experimental framework to gather data on face-to-face social interactions between individuals, with a high spatial and temporal resolution. We use active Radio Frequency Identification (RFID) devices that assess contacts with one another by exchanging low-power radio packets. When individuals wear the beacons as a badge, a persistent radio contact between the RFID devices can be used as a proxy for a social interaction between individuals. We present the results of a pilot study recently performed during a conference, and a subsequent preliminary data analysis, that provides an assessment of our method and highlights its versatility and applicability in many areas concerned with human dynamics.
[ { "created": "Tue, 25 Nov 2008 20:54:34 GMT", "version": "v1" }, { "created": "Tue, 25 Nov 2008 21:01:28 GMT", "version": "v2" } ]
2010-08-18
[ [ "Barrat", "Alain", "" ], [ "Cattuto", "Ciro", "" ], [ "Colizza", "Vittoria", "" ], [ "Pinton", "Jean-Francois", "" ], [ "Broeck", "Wouter Van den", "" ], [ "Vespignani", "Alessandro", "" ] ]
In this paper we present an experimental framework to gather data on face-to-face social interactions between individuals, with a high spatial and temporal resolution. We use active Radio Frequency Identification (RFID) devices that assess contacts with one another by exchanging low-power radio packets. When individuals wear the beacons as a badge, a persistent radio contact between the RFID devices can be used as a proxy for a social interaction between individuals. We present the results of a pilot study recently performed during a conference, and a subsequent preliminary data analysis, that provides an assessment of our method and highlights its versatility and applicability in many areas concerned with human dynamics.
2009.04426
Felipe Del Rio
Pablo Messina, Manuel Cartagena, Patricio Cerda-Mardini, Felipe del Rio and Denis Parra
CuratorNet: Visually-aware Recommendation of Art Images
null
null
null
null
cs.IR cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Although there are several visually-aware recommendation models in domains like fashion or even movies, the art domain lacks thesame level of research attention, despite the recent growth of the online artwork market. To reduce this gap, in this article we introduceCuratorNet, a neural network architecture for visually-aware recommendation of art images. CuratorNet is designed at the core withthe goal of maximizing generalization: the network has a fixed set of parameters that only need to be trained once, and thereafter themodel is able to generalize to new users or items never seen before, without further training. This is achieved by leveraging visualcontent: items are mapped to item vectors through visual embeddings, and users are mapped to user vectors by aggregating the visualcontent of items they have consumed. Besides the model architecture, we also introduce novel triplet sampling strategies to build atraining set for rank learning in the art domain, resulting in more effective learning than naive random sampling. With an evaluationover a real-world dataset of physical paintings, we show that CuratorNet achieves the best performance among several baselines,including the state-of-the-art model VBPR. CuratorNet is motivated and evaluated in the art domain, but its architecture and trainingscheme could be adapted to recommend images in other areas
[ { "created": "Wed, 9 Sep 2020 17:22:17 GMT", "version": "v1" }, { "created": "Wed, 30 Sep 2020 12:35:08 GMT", "version": "v2" } ]
2020-10-01
[ [ "Messina", "Pablo", "" ], [ "Cartagena", "Manuel", "" ], [ "Cerda-Mardini", "Patricio", "" ], [ "del Rio", "Felipe", "" ], [ "Parra", "Denis", "" ] ]
Although there are several visually-aware recommendation models in domains like fashion or even movies, the art domain lacks thesame level of research attention, despite the recent growth of the online artwork market. To reduce this gap, in this article we introduceCuratorNet, a neural network architecture for visually-aware recommendation of art images. CuratorNet is designed at the core withthe goal of maximizing generalization: the network has a fixed set of parameters that only need to be trained once, and thereafter themodel is able to generalize to new users or items never seen before, without further training. This is achieved by leveraging visualcontent: items are mapped to item vectors through visual embeddings, and users are mapped to user vectors by aggregating the visualcontent of items they have consumed. Besides the model architecture, we also introduce novel triplet sampling strategies to build atraining set for rank learning in the art domain, resulting in more effective learning than naive random sampling. With an evaluationover a real-world dataset of physical paintings, we show that CuratorNet achieves the best performance among several baselines,including the state-of-the-art model VBPR. CuratorNet is motivated and evaluated in the art domain, but its architecture and trainingscheme could be adapted to recommend images in other areas
2407.05458
Fei Wang
Fei Wang, Weibo Gao, Qi Liu, Jiatong Li, Guanhao Zhao, Zheng Zhang, Zhenya Huang, Mengxiao Zhu, Shijin Wang, Wei Tong, Enhong Chen
A Survey of Models for Cognitive Diagnosis: New Developments and Future Directions
null
null
null
null
cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Cognitive diagnosis has been developed for decades as an effective measurement tool to evaluate human cognitive status such as ability level and knowledge mastery. It has been applied to a wide range of fields including education, sport, psychological diagnosis, etc. By providing better awareness of cognitive status, it can serve as the basis for personalized services such as well-designed medical treatment, teaching strategy and vocational training. This paper aims to provide a survey of current models for cognitive diagnosis, with more attention on new developments using machine learning-based methods. By comparing the model structures, parameter estimation algorithms, model evaluation methods and applications, we provide a relatively comprehensive review of the recent trends in cognitive diagnosis models. Further, we discuss future directions that are worthy of exploration. In addition, we release two Python libraries: EduData for easy access to some relevant public datasets we have collected, and EduCDM that implements popular CDMs to facilitate both applications and research purposes.
[ { "created": "Sun, 7 Jul 2024 18:02:00 GMT", "version": "v1" } ]
2024-07-09
[ [ "Wang", "Fei", "" ], [ "Gao", "Weibo", "" ], [ "Liu", "Qi", "" ], [ "Li", "Jiatong", "" ], [ "Zhao", "Guanhao", "" ], [ "Zhang", "Zheng", "" ], [ "Huang", "Zhenya", "" ], [ "Zhu", "Mengxiao", "" ], [ "Wang", "Shijin", "" ], [ "Tong", "Wei", "" ], [ "Chen", "Enhong", "" ] ]
Cognitive diagnosis has been developed for decades as an effective measurement tool to evaluate human cognitive status such as ability level and knowledge mastery. It has been applied to a wide range of fields including education, sport, psychological diagnosis, etc. By providing better awareness of cognitive status, it can serve as the basis for personalized services such as well-designed medical treatment, teaching strategy and vocational training. This paper aims to provide a survey of current models for cognitive diagnosis, with more attention on new developments using machine learning-based methods. By comparing the model structures, parameter estimation algorithms, model evaluation methods and applications, we provide a relatively comprehensive review of the recent trends in cognitive diagnosis models. Further, we discuss future directions that are worthy of exploration. In addition, we release two Python libraries: EduData for easy access to some relevant public datasets we have collected, and EduCDM that implements popular CDMs to facilitate both applications and research purposes.
2312.07685
Yinmin Zhang
Yinmin Zhang, Jie Liu, Chuming Li, Yazhe Niu, Yaodong Yang, Yu Liu, Wanli Ouyang
A Perspective of Q-value Estimation on Offline-to-Online Reinforcement Learning
Accepted at AAAI 2024
null
null
null
cs.LG cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Offline-to-online Reinforcement Learning (O2O RL) aims to improve the performance of offline pretrained policy using only a few online samples. Built on offline RL algorithms, most O2O methods focus on the balance between RL objective and pessimism, or the utilization of offline and online samples. In this paper, from a novel perspective, we systematically study the challenges that remain in O2O RL and identify that the reason behind the slow improvement of the performance and the instability of online finetuning lies in the inaccurate Q-value estimation inherited from offline pretraining. Specifically, we demonstrate that the estimation bias and the inaccurate rank of Q-value cause a misleading signal for the policy update, making the standard offline RL algorithms, such as CQL and TD3-BC, ineffective in the online finetuning. Based on this observation, we address the problem of Q-value estimation by two techniques: (1) perturbed value update and (2) increased frequency of Q-value updates. The first technique smooths out biased Q-value estimation with sharp peaks, preventing early-stage policy exploitation of sub-optimal actions. The second one alleviates the estimation bias inherited from offline pretraining by accelerating learning. Extensive experiments on the MuJoco and Adroit environments demonstrate that the proposed method, named SO2, significantly alleviates Q-value estimation issues, and consistently improves the performance against the state-of-the-art methods by up to 83.1%.
[ { "created": "Tue, 12 Dec 2023 19:24:35 GMT", "version": "v1" } ]
2023-12-14
[ [ "Zhang", "Yinmin", "" ], [ "Liu", "Jie", "" ], [ "Li", "Chuming", "" ], [ "Niu", "Yazhe", "" ], [ "Yang", "Yaodong", "" ], [ "Liu", "Yu", "" ], [ "Ouyang", "Wanli", "" ] ]
Offline-to-online Reinforcement Learning (O2O RL) aims to improve the performance of offline pretrained policy using only a few online samples. Built on offline RL algorithms, most O2O methods focus on the balance between RL objective and pessimism, or the utilization of offline and online samples. In this paper, from a novel perspective, we systematically study the challenges that remain in O2O RL and identify that the reason behind the slow improvement of the performance and the instability of online finetuning lies in the inaccurate Q-value estimation inherited from offline pretraining. Specifically, we demonstrate that the estimation bias and the inaccurate rank of Q-value cause a misleading signal for the policy update, making the standard offline RL algorithms, such as CQL and TD3-BC, ineffective in the online finetuning. Based on this observation, we address the problem of Q-value estimation by two techniques: (1) perturbed value update and (2) increased frequency of Q-value updates. The first technique smooths out biased Q-value estimation with sharp peaks, preventing early-stage policy exploitation of sub-optimal actions. The second one alleviates the estimation bias inherited from offline pretraining by accelerating learning. Extensive experiments on the MuJoco and Adroit environments demonstrate that the proposed method, named SO2, significantly alleviates Q-value estimation issues, and consistently improves the performance against the state-of-the-art methods by up to 83.1%.
1909.07818
Lasse Hansen
Lasse Hansen, Doris Dittmer, Mattias P. Heinrich
Learning Deformable Point Set Registration with Regularized Dynamic Graph CNNs for Large Lung Motion in COPD Patients
accepted for MICCAI 2019 Workshop Graph Learning in Medical Imaging
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Deformable registration continues to be one of the key challenges in medical image analysis. While iconic registration methods have started to benefit from the recent advances in medical deep learning, the same does not yet apply for the registration of point sets, e.g. registration based on surfaces, keypoints or landmarks. This is mainly due to the restriction of the convolution operator in modern CNNs to densely gridded input. However, with the newly developed methods from the field of geometric deep learning suitable tools are now emerging, which enable powerful analysis of medical data on irregular domains. In this work, we present a new method that enables the learning of regularized feature descriptors with dynamic graph CNNs. By incorporating the learned geometric features as prior probabilities into the well-established coherent point drift (CPD) algorithm, formulated as differentiable network layer, we establish an end-to-end framework for robust registration of two point sets. Our approach is evaluated on the challenging task of aligning keypoints extracted from lung CT scans in inhale and exhale states with large deformations and without any additional intensity information. Our results indicate that the inherent geometric structure of the extracted keypoints is sufficient to establish descriptive point features, which yield a significantly improved performance and robustness of our registration framework.
[ { "created": "Tue, 17 Sep 2019 13:59:04 GMT", "version": "v1" } ]
2019-09-18
[ [ "Hansen", "Lasse", "" ], [ "Dittmer", "Doris", "" ], [ "Heinrich", "Mattias P.", "" ] ]
Deformable registration continues to be one of the key challenges in medical image analysis. While iconic registration methods have started to benefit from the recent advances in medical deep learning, the same does not yet apply for the registration of point sets, e.g. registration based on surfaces, keypoints or landmarks. This is mainly due to the restriction of the convolution operator in modern CNNs to densely gridded input. However, with the newly developed methods from the field of geometric deep learning suitable tools are now emerging, which enable powerful analysis of medical data on irregular domains. In this work, we present a new method that enables the learning of regularized feature descriptors with dynamic graph CNNs. By incorporating the learned geometric features as prior probabilities into the well-established coherent point drift (CPD) algorithm, formulated as differentiable network layer, we establish an end-to-end framework for robust registration of two point sets. Our approach is evaluated on the challenging task of aligning keypoints extracted from lung CT scans in inhale and exhale states with large deformations and without any additional intensity information. Our results indicate that the inherent geometric structure of the extracted keypoints is sufficient to establish descriptive point features, which yield a significantly improved performance and robustness of our registration framework.
2403.11369
KV Aditya Srivatsa
KV Aditya Srivatsa and Ekaterina Kochmar
What Makes Math Word Problems Challenging for LLMs?
Accepted to NAACL Findings 2024
null
null
null
cs.CL
http://creativecommons.org/licenses/by/4.0/
This paper investigates the question of what makes math word problems (MWPs) in English challenging for large language models (LLMs). We conduct an in-depth analysis of the key linguistic and mathematical characteristics of MWPs. In addition, we train feature-based classifiers to better understand the impact of each feature on the overall difficulty of MWPs for prominent LLMs and investigate whether this helps predict how well LLMs fare against specific categories of MWPs.
[ { "created": "Sun, 17 Mar 2024 23:18:40 GMT", "version": "v1" }, { "created": "Mon, 1 Apr 2024 13:58:34 GMT", "version": "v2" } ]
2024-04-02
[ [ "Srivatsa", "KV Aditya", "" ], [ "Kochmar", "Ekaterina", "" ] ]
This paper investigates the question of what makes math word problems (MWPs) in English challenging for large language models (LLMs). We conduct an in-depth analysis of the key linguistic and mathematical characteristics of MWPs. In addition, we train feature-based classifiers to better understand the impact of each feature on the overall difficulty of MWPs for prominent LLMs and investigate whether this helps predict how well LLMs fare against specific categories of MWPs.
2301.04727
Joaquin Garcia-Alfaro
Iain Burge, Michel Barbeau, Joaquin Garcia-Alfaro
A Quantum Algorithm for Shapley Value Estimation
29 pages, 8 figures, 21 references, baseline (preprint) QCE 2023 (IEEE International Conference on Quantum Computing and Engineering) Technical Paper (Quantum Algorithms for Shapley Value Calculation)
null
null
null
cs.ET cs.CR math.QA
http://creativecommons.org/licenses/by/4.0/
The introduction of the European Union's (EU) set of comprehensive regulations relating to technology, the General Data Protection Regulation, grants EU citizens the right to explanations for automated decisions that have significant effects on their life. This poses a substantial challenge, as many of today's state-of-the-art algorithms are generally unexplainable black boxes. Simultaneously, we have seen an emergence of the fields of quantum computation and quantum AI. Due to the fickle nature of quantum information, the problem of explainability is amplified, as measuring a quantum system destroys the information. As a result, there is a need for post-hoc explanations for quantum AI algorithms. In the classical context, the cooperative game theory concept of the Shapley value has been adapted for post-hoc explanations. However, this approach does not translate to use in quantum computing trivially and can be exponentially difficult to implement if not handled with care. We propose a novel algorithm which reduces the problem of accurately estimating the Shapley values of a quantum algorithm into a far simpler problem of estimating the true average of a binomial distribution in polynomial time.
[ { "created": "Wed, 11 Jan 2023 21:32:59 GMT", "version": "v1" }, { "created": "Wed, 15 Mar 2023 16:10:32 GMT", "version": "v2" }, { "created": "Fri, 14 Jul 2023 08:17:46 GMT", "version": "v3" } ]
2023-08-24
[ [ "Burge", "Iain", "" ], [ "Barbeau", "Michel", "" ], [ "Garcia-Alfaro", "Joaquin", "" ] ]
The introduction of the European Union's (EU) set of comprehensive regulations relating to technology, the General Data Protection Regulation, grants EU citizens the right to explanations for automated decisions that have significant effects on their life. This poses a substantial challenge, as many of today's state-of-the-art algorithms are generally unexplainable black boxes. Simultaneously, we have seen an emergence of the fields of quantum computation and quantum AI. Due to the fickle nature of quantum information, the problem of explainability is amplified, as measuring a quantum system destroys the information. As a result, there is a need for post-hoc explanations for quantum AI algorithms. In the classical context, the cooperative game theory concept of the Shapley value has been adapted for post-hoc explanations. However, this approach does not translate to use in quantum computing trivially and can be exponentially difficult to implement if not handled with care. We propose a novel algorithm which reduces the problem of accurately estimating the Shapley values of a quantum algorithm into a far simpler problem of estimating the true average of a binomial distribution in polynomial time.
2311.09394
Marco Elver
Kostya Serebryany, Chris Kennelly, Mitch Phillips, Matt Denton, Marco Elver, Alexander Potapenko, Matt Morehouse, Vlad Tsyrklevich, Christian Holler, Julian Lettner, David Kilzer, Lander Brandt
GWP-ASan: Sampling-Based Detection of Memory-Safety Bugs in Production
null
null
null
null
cs.SE cs.PL
http://creativecommons.org/licenses/by/4.0/
Despite the recent advances in pre-production bug detection, heap-use-after-free and heap-buffer-overflow bugs remain the primary problem for security, reliability, and developer productivity for applications written in C or C++, across all major software ecosystems. Memory-safe languages solve this problem when they are used, but the existing code bases consisting of billions of lines of C and C++ continue to grow, and we need additional bug detection mechanisms. This paper describes a family of tools that detect these two classes of memory-safety bugs, while running in production, at near-zero overhead. These tools combine page-granular guarded allocation and low-rate sampling. In other words, we added an "if" statement to a 36-year-old idea and made it work at scale. We describe the basic algorithm, several of its variants and implementations, and the results of multi-year deployments across mobile, desktop, and server applications.
[ { "created": "Wed, 15 Nov 2023 21:41:53 GMT", "version": "v1" }, { "created": "Sat, 13 Jan 2024 14:42:26 GMT", "version": "v2" } ]
2024-01-17
[ [ "Serebryany", "Kostya", "" ], [ "Kennelly", "Chris", "" ], [ "Phillips", "Mitch", "" ], [ "Denton", "Matt", "" ], [ "Elver", "Marco", "" ], [ "Potapenko", "Alexander", "" ], [ "Morehouse", "Matt", "" ], [ "Tsyrklevich", "Vlad", "" ], [ "Holler", "Christian", "" ], [ "Lettner", "Julian", "" ], [ "Kilzer", "David", "" ], [ "Brandt", "Lander", "" ] ]
Despite the recent advances in pre-production bug detection, heap-use-after-free and heap-buffer-overflow bugs remain the primary problem for security, reliability, and developer productivity for applications written in C or C++, across all major software ecosystems. Memory-safe languages solve this problem when they are used, but the existing code bases consisting of billions of lines of C and C++ continue to grow, and we need additional bug detection mechanisms. This paper describes a family of tools that detect these two classes of memory-safety bugs, while running in production, at near-zero overhead. These tools combine page-granular guarded allocation and low-rate sampling. In other words, we added an "if" statement to a 36-year-old idea and made it work at scale. We describe the basic algorithm, several of its variants and implementations, and the results of multi-year deployments across mobile, desktop, and server applications.
2103.05469
Mark Stamp
Andy Phung and Mark Stamp
Universal Adversarial Perturbations and Image Spam Classifiers
null
null
null
null
cs.CR cs.CV cs.LG
http://creativecommons.org/licenses/by/4.0/
As the name suggests, image spam is spam email that has been embedded in an image. Image spam was developed in an effort to evade text-based filters. Modern deep learning-based classifiers perform well in detecting typical image spam that is seen in the wild. In this chapter, we evaluate numerous adversarial techniques for the purpose of attacking deep learning-based image spam classifiers. Of the techniques tested, we find that universal perturbation performs best. Using universal adversarial perturbations, we propose and analyze a new transformation-based adversarial attack that enables us to create tailored "natural perturbations" in image spam. The resulting spam images benefit from both the presence of concentrated natural features and a universal adversarial perturbation. We show that the proposed technique outperforms existing adversarial attacks in terms of accuracy reduction, computation time per example, and perturbation distance. We apply our technique to create a dataset of adversarial spam images, which can serve as a challenge dataset for future research in image spam detection.
[ { "created": "Sun, 7 Mar 2021 14:36:02 GMT", "version": "v1" } ]
2021-03-10
[ [ "Phung", "Andy", "" ], [ "Stamp", "Mark", "" ] ]
As the name suggests, image spam is spam email that has been embedded in an image. Image spam was developed in an effort to evade text-based filters. Modern deep learning-based classifiers perform well in detecting typical image spam that is seen in the wild. In this chapter, we evaluate numerous adversarial techniques for the purpose of attacking deep learning-based image spam classifiers. Of the techniques tested, we find that universal perturbation performs best. Using universal adversarial perturbations, we propose and analyze a new transformation-based adversarial attack that enables us to create tailored "natural perturbations" in image spam. The resulting spam images benefit from both the presence of concentrated natural features and a universal adversarial perturbation. We show that the proposed technique outperforms existing adversarial attacks in terms of accuracy reduction, computation time per example, and perturbation distance. We apply our technique to create a dataset of adversarial spam images, which can serve as a challenge dataset for future research in image spam detection.
1606.02409
Zeng Yulong
Pingzhong Tang and Yulong Zeng
How to manipulate truthful prior-dependent mechanisms?
29 pages, 1 figure
null
null
null
cs.GT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In the standard formulation of mechanism design, a key assumption is that the designer has reliable information and technology to determine a prior distribution on types of the agents. In the meanwhile, as pointed out by the Wilson's Principle, a mechanism should reply as little as possible on the accuracy of prior type distribution. In this paper, we put forward a model to formalize and quantify this statement. In our model, each agent has a type distribution. In addition, the agent can commit to a fake distribution and bids consistently and credibly with respect to the fake distribution (i.e., plays Bayes equilibrium under the fake distributions). We study the equilibria of the induced distribution-committing games in several well-known mechanisms. Our results can be summarized as follows: (1) the game induced by Myerson's auction under our model is strategically equivalent to the first price auction under the standard model. As a consequence, they are revenue-equivalent as well. (2) the second-price auction yields weakly better revenue than several reserve-based and virtual-value-based auctions, under our fake distribution model. These results echo the recent literature on prior-independent mechanism design.
[ { "created": "Wed, 8 Jun 2016 05:59:14 GMT", "version": "v1" }, { "created": "Tue, 19 Jul 2016 05:18:21 GMT", "version": "v2" } ]
2016-07-20
[ [ "Tang", "Pingzhong", "" ], [ "Zeng", "Yulong", "" ] ]
In the standard formulation of mechanism design, a key assumption is that the designer has reliable information and technology to determine a prior distribution on types of the agents. In the meanwhile, as pointed out by the Wilson's Principle, a mechanism should reply as little as possible on the accuracy of prior type distribution. In this paper, we put forward a model to formalize and quantify this statement. In our model, each agent has a type distribution. In addition, the agent can commit to a fake distribution and bids consistently and credibly with respect to the fake distribution (i.e., plays Bayes equilibrium under the fake distributions). We study the equilibria of the induced distribution-committing games in several well-known mechanisms. Our results can be summarized as follows: (1) the game induced by Myerson's auction under our model is strategically equivalent to the first price auction under the standard model. As a consequence, they are revenue-equivalent as well. (2) the second-price auction yields weakly better revenue than several reserve-based and virtual-value-based auctions, under our fake distribution model. These results echo the recent literature on prior-independent mechanism design.
1904.06903
Xiangyu Xu
Xiangyu Xu, Muchen Li, Wenxiu Sun
Learning Deformable Kernels for Image and Video Denoising
10 pages
null
null
null
cs.CV cs.AI cs.GR
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Most of the classical denoising methods restore clear results by selecting and averaging pixels in the noisy input. Instead of relying on hand-crafted selecting and averaging strategies, we propose to explicitly learn this process with deep neural networks. Specifically, we propose deformable 2D kernels for image denoising where the sampling locations and kernel weights are both learned. The proposed kernel naturally adapts to image structures and could effectively reduce the oversmoothing artifacts. Furthermore, we develop 3D deformable kernels for video denoising to more efficiently sample pixels across the spatial-temporal space. Our method is able to solve the misalignment issues of large motion from dynamic scenes. For better training our video denoising model, we introduce the trilinear sampler and a new regularization term. We demonstrate that the proposed method performs favorably against the state-of-the-art image and video denoising approaches on both synthetic and real-world data.
[ { "created": "Mon, 15 Apr 2019 08:15:09 GMT", "version": "v1" } ]
2019-04-16
[ [ "Xu", "Xiangyu", "" ], [ "Li", "Muchen", "" ], [ "Sun", "Wenxiu", "" ] ]
Most of the classical denoising methods restore clear results by selecting and averaging pixels in the noisy input. Instead of relying on hand-crafted selecting and averaging strategies, we propose to explicitly learn this process with deep neural networks. Specifically, we propose deformable 2D kernels for image denoising where the sampling locations and kernel weights are both learned. The proposed kernel naturally adapts to image structures and could effectively reduce the oversmoothing artifacts. Furthermore, we develop 3D deformable kernels for video denoising to more efficiently sample pixels across the spatial-temporal space. Our method is able to solve the misalignment issues of large motion from dynamic scenes. For better training our video denoising model, we introduce the trilinear sampler and a new regularization term. We demonstrate that the proposed method performs favorably against the state-of-the-art image and video denoising approaches on both synthetic and real-world data.
1610.04872
Bodhisattwa Majumder
Bodhisattwa Prasad Majumder, Ayan Sengupta, Sajal jain, Parikshit Bhaduri
Fault Detection Engine in Intelligent Predictive Analytics Platform for DCIM
Accepted in 4th International Conference on Business Analytics and Intelligence (ICBAI 2016)
null
null
null
cs.AI cs.DC
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
With the advancement of huge data generation and data handling capability, Machine Learning and Probabilistic modelling enables an immense opportunity to employ predictive analytics platform in high security critical industries namely data centers, electricity grids, utilities, airport etc. where downtime minimization is one of the primary objectives. This paper proposes a novel, complete architecture of an intelligent predictive analytics platform, Fault Engine, for huge device network connected with electrical/information flow. Three unique modules, here proposed, seamlessly integrate with available technology stack of data handling and connect with middleware to produce online intelligent prediction in critical failure scenarios. The Markov Failure module predicts the severity of a failure along with survival probability of a device at any given instances. The Root Cause Analysis model indicates probable devices as potential root cause employing Bayesian probability assignment and topological sort. Finally, a community detection algorithm produces correlated clusters of device in terms of failure probability which will further narrow down the search space of finding route cause. The whole Engine has been tested with different size of network with simulated failure environments and shows its potential to be scalable in real-time implementation.
[ { "created": "Sun, 16 Oct 2016 15:14:36 GMT", "version": "v1" } ]
2016-10-18
[ [ "Majumder", "Bodhisattwa Prasad", "" ], [ "Sengupta", "Ayan", "" ], [ "jain", "Sajal", "" ], [ "Bhaduri", "Parikshit", "" ] ]
With the advancement of huge data generation and data handling capability, Machine Learning and Probabilistic modelling enables an immense opportunity to employ predictive analytics platform in high security critical industries namely data centers, electricity grids, utilities, airport etc. where downtime minimization is one of the primary objectives. This paper proposes a novel, complete architecture of an intelligent predictive analytics platform, Fault Engine, for huge device network connected with electrical/information flow. Three unique modules, here proposed, seamlessly integrate with available technology stack of data handling and connect with middleware to produce online intelligent prediction in critical failure scenarios. The Markov Failure module predicts the severity of a failure along with survival probability of a device at any given instances. The Root Cause Analysis model indicates probable devices as potential root cause employing Bayesian probability assignment and topological sort. Finally, a community detection algorithm produces correlated clusters of device in terms of failure probability which will further narrow down the search space of finding route cause. The whole Engine has been tested with different size of network with simulated failure environments and shows its potential to be scalable in real-time implementation.
2106.11196
Benedikt Boenninghoff
Benedikt Boenninghoff, Dorothea Kolossa, Robert M. Nickel
Self-Calibrating Neural-Probabilistic Model for Authorship Verification Under Covariate Shift
12th International Conference of the CLEF Association, 2021
null
null
null
cs.CL
http://creativecommons.org/licenses/by/4.0/
We are addressing two fundamental problems in authorship verification (AV): Topic variability and miscalibration. Variations in the topic of two disputed texts are a major cause of error for most AV systems. In addition, it is observed that the underlying probability estimates produced by deep learning AV mechanisms oftentimes do not match the actual case counts in the respective training data. As such, probability estimates are poorly calibrated. We are expanding our framework from PAN 2020 to include Bayes factor scoring (BFS) and an uncertainty adaptation layer (UAL) to address both problems. Experiments with the 2020/21 PAN AV shared task data show that the proposed method significantly reduces sensitivities to topical variations and significantly improves the system's calibration.
[ { "created": "Mon, 21 Jun 2021 15:33:48 GMT", "version": "v1" } ]
2021-06-22
[ [ "Boenninghoff", "Benedikt", "" ], [ "Kolossa", "Dorothea", "" ], [ "Nickel", "Robert M.", "" ] ]
We are addressing two fundamental problems in authorship verification (AV): Topic variability and miscalibration. Variations in the topic of two disputed texts are a major cause of error for most AV systems. In addition, it is observed that the underlying probability estimates produced by deep learning AV mechanisms oftentimes do not match the actual case counts in the respective training data. As such, probability estimates are poorly calibrated. We are expanding our framework from PAN 2020 to include Bayes factor scoring (BFS) and an uncertainty adaptation layer (UAL) to address both problems. Experiments with the 2020/21 PAN AV shared task data show that the proposed method significantly reduces sensitivities to topical variations and significantly improves the system's calibration.
2107.12429
Pan Ji
Pan Ji, Runze Li, Bir Bhanu, Yi Xu
MonoIndoor: Towards Good Practice of Self-Supervised Monocular Depth Estimation for Indoor Environments
ICCV 2021
null
null
null
cs.CV
http://creativecommons.org/licenses/by-nc-sa/4.0/
Self-supervised depth estimation for indoor environments is more challenging than its outdoor counterpart in at least the following two aspects: (i) the depth range of indoor sequences varies a lot across different frames, making it difficult for the depth network to induce consistent depth cues, whereas the maximum distance in outdoor scenes mostly stays the same as the camera usually sees the sky; (ii) the indoor sequences contain much more rotational motions, which cause difficulties for the pose network, while the motions of outdoor sequences are pre-dominantly translational, especially for driving datasets such as KITTI. In this paper, special considerations are given to those challenges and a set of good practices are consolidated for improving the performance of self-supervised monocular depth estimation in indoor environments. The proposed method mainly consists of two novel modules, \ie, a depth factorization module and a residual pose estimation module, each of which is designed to respectively tackle the aforementioned challenges. The effectiveness of each module is shown through a carefully conducted ablation study and the demonstration of the state-of-the-art performance on three indoor datasets, \ie, EuRoC, NYUv2, and 7-scenes.
[ { "created": "Mon, 26 Jul 2021 18:45:14 GMT", "version": "v1" }, { "created": "Wed, 28 Jul 2021 00:32:57 GMT", "version": "v2" } ]
2021-07-29
[ [ "Ji", "Pan", "" ], [ "Li", "Runze", "" ], [ "Bhanu", "Bir", "" ], [ "Xu", "Yi", "" ] ]
Self-supervised depth estimation for indoor environments is more challenging than its outdoor counterpart in at least the following two aspects: (i) the depth range of indoor sequences varies a lot across different frames, making it difficult for the depth network to induce consistent depth cues, whereas the maximum distance in outdoor scenes mostly stays the same as the camera usually sees the sky; (ii) the indoor sequences contain much more rotational motions, which cause difficulties for the pose network, while the motions of outdoor sequences are pre-dominantly translational, especially for driving datasets such as KITTI. In this paper, special considerations are given to those challenges and a set of good practices are consolidated for improving the performance of self-supervised monocular depth estimation in indoor environments. The proposed method mainly consists of two novel modules, \ie, a depth factorization module and a residual pose estimation module, each of which is designed to respectively tackle the aforementioned challenges. The effectiveness of each module is shown through a carefully conducted ablation study and the demonstration of the state-of-the-art performance on three indoor datasets, \ie, EuRoC, NYUv2, and 7-scenes.
2208.11904
Gayan Kulatilleke
Gayan K. Kulatilleke, Sugandika Samarakoon
Empirical study of Machine Learning Classifier Evaluation Metrics behavior in Massively Imbalanced and Noisy data
null
null
null
null
cs.LG cs.CR
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
With growing credit card transaction volumes, the fraud percentages are also rising, including overhead costs for institutions to combat and compensate victims. The use of machine learning into the financial sector permits more effective protection against fraud and other economic crime. Suitably trained machine learning classifiers help proactive fraud detection, improving stakeholder trust and robustness against illicit transactions. However, the design of machine learning based fraud detection algorithms has been challenging and slow due the massively unbalanced nature of fraud data and the challenges of identifying the frauds accurately and completely to create a gold standard ground truth. Furthermore, there are no benchmarks or standard classifier evaluation metrics to measure and identify better performing classifiers, thus keeping researchers in the dark. In this work, we develop a theoretical foundation to model human annotation errors and extreme imbalance typical in real world fraud detection data sets. By conducting empirical experiments on a hypothetical classifier, with a synthetic data distribution approximated to a popular real world credit card fraud data set, we simulate human annotation errors and extreme imbalance to observe the behavior of popular machine learning classifier evaluation matrices. We demonstrate that a combined F1 score and g-mean, in that specific order, is the best evaluation metric for typical imbalanced fraud detection model classification.
[ { "created": "Thu, 25 Aug 2022 07:30:31 GMT", "version": "v1" } ]
2022-08-26
[ [ "Kulatilleke", "Gayan K.", "" ], [ "Samarakoon", "Sugandika", "" ] ]
With growing credit card transaction volumes, the fraud percentages are also rising, including overhead costs for institutions to combat and compensate victims. The use of machine learning into the financial sector permits more effective protection against fraud and other economic crime. Suitably trained machine learning classifiers help proactive fraud detection, improving stakeholder trust and robustness against illicit transactions. However, the design of machine learning based fraud detection algorithms has been challenging and slow due the massively unbalanced nature of fraud data and the challenges of identifying the frauds accurately and completely to create a gold standard ground truth. Furthermore, there are no benchmarks or standard classifier evaluation metrics to measure and identify better performing classifiers, thus keeping researchers in the dark. In this work, we develop a theoretical foundation to model human annotation errors and extreme imbalance typical in real world fraud detection data sets. By conducting empirical experiments on a hypothetical classifier, with a synthetic data distribution approximated to a popular real world credit card fraud data set, we simulate human annotation errors and extreme imbalance to observe the behavior of popular machine learning classifier evaluation matrices. We demonstrate that a combined F1 score and g-mean, in that specific order, is the best evaluation metric for typical imbalanced fraud detection model classification.
1807.11164
Xiangyu Zhang
Ningning Ma, Xiangyu Zhang, Hai-Tao Zheng, Jian Sun
ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design
null
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Currently, the neural network architecture design is mostly guided by the \emph{indirect} metric of computation complexity, i.e., FLOPs. However, the \emph{direct} metric, e.g., speed, also depends on the other factors such as memory access cost and platform characterics. Thus, this work proposes to evaluate the direct metric on the target platform, beyond only considering FLOPs. Based on a series of controlled experiments, this work derives several practical \emph{guidelines} for efficient network design. Accordingly, a new architecture is presented, called \emph{ShuffleNet V2}. Comprehensive ablation experiments verify that our model is the state-of-the-art in terms of speed and accuracy tradeoff.
[ { "created": "Mon, 30 Jul 2018 04:18:25 GMT", "version": "v1" } ]
2018-07-31
[ [ "Ma", "Ningning", "" ], [ "Zhang", "Xiangyu", "" ], [ "Zheng", "Hai-Tao", "" ], [ "Sun", "Jian", "" ] ]
Currently, the neural network architecture design is mostly guided by the \emph{indirect} metric of computation complexity, i.e., FLOPs. However, the \emph{direct} metric, e.g., speed, also depends on the other factors such as memory access cost and platform characterics. Thus, this work proposes to evaluate the direct metric on the target platform, beyond only considering FLOPs. Based on a series of controlled experiments, this work derives several practical \emph{guidelines} for efficient network design. Accordingly, a new architecture is presented, called \emph{ShuffleNet V2}. Comprehensive ablation experiments verify that our model is the state-of-the-art in terms of speed and accuracy tradeoff.
2105.10011
Leonard Berrada
Leonard Berrada, Andrew Zisserman, M. Pawan Kumar
Comment on Stochastic Polyak Step-Size: Performance of ALI-G
null
null
null
null
cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This is a short note on the performance of the ALI-G algorithm (Berrada et al., 2020) as reported in (Loizou et al., 2021). ALI-G (Berrada et al., 2020) and SPS (Loizou et al., 2021) are both adaptations of the Polyak step-size to optimize machine learning models that can interpolate the training data. The main algorithmic differences are that (1) SPS employs a multiplicative constant in the denominator of the learning-rate while ALI-G uses an additive constant, and (2) SPS uses an iteration-dependent maximal learning-rate while ALI-G uses a constant one. There are also differences in the analysis provided by the two works, with less restrictive assumptions proposed in (Loizou et al., 2021). In their experiments, (Loizou et al., 2021) did not use momentum for ALI-G (which is a standard part of the algorithm) or standard hyper-parameter tuning (for e.g. learning-rate and regularization). Hence this note as a reference for the improved performance that ALI-G can obtain with well-chosen hyper-parameters. In particular, we show that when training a ResNet-34 on CIFAR-10 and CIFAR-100, the performance of ALI-G can reach respectively 93.5% (+6%) and 76% (+8%) with a very small amount of tuning. Thus ALI-G remains a very competitive method for training interpolating neural networks.
[ { "created": "Thu, 20 May 2021 19:57:34 GMT", "version": "v1" } ]
2021-05-24
[ [ "Berrada", "Leonard", "" ], [ "Zisserman", "Andrew", "" ], [ "Kumar", "M. Pawan", "" ] ]
This is a short note on the performance of the ALI-G algorithm (Berrada et al., 2020) as reported in (Loizou et al., 2021). ALI-G (Berrada et al., 2020) and SPS (Loizou et al., 2021) are both adaptations of the Polyak step-size to optimize machine learning models that can interpolate the training data. The main algorithmic differences are that (1) SPS employs a multiplicative constant in the denominator of the learning-rate while ALI-G uses an additive constant, and (2) SPS uses an iteration-dependent maximal learning-rate while ALI-G uses a constant one. There are also differences in the analysis provided by the two works, with less restrictive assumptions proposed in (Loizou et al., 2021). In their experiments, (Loizou et al., 2021) did not use momentum for ALI-G (which is a standard part of the algorithm) or standard hyper-parameter tuning (for e.g. learning-rate and regularization). Hence this note as a reference for the improved performance that ALI-G can obtain with well-chosen hyper-parameters. In particular, we show that when training a ResNet-34 on CIFAR-10 and CIFAR-100, the performance of ALI-G can reach respectively 93.5% (+6%) and 76% (+8%) with a very small amount of tuning. Thus ALI-G remains a very competitive method for training interpolating neural networks.
cs/0208012
Jim Gray
Jim Gray, Alexander S. Szalay, Ani R. Thakar, Christopher Stoughton, Jan vandenBerg
Online Scientific Data Curation, Publication, and Archiving
original at http://research.microsoft.com/scripts/pubs/view.asp?TR_ID=MSR-TR-2002-74
null
10.1117/12.461524
MSR-TR-2002-74
cs.DL
null
Science projects are data publishers. The scale and complexity of current and future science data changes the nature of the publication process. Publication is becoming a major project component. At a minimum, a project must preserve the ephemeral data it gathers. Derived data can be reconstructed from metadata, but metadata is ephemeral. Longer term, a project should expect some archive to preserve the data. We observe that pub-lished scientific data needs to be available forever ? this gives rise to the data pyramid of versions and to data inflation where the derived data volumes explode. As an example, this article describes the Sloan Digital Sky Survey (SDSS) strategies for data publication, data access, curation, and preservation.
[ { "created": "Wed, 7 Aug 2002 22:42:31 GMT", "version": "v1" } ]
2015-06-25
[ [ "Gray", "Jim", "" ], [ "Szalay", "Alexander S.", "" ], [ "Thakar", "Ani R.", "" ], [ "Stoughton", "Christopher", "" ], [ "vandenBerg", "Jan", "" ] ]
Science projects are data publishers. The scale and complexity of current and future science data changes the nature of the publication process. Publication is becoming a major project component. At a minimum, a project must preserve the ephemeral data it gathers. Derived data can be reconstructed from metadata, but metadata is ephemeral. Longer term, a project should expect some archive to preserve the data. We observe that pub-lished scientific data needs to be available forever ? this gives rise to the data pyramid of versions and to data inflation where the derived data volumes explode. As an example, this article describes the Sloan Digital Sky Survey (SDSS) strategies for data publication, data access, curation, and preservation.
2207.04606
Zihao Ye
Zihao Ye, Ruihang Lai, Junru Shao, Tianqi Chen, Luis Ceze
SparseTIR: Composable Abstractions for Sparse Compilation in Deep Learning
To appear at ASPLOS 2023 (19 pages, 23 figures), source code available at https://github.com/uwsampl/sparsetir, artifact available at https://github.com/uwsampl/sparsetir-artifact
null
null
null
cs.LG cs.AI cs.PL
http://creativecommons.org/licenses/by-nc-sa/4.0/
Sparse tensors are rapidly becoming critical components of modern deep learning workloads. However, developing high-performance sparse operators can be difficult and tedious, and existing vendor libraries cannot satisfy the escalating demands from new operators. Sparse tensor compilers simplify the development of operators, but efficient sparse compilation for deep learning remains challenging because a single sparse format cannot maximize hardware efficiency, and single-shot compilers cannot keep up with latest hardware and system advances. In this paper, we observe that the key to addressing both these challenges is to leverage composable formats and composable transformations. We propose SparseTIR, a sparse tensor compilation abstraction that offers composable formats and composable transformations for deep learning workloads. SparseTIR constructs a search space over these composable components for performance tuning. With these improvements, SparseTIR obtains consistent performance speedups vs vendor libraries on GPUs for single operators: 1.20-2.34x for GNN operators, 1.05-2.98x for sparse attention operators, and 0.56-7.45x for sparse convolution operators. SparseTIR also accelerates end-to-end GNNs by 1.08-1.52x for GraphSAGE training, and 4.20-40.18x for RGCN inference.
[ { "created": "Mon, 11 Jul 2022 03:49:53 GMT", "version": "v1" }, { "created": "Fri, 26 Aug 2022 03:57:10 GMT", "version": "v2" }, { "created": "Fri, 11 Nov 2022 01:48:13 GMT", "version": "v3" }, { "created": "Tue, 21 Feb 2023 16:51:55 GMT", "version": "v4" } ]
2023-02-22
[ [ "Ye", "Zihao", "" ], [ "Lai", "Ruihang", "" ], [ "Shao", "Junru", "" ], [ "Chen", "Tianqi", "" ], [ "Ceze", "Luis", "" ] ]
Sparse tensors are rapidly becoming critical components of modern deep learning workloads. However, developing high-performance sparse operators can be difficult and tedious, and existing vendor libraries cannot satisfy the escalating demands from new operators. Sparse tensor compilers simplify the development of operators, but efficient sparse compilation for deep learning remains challenging because a single sparse format cannot maximize hardware efficiency, and single-shot compilers cannot keep up with latest hardware and system advances. In this paper, we observe that the key to addressing both these challenges is to leverage composable formats and composable transformations. We propose SparseTIR, a sparse tensor compilation abstraction that offers composable formats and composable transformations for deep learning workloads. SparseTIR constructs a search space over these composable components for performance tuning. With these improvements, SparseTIR obtains consistent performance speedups vs vendor libraries on GPUs for single operators: 1.20-2.34x for GNN operators, 1.05-2.98x for sparse attention operators, and 0.56-7.45x for sparse convolution operators. SparseTIR also accelerates end-to-end GNNs by 1.08-1.52x for GraphSAGE training, and 4.20-40.18x for RGCN inference.
1602.00248
Adam Kucharski
Adam J. Kucharski
Modelling the transmission dynamics of online social contagion
13 pages, 6 figures, 2 tables
null
null
null
cs.SI physics.soc-ph
http://creativecommons.org/licenses/by/4.0/
During 2014-15, there were several outbreaks of nominated-based online social contagion. These infections, which were transmitted from one individual to another via posts on social media, included games such as 'neknomination', 'ice bucket challenge', 'no make up selfies', and Facebook users re-posting their first profile pictures. Fitting a mathematical model of infectious disease transmission to outbreaks of these four games in the United Kingdom, I estimated the basic reproduction number, $R_0$, and generation time of each infection. Median estimates for $R_0$ ranged from 1.9-2.5 across the four outbreaks, and the estimated generation times were between 1.0 and 2.0 days. Tests using out-of-sample data from Australia suggested that the model had reasonable predictive power, with $R^2$ values between 0.52-0.70 across the four Australian datasets. Further, the relatively low basic reproduction numbers for the infections suggests that only 48-60% of index cases in nomination-based games may subsequently generate major outbreaks.
[ { "created": "Sun, 31 Jan 2016 13:58:17 GMT", "version": "v1" } ]
2016-02-02
[ [ "Kucharski", "Adam J.", "" ] ]
During 2014-15, there were several outbreaks of nominated-based online social contagion. These infections, which were transmitted from one individual to another via posts on social media, included games such as 'neknomination', 'ice bucket challenge', 'no make up selfies', and Facebook users re-posting their first profile pictures. Fitting a mathematical model of infectious disease transmission to outbreaks of these four games in the United Kingdom, I estimated the basic reproduction number, $R_0$, and generation time of each infection. Median estimates for $R_0$ ranged from 1.9-2.5 across the four outbreaks, and the estimated generation times were between 1.0 and 2.0 days. Tests using out-of-sample data from Australia suggested that the model had reasonable predictive power, with $R^2$ values between 0.52-0.70 across the four Australian datasets. Further, the relatively low basic reproduction numbers for the infections suggests that only 48-60% of index cases in nomination-based games may subsequently generate major outbreaks.
2403.00278
Jinho Bok
Jinho Bok, Weijie Su, Jason M. Altschuler
Shifted Interpolation for Differential Privacy
45 pages, ICML 2024. v2: added lower bounds (Appendix C.5)
null
null
null
cs.LG cs.CR math.OC math.ST stat.ML stat.TH
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Noisy gradient descent and its variants are the predominant algorithms for differentially private machine learning. It is a fundamental question to quantify their privacy leakage, yet tight characterizations remain open even in the foundational setting of convex losses. This paper improves over previous analyses by establishing (and refining) the "privacy amplification by iteration" phenomenon in the unifying framework of $f$-differential privacy--which tightly captures all aspects of the privacy loss and immediately implies tighter privacy accounting in other notions of differential privacy, e.g., $(\varepsilon,\delta)$-DP and R\'enyi DP. Our key technical insight is the construction of shifted interpolated processes that unravel the popular shifted-divergences argument, enabling generalizations beyond divergence-based relaxations of DP. Notably, this leads to the first exact privacy analysis in the foundational setting of strongly convex optimization. Our techniques extend to many settings: convex/strongly convex, constrained/unconstrained, full/cyclic/stochastic batches, and all combinations thereof. As an immediate corollary, we recover the $f$-DP characterization of the exponential mechanism for strongly convex optimization in Gopi et al. (2022), and moreover extend this result to more general settings.
[ { "created": "Fri, 1 Mar 2024 04:50:04 GMT", "version": "v1" }, { "created": "Wed, 12 Jun 2024 04:08:27 GMT", "version": "v2" } ]
2024-06-13
[ [ "Bok", "Jinho", "" ], [ "Su", "Weijie", "" ], [ "Altschuler", "Jason M.", "" ] ]
Noisy gradient descent and its variants are the predominant algorithms for differentially private machine learning. It is a fundamental question to quantify their privacy leakage, yet tight characterizations remain open even in the foundational setting of convex losses. This paper improves over previous analyses by establishing (and refining) the "privacy amplification by iteration" phenomenon in the unifying framework of $f$-differential privacy--which tightly captures all aspects of the privacy loss and immediately implies tighter privacy accounting in other notions of differential privacy, e.g., $(\varepsilon,\delta)$-DP and R\'enyi DP. Our key technical insight is the construction of shifted interpolated processes that unravel the popular shifted-divergences argument, enabling generalizations beyond divergence-based relaxations of DP. Notably, this leads to the first exact privacy analysis in the foundational setting of strongly convex optimization. Our techniques extend to many settings: convex/strongly convex, constrained/unconstrained, full/cyclic/stochastic batches, and all combinations thereof. As an immediate corollary, we recover the $f$-DP characterization of the exponential mechanism for strongly convex optimization in Gopi et al. (2022), and moreover extend this result to more general settings.
2207.03618
Shannan Guan
Shannan Guan, Haiyan Lu, Linchao Zhu, Gengfa Fang
PoseGU: 3D Human Pose Estimation with Novel Human Pose Generator and Unbiased Learning
null
null
null
null
cs.CV cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
3D pose estimation has recently gained substantial interests in computer vision domain. Existing 3D pose estimation methods have a strong reliance on large size well-annotated 3D pose datasets, and they suffer poor model generalization on unseen poses due to limited diversity of 3D poses in training sets. In this work, we propose PoseGU, a novel human pose generator that generates diverse poses with access only to a small size of seed samples, while equipping the Counterfactual Risk Minimization to pursue an unbiased evaluation objective. Extensive experiments demonstrate PoseGU outforms almost all the state-of-the-art 3D human pose methods under consideration over three popular benchmark datasets. Empirical analysis also proves PoseGU generates 3D poses with improved data diversity and better generalization ability.
[ { "created": "Thu, 7 Jul 2022 23:43:53 GMT", "version": "v1" } ]
2022-07-11
[ [ "Guan", "Shannan", "" ], [ "Lu", "Haiyan", "" ], [ "Zhu", "Linchao", "" ], [ "Fang", "Gengfa", "" ] ]
3D pose estimation has recently gained substantial interests in computer vision domain. Existing 3D pose estimation methods have a strong reliance on large size well-annotated 3D pose datasets, and they suffer poor model generalization on unseen poses due to limited diversity of 3D poses in training sets. In this work, we propose PoseGU, a novel human pose generator that generates diverse poses with access only to a small size of seed samples, while equipping the Counterfactual Risk Minimization to pursue an unbiased evaluation objective. Extensive experiments demonstrate PoseGU outforms almost all the state-of-the-art 3D human pose methods under consideration over three popular benchmark datasets. Empirical analysis also proves PoseGU generates 3D poses with improved data diversity and better generalization ability.
2104.07516
Chengtang Yao
Chengtang Yao, Yunde Jia, Huijun Di, Pengxiang Li, Yuwei Wu
A Decomposition Model for Stereo Matching
CVPR 2021
null
null
null
cs.CV
http://creativecommons.org/licenses/by/4.0/
In this paper, we present a decomposition model for stereo matching to solve the problem of excessive growth in computational cost (time and memory cost) as the resolution increases. In order to reduce the huge cost of stereo matching at the original resolution, our model only runs dense matching at a very low resolution and uses sparse matching at different higher resolutions to recover the disparity of lost details scale-by-scale. After the decomposition of stereo matching, our model iteratively fuses the sparse and dense disparity maps from adjacent scales with an occlusion-aware mask. A refinement network is also applied to improving the fusion result. Compared with high-performance methods like PSMNet and GANet, our method achieves $10-100\times$ speed increase while obtaining comparable disparity estimation results.
[ { "created": "Thu, 15 Apr 2021 15:16:23 GMT", "version": "v1" } ]
2021-04-16
[ [ "Yao", "Chengtang", "" ], [ "Jia", "Yunde", "" ], [ "Di", "Huijun", "" ], [ "Li", "Pengxiang", "" ], [ "Wu", "Yuwei", "" ] ]
In this paper, we present a decomposition model for stereo matching to solve the problem of excessive growth in computational cost (time and memory cost) as the resolution increases. In order to reduce the huge cost of stereo matching at the original resolution, our model only runs dense matching at a very low resolution and uses sparse matching at different higher resolutions to recover the disparity of lost details scale-by-scale. After the decomposition of stereo matching, our model iteratively fuses the sparse and dense disparity maps from adjacent scales with an occlusion-aware mask. A refinement network is also applied to improving the fusion result. Compared with high-performance methods like PSMNet and GANet, our method achieves $10-100\times$ speed increase while obtaining comparable disparity estimation results.
2208.01250
Chaozhuo Li
Yiding Zhang, Chaozhuo Li, Senzhang Wang, Jianxun Lian, Xing Xie
Geometric Interaction Augmented Graph Collaborative Filtering
null
null
null
null
cs.IR
http://creativecommons.org/licenses/by/4.0/
Graph-based collaborative filtering is capable of capturing the essential and abundant collaborative signals from the high-order interactions, and thus received increasingly research interests. Conventionally, the embeddings of users and items are defined in the Euclidean spaces, along with the propagation on the interaction graphs. Meanwhile, recent works point out that the high-order interactions naturally form up the tree-likeness structures, which the hyperbolic models thrive on. However, the interaction graphs inherently exhibit the hybrid and nested geometric characteristics, while the existing single geometry-based models are inadequate to fully capture such sophisticated topological patterns. In this paper, we propose to model the user-item interactions in a hybrid geometric space, in which the merits of Euclidean and hyperbolic spaces are simultaneously enjoyed to learn expressive representations. Experimental results on public datasets validate the effectiveness of our proposal.
[ { "created": "Tue, 2 Aug 2022 04:53:17 GMT", "version": "v1" } ]
2022-08-03
[ [ "Zhang", "Yiding", "" ], [ "Li", "Chaozhuo", "" ], [ "Wang", "Senzhang", "" ], [ "Lian", "Jianxun", "" ], [ "Xie", "Xing", "" ] ]
Graph-based collaborative filtering is capable of capturing the essential and abundant collaborative signals from the high-order interactions, and thus received increasingly research interests. Conventionally, the embeddings of users and items are defined in the Euclidean spaces, along with the propagation on the interaction graphs. Meanwhile, recent works point out that the high-order interactions naturally form up the tree-likeness structures, which the hyperbolic models thrive on. However, the interaction graphs inherently exhibit the hybrid and nested geometric characteristics, while the existing single geometry-based models are inadequate to fully capture such sophisticated topological patterns. In this paper, we propose to model the user-item interactions in a hybrid geometric space, in which the merits of Euclidean and hyperbolic spaces are simultaneously enjoyed to learn expressive representations. Experimental results on public datasets validate the effectiveness of our proposal.
2003.13320
Kai Niu
Kai Niu and Yan Li
Polar Coded Diversity on Block Fading Channels via Polar Spectrum
13 pages, 5 figues
null
10.1109/TSP.2021.3094652
null
cs.IT math.IT
http://creativecommons.org/licenses/by/4.0/
Due to the advantage of capacity-achieving, polar codes have been extended to the block fading channel whereas most constructions involve complex iterative-calculation. In this paper, we establish a systematic framework to analyze the error performance of polar codes in the case of block mapping and random mapping. For both the mappings, by introducing the new concept, named split polar spectrum, we derive the upper bound on the error probability of polarized channel which explicitly reveals the relationship between the diversity order L and the block-wise weight distribution of the codeword. For the special case L=2 in the block mapping, we design the enumeration algorithm to calculate the exact split polar spectrum based on the general MacWilliams identities. For arbitrary diversity order in the random mapping, with the help of uniform interleaving, we derive the approximate split polar spectrum by combining the polar spectrum and the probability of fading pattern for a specific weight. Furthermore, we propose the design criteria to construct polar codes over the block fading channel. The full diversity criterion is the primary target so as to achieve the diversity gain and the product distance criterion requires to maximize the product of the block-wise Hamming distance whereby obtain the coding gain. Guided by these design criteria, the construction metric, named polarized diversity weight (PDW) is proposed to design the polar codes in both mappings. Such a simple metric can construct polar codes with similar or better performance over those based on traditional methods in block fading channel.
[ { "created": "Mon, 30 Mar 2020 10:12:14 GMT", "version": "v1" } ]
2021-08-11
[ [ "Niu", "Kai", "" ], [ "Li", "Yan", "" ] ]
Due to the advantage of capacity-achieving, polar codes have been extended to the block fading channel whereas most constructions involve complex iterative-calculation. In this paper, we establish a systematic framework to analyze the error performance of polar codes in the case of block mapping and random mapping. For both the mappings, by introducing the new concept, named split polar spectrum, we derive the upper bound on the error probability of polarized channel which explicitly reveals the relationship between the diversity order L and the block-wise weight distribution of the codeword. For the special case L=2 in the block mapping, we design the enumeration algorithm to calculate the exact split polar spectrum based on the general MacWilliams identities. For arbitrary diversity order in the random mapping, with the help of uniform interleaving, we derive the approximate split polar spectrum by combining the polar spectrum and the probability of fading pattern for a specific weight. Furthermore, we propose the design criteria to construct polar codes over the block fading channel. The full diversity criterion is the primary target so as to achieve the diversity gain and the product distance criterion requires to maximize the product of the block-wise Hamming distance whereby obtain the coding gain. Guided by these design criteria, the construction metric, named polarized diversity weight (PDW) is proposed to design the polar codes in both mappings. Such a simple metric can construct polar codes with similar or better performance over those based on traditional methods in block fading channel.
2209.13803
Luo Ping
Ping Luo, Jieren Cheng, Zhenhao Liu, N.Xiong, Jie Wu
FedVeca: Federated Vectorized Averaging on Non-IID Data with Adaptive Bi-directional Global Objective
null
null
null
null
cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Federated Learning (FL) is a distributed machine learning framework to alleviate the data silos, where decentralized clients collaboratively learn a global model without sharing their private data. However, the clients' Non-Independent and Identically Distributed (Non-IID) data negatively affect the trained model, and clients with different numbers of local updates may cause significant gaps to the local gradients in each communication round. In this paper, we propose a Federated Vectorized Averaging (FedVeca) method to address the above problem on Non-IID data. Specifically, we set a novel objective for the global model which is related to the local gradients. The local gradient is defined as a bi-directional vector with step size and direction, where the step size is the number of local updates and the direction is divided into positive and negative according to our definition. In FedVeca, the direction is influenced by the step size, thus we average the bi-directional vectors to reduce the effect of different step sizes. Then, we theoretically analyze the relationship between the step sizes and the global objective, and obtain upper bounds on the step sizes per communication round. Based on the upper bounds, we design an algorithm for the server and the client to adaptively adjusts the step sizes that make the objective close to the optimum. Finally, we conduct experiments on different datasets, models and scenarios by building a prototype system, and the experimental results demonstrate the effectiveness and efficiency of the FedVeca method.
[ { "created": "Wed, 28 Sep 2022 03:14:10 GMT", "version": "v1" }, { "created": "Tue, 7 Feb 2023 11:14:06 GMT", "version": "v2" } ]
2023-02-08
[ [ "Luo", "Ping", "" ], [ "Cheng", "Jieren", "" ], [ "Liu", "Zhenhao", "" ], [ "Xiong", "N.", "" ], [ "Wu", "Jie", "" ] ]
Federated Learning (FL) is a distributed machine learning framework to alleviate the data silos, where decentralized clients collaboratively learn a global model without sharing their private data. However, the clients' Non-Independent and Identically Distributed (Non-IID) data negatively affect the trained model, and clients with different numbers of local updates may cause significant gaps to the local gradients in each communication round. In this paper, we propose a Federated Vectorized Averaging (FedVeca) method to address the above problem on Non-IID data. Specifically, we set a novel objective for the global model which is related to the local gradients. The local gradient is defined as a bi-directional vector with step size and direction, where the step size is the number of local updates and the direction is divided into positive and negative according to our definition. In FedVeca, the direction is influenced by the step size, thus we average the bi-directional vectors to reduce the effect of different step sizes. Then, we theoretically analyze the relationship between the step sizes and the global objective, and obtain upper bounds on the step sizes per communication round. Based on the upper bounds, we design an algorithm for the server and the client to adaptively adjusts the step sizes that make the objective close to the optimum. Finally, we conduct experiments on different datasets, models and scenarios by building a prototype system, and the experimental results demonstrate the effectiveness and efficiency of the FedVeca method.
2002.01358
Shan Zhang
Xiao Ma, Ao Zhou, Shan Zhang, Shangguang Wang
Cooperative Service Caching and Workload Scheduling in Mobile Edge Computing
INFOCOM 2020
null
null
null
cs.NI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Mobile edge computing is beneficial to reduce service response time and core network traffic by pushing cloud functionalities to network edge. Equipped with storage and computation capacities, edge nodes can cache services of resource-intensive and delay-sensitive mobile applications and process the corresponding computation tasks without outsourcing to central clouds. However, the heterogeneity of edge resource capacities and inconsistence of edge storage and computation capacities make it difficult to jointly fully utilize the storage and computation capacities when there is no cooperation among edge nodes. To address this issue, we consider cooperation among edge nodes and investigate cooperative service caching and workload scheduling in mobile edge computing. This problem can be formulated as a mixed integer nonlinear programming problem, which has non-polynomial computation complexity. To overcome the challenges of subproblem coupling, computation-communication tradeoff, and edge node heterogeneity, we develop an iterative algorithm called ICE. This algorithm is designed based on Gibbs sampling, which has provably near-optimal results, and the idea of water filling, which has polynomial computation complexity. Simulations are conducted and the results demonstrate that our algorithm can jointly reduce the service response time and the outsourcing traffic compared with the benchmark algorithms.
[ { "created": "Tue, 4 Feb 2020 15:06:44 GMT", "version": "v1" } ]
2020-02-05
[ [ "Ma", "Xiao", "" ], [ "Zhou", "Ao", "" ], [ "Zhang", "Shan", "" ], [ "Wang", "Shangguang", "" ] ]
Mobile edge computing is beneficial to reduce service response time and core network traffic by pushing cloud functionalities to network edge. Equipped with storage and computation capacities, edge nodes can cache services of resource-intensive and delay-sensitive mobile applications and process the corresponding computation tasks without outsourcing to central clouds. However, the heterogeneity of edge resource capacities and inconsistence of edge storage and computation capacities make it difficult to jointly fully utilize the storage and computation capacities when there is no cooperation among edge nodes. To address this issue, we consider cooperation among edge nodes and investigate cooperative service caching and workload scheduling in mobile edge computing. This problem can be formulated as a mixed integer nonlinear programming problem, which has non-polynomial computation complexity. To overcome the challenges of subproblem coupling, computation-communication tradeoff, and edge node heterogeneity, we develop an iterative algorithm called ICE. This algorithm is designed based on Gibbs sampling, which has provably near-optimal results, and the idea of water filling, which has polynomial computation complexity. Simulations are conducted and the results demonstrate that our algorithm can jointly reduce the service response time and the outsourcing traffic compared with the benchmark algorithms.
1907.12891
Olivier Rukundo
Olivier Rukundo
4X4 Census Transform
3 pages, 9 figures, 2 tables
null
null
null
cs.CV eess.IV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This paper proposes a 4X4 Census Transform (4X4CT) to encourage further research in computer vision and visual computing. Unlike the traditional 3X3 CT which uses a nine pixels kernel, the proposed 4X4CT uses a sixteen pixels kernel with four overlapped groups of 3X3 kernel size. In each overlapping group, a reference input pixel profits from its nearest eight pixels to produce an eight bits binary string convertible to a grayscale integer of the 4X4CT's output pixel. Preliminary experiments demonstrated more image textural crispness and contrast than the CT as well as alternativeness to enable meaningful solutions to be achieved.
[ { "created": "Tue, 30 Jul 2019 13:30:45 GMT", "version": "v1" } ]
2019-07-31
[ [ "Rukundo", "Olivier", "" ] ]
This paper proposes a 4X4 Census Transform (4X4CT) to encourage further research in computer vision and visual computing. Unlike the traditional 3X3 CT which uses a nine pixels kernel, the proposed 4X4CT uses a sixteen pixels kernel with four overlapped groups of 3X3 kernel size. In each overlapping group, a reference input pixel profits from its nearest eight pixels to produce an eight bits binary string convertible to a grayscale integer of the 4X4CT's output pixel. Preliminary experiments demonstrated more image textural crispness and contrast than the CT as well as alternativeness to enable meaningful solutions to be achieved.
2108.01454
Albert Weichselbraun
Albert Weichselbraun
Inscriptis -- A Python-based HTML to text conversion library optimized for knowledge extraction from the Web
Preprint of the published version, which includes all improvements made during the review process
Journal of Open Source Software (2021), 6(66), 3557
10.21105/joss.03557
null
cs.IR
http://creativecommons.org/licenses/by/4.0/
Inscriptis provides a library, command line client and Web service for converting HTML to plain text. Its development has been triggered by the need to obtain accurate text representations for knowledge extraction tasks that preserve the spatial alignment of text without drawing upon heavyweight, browser-based solutions such as Selenium. In contrast to related software packages, Inscriptis (i) provides a layout-aware conversion of HTML that more closely resembles the rendering obtained from standard Web browsers; and (ii) supports annotation rules, i.e., user-provided mappings that allow for annotating the extracted text based on structural and semantic information encoded in HTML tags and attributes. These unique features ensure that downstream knowledge extraction components can operate on accurate text representations, and may even use information on the semantics and structure of the original HTML document.
[ { "created": "Mon, 12 Jul 2021 12:40:43 GMT", "version": "v1" }, { "created": "Fri, 22 Oct 2021 07:04:02 GMT", "version": "v2" } ]
2021-10-25
[ [ "Weichselbraun", "Albert", "" ] ]
Inscriptis provides a library, command line client and Web service for converting HTML to plain text. Its development has been triggered by the need to obtain accurate text representations for knowledge extraction tasks that preserve the spatial alignment of text without drawing upon heavyweight, browser-based solutions such as Selenium. In contrast to related software packages, Inscriptis (i) provides a layout-aware conversion of HTML that more closely resembles the rendering obtained from standard Web browsers; and (ii) supports annotation rules, i.e., user-provided mappings that allow for annotating the extracted text based on structural and semantic information encoded in HTML tags and attributes. These unique features ensure that downstream knowledge extraction components can operate on accurate text representations, and may even use information on the semantics and structure of the original HTML document.
2112.04019
Hongwei Zhu
Hongwei Zhu, Minjia Shi
The b-symbol weight hierarchy of the Kasami codes
null
null
null
null
cs.IT math.GR math.IT
http://creativecommons.org/publicdomain/zero/1.0/
The symbol-pair read channel was first proposed by Cassuto and Blaum. Later, Yaakobi et al. generalized it to the $b$-symbol read channel. It is motivated by the limitations of the reading process in high density data storage systems. One main task in $b$-symbol coding theory is to determine the $b$-symbol weight hierarchy of codes. In this paper, we study the $b$-symbol weight hierarchy of the Kasami codes, which are well known for their applications to construct sequences with optimal correlation magnitudes. The complete symbol-pair weight distribution of the Kasami codes is determined.
[ { "created": "Tue, 7 Dec 2021 22:19:05 GMT", "version": "v1" } ]
2021-12-09
[ [ "Zhu", "Hongwei", "" ], [ "Shi", "Minjia", "" ] ]
The symbol-pair read channel was first proposed by Cassuto and Blaum. Later, Yaakobi et al. generalized it to the $b$-symbol read channel. It is motivated by the limitations of the reading process in high density data storage systems. One main task in $b$-symbol coding theory is to determine the $b$-symbol weight hierarchy of codes. In this paper, we study the $b$-symbol weight hierarchy of the Kasami codes, which are well known for their applications to construct sequences with optimal correlation magnitudes. The complete symbol-pair weight distribution of the Kasami codes is determined.
1009.1407
Grenville Croll
Sebastian Dewhurst
Transforming Critical Spreadsheets into Web Applications at Zurich Financial
10 pages, 6 colour figures; ISBN 978-1-905404-50-6
Proc. European Spreadsheet Risks Int. Grp. (EuSpRIG) 2010 23-32
null
null
cs.SE cs.CY
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In the insurance industry, spreadsheets have emerged as an invaluable tool to for product pricing, because it is relatively straightforward to create and maintain complex pricing models using Excel. In fact, Excel is often preferred to "hard-code" whenever there are frequent changes to the calculations and business logic which under-pin the pricing of an insurance product. However, problems arise as soon as spreadsheets are deployed to end-users: version control, security of intellectual property, and ensuring correct usage are obvious issues; frequently, integration with other systems is also a requirement. Zurich Financial Services Group is a leading financial services provider; several possible solutions to these problems have been evaluated, and EASA has been selected as the preferred technology. Other spreadsheet collaboration approaches which were considered include Excel Services, and/or custom-built software; however, EASA has provided clear benefits over these strategies.
[ { "created": "Tue, 7 Sep 2010 21:15:47 GMT", "version": "v1" } ]
2010-09-09
[ [ "Dewhurst", "Sebastian", "" ] ]
In the insurance industry, spreadsheets have emerged as an invaluable tool to for product pricing, because it is relatively straightforward to create and maintain complex pricing models using Excel. In fact, Excel is often preferred to "hard-code" whenever there are frequent changes to the calculations and business logic which under-pin the pricing of an insurance product. However, problems arise as soon as spreadsheets are deployed to end-users: version control, security of intellectual property, and ensuring correct usage are obvious issues; frequently, integration with other systems is also a requirement. Zurich Financial Services Group is a leading financial services provider; several possible solutions to these problems have been evaluated, and EASA has been selected as the preferred technology. Other spreadsheet collaboration approaches which were considered include Excel Services, and/or custom-built software; however, EASA has provided clear benefits over these strategies.
2310.18912
Hao Zhang
Hao Zhang, Yang Liu, Xiaoyan Liu, Tianming Liang, Gaurav Sharma, Liang Xue, and Maozu Guo
Sentence Bag Graph Formulation for Biomedical Distant Supervision Relation Extraction
in IEEE Transactions on Knowledge and Data Engineering, 2024
null
10.1109/TKDE.2024.3377229
null
cs.LG cs.CL
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We introduce a novel graph-based framework for alleviating key challenges in distantly-supervised relation extraction and demonstrate its effectiveness in the challenging and important domain of biomedical data. Specifically, we propose a graph view of sentence bags referring to an entity pair, which enables message-passing based aggregation of information related to the entity pair over the sentence bag. The proposed framework alleviates the common problem of noisy labeling in distantly supervised relation extraction and also effectively incorporates inter-dependencies between sentences within a bag. Extensive experiments on two large-scale biomedical relation datasets and the widely utilized NYT dataset demonstrate that our proposed framework significantly outperforms the state-of-the-art methods for biomedical distant supervision relation extraction while also providing excellent performance for relation extraction in the general text mining domain.
[ { "created": "Sun, 29 Oct 2023 05:48:04 GMT", "version": "v1" } ]
2024-04-08
[ [ "Zhang", "Hao", "" ], [ "Liu", "Yang", "" ], [ "Liu", "Xiaoyan", "" ], [ "Liang", "Tianming", "" ], [ "Sharma", "Gaurav", "" ], [ "Xue", "Liang", "" ], [ "Guo", "Maozu", "" ] ]
We introduce a novel graph-based framework for alleviating key challenges in distantly-supervised relation extraction and demonstrate its effectiveness in the challenging and important domain of biomedical data. Specifically, we propose a graph view of sentence bags referring to an entity pair, which enables message-passing based aggregation of information related to the entity pair over the sentence bag. The proposed framework alleviates the common problem of noisy labeling in distantly supervised relation extraction and also effectively incorporates inter-dependencies between sentences within a bag. Extensive experiments on two large-scale biomedical relation datasets and the widely utilized NYT dataset demonstrate that our proposed framework significantly outperforms the state-of-the-art methods for biomedical distant supervision relation extraction while also providing excellent performance for relation extraction in the general text mining domain.
2405.19773
Jasper Uijlings
Tautvydas Misiunas and Hassan Mansoor and Jasper Uijlings and Oriana Riva and Victor Carbune
VQA Training Sets are Self-play Environments for Generating Few-shot Pools
null
null
null
null
cs.CV
http://creativecommons.org/licenses/by/4.0/
Large-language models and large-vision models are increasingly capable of solving compositional reasoning tasks, as measured by breakthroughs in visual-question answering benchmarks. However, state-of-the-art solutions often involve careful construction of large pre-training and fine-tuning datasets, which can be expensive. The use of external tools, whether other ML models, search engines, or APIs, can significantly improve performance by breaking down high-level reasoning questions into sub-questions that are answerable by individual tools, but this approach has similar dataset construction costs to teach fine-tuned models how to use the available tools. We propose a technique in which existing training sets can be directly used for constructing computational environments with task metrics as rewards. This enables a model to autonomously teach itself to use itself or another model as a tool. By doing so, we augment training sets by integrating external signals. The proposed method starts with zero-shot prompts and iteratively refines them by selecting few-shot examples that maximize the task metric on the training set. Our experiments showcase how Gemini learns how to use itself, or another smaller and specialized model such as ScreenAI, to iteratively improve performance on training sets. Our approach successfully generalizes and improves upon zeroshot performance on charts, infographics, and document visual question-answering datasets
[ { "created": "Thu, 30 May 2024 07:38:58 GMT", "version": "v1" } ]
2024-05-31
[ [ "Misiunas", "Tautvydas", "" ], [ "Mansoor", "Hassan", "" ], [ "Uijlings", "Jasper", "" ], [ "Riva", "Oriana", "" ], [ "Carbune", "Victor", "" ] ]
Large-language models and large-vision models are increasingly capable of solving compositional reasoning tasks, as measured by breakthroughs in visual-question answering benchmarks. However, state-of-the-art solutions often involve careful construction of large pre-training and fine-tuning datasets, which can be expensive. The use of external tools, whether other ML models, search engines, or APIs, can significantly improve performance by breaking down high-level reasoning questions into sub-questions that are answerable by individual tools, but this approach has similar dataset construction costs to teach fine-tuned models how to use the available tools. We propose a technique in which existing training sets can be directly used for constructing computational environments with task metrics as rewards. This enables a model to autonomously teach itself to use itself or another model as a tool. By doing so, we augment training sets by integrating external signals. The proposed method starts with zero-shot prompts and iteratively refines them by selecting few-shot examples that maximize the task metric on the training set. Our experiments showcase how Gemini learns how to use itself, or another smaller and specialized model such as ScreenAI, to iteratively improve performance on training sets. Our approach successfully generalizes and improves upon zeroshot performance on charts, infographics, and document visual question-answering datasets
cs/0511012
Hamilton Link
Hamilton Link and Randall A. LaViolette and Jared Saia and Terran Lane
Parameters Affecting the Resilience of Scale-Free Networks to Random Failures
12 pages, 7 figures. Submitting to Phys. Rev. Lett
null
null
null
cs.NI cs.AR cs.MA
null
It is commonly believed that scale-free networks are robust to massive numbers of random node deletions. For example, Cohen et al. study scale-free networks including some which approximate the measured degree distribution of the Internet. Their results suggest that if each node in this network failed independently with probability 0.99, the remaining network would continue to have a giant component. In this paper, we show that a large and important subclass of scale-free networks are not robust to massive numbers of random node deletions for practical purposes. In particular, we study finite scale-free networks which have minimum node degree of 1 and a power-law degree distribution beginning with nodes of degree 1 (power-law networks). We show that, in a power-law network approximating the Internet's reported distribution, when the probability of deletion of each node is 0.5 only about 25% of the surviving nodes in the network remain connected in a giant component, and the giant component does not persist beyond a critical failure rate of 0.9. The new result is partially due to improved analytical accommodation of the large number of degree-0 nodes that result after node deletions. Our results apply to finite power-law networks with a wide range of power-law exponents, including Internet-like networks. We give both analytical and empirical evidence that such networks are not generally robust to massive random node deletions.
[ { "created": "Wed, 2 Nov 2005 22:22:14 GMT", "version": "v1" } ]
2007-05-23
[ [ "Link", "Hamilton", "" ], [ "LaViolette", "Randall A.", "" ], [ "Saia", "Jared", "" ], [ "Lane", "Terran", "" ] ]
It is commonly believed that scale-free networks are robust to massive numbers of random node deletions. For example, Cohen et al. study scale-free networks including some which approximate the measured degree distribution of the Internet. Their results suggest that if each node in this network failed independently with probability 0.99, the remaining network would continue to have a giant component. In this paper, we show that a large and important subclass of scale-free networks are not robust to massive numbers of random node deletions for practical purposes. In particular, we study finite scale-free networks which have minimum node degree of 1 and a power-law degree distribution beginning with nodes of degree 1 (power-law networks). We show that, in a power-law network approximating the Internet's reported distribution, when the probability of deletion of each node is 0.5 only about 25% of the surviving nodes in the network remain connected in a giant component, and the giant component does not persist beyond a critical failure rate of 0.9. The new result is partially due to improved analytical accommodation of the large number of degree-0 nodes that result after node deletions. Our results apply to finite power-law networks with a wide range of power-law exponents, including Internet-like networks. We give both analytical and empirical evidence that such networks are not generally robust to massive random node deletions.
2406.10738
Yao Zhao
Yao Zhao, Kwang-Sung Jun, Tanner Fiez, Lalit Jain
Adaptive Experimentation When You Can't Experiment
null
null
null
null
cs.LG stat.ME
http://creativecommons.org/licenses/by/4.0/
This paper introduces the \emph{confounded pure exploration transductive linear bandit} (\texttt{CPET-LB}) problem. As a motivating example, often online services cannot directly assign users to specific control or treatment experiences either for business or practical reasons. In these settings, naively comparing treatment and control groups that may result from self-selection can lead to biased estimates of underlying treatment effects. Instead, online services can employ a properly randomized encouragement that incentivizes users toward a specific treatment. Our methodology provides online services with an adaptive experimental design approach for learning the best-performing treatment for such \textit{encouragement designs}. We consider a more general underlying model captured by a linear structural equation and formulate pure exploration linear bandits in this setting. Though pure exploration has been extensively studied in standard adaptive experimental design settings, we believe this is the first work considering a setting where noise is confounded. Elimination-style algorithms using experimental design methods in combination with a novel finite-time confidence interval on an instrumental variable style estimator are presented with sample complexity upper bounds nearly matching a minimax lower bound. Finally, experiments are conducted that demonstrate the efficacy of our approach.
[ { "created": "Sat, 15 Jun 2024 20:54:48 GMT", "version": "v1" } ]
2024-06-18
[ [ "Zhao", "Yao", "" ], [ "Jun", "Kwang-Sung", "" ], [ "Fiez", "Tanner", "" ], [ "Jain", "Lalit", "" ] ]
This paper introduces the \emph{confounded pure exploration transductive linear bandit} (\texttt{CPET-LB}) problem. As a motivating example, often online services cannot directly assign users to specific control or treatment experiences either for business or practical reasons. In these settings, naively comparing treatment and control groups that may result from self-selection can lead to biased estimates of underlying treatment effects. Instead, online services can employ a properly randomized encouragement that incentivizes users toward a specific treatment. Our methodology provides online services with an adaptive experimental design approach for learning the best-performing treatment for such \textit{encouragement designs}. We consider a more general underlying model captured by a linear structural equation and formulate pure exploration linear bandits in this setting. Though pure exploration has been extensively studied in standard adaptive experimental design settings, we believe this is the first work considering a setting where noise is confounded. Elimination-style algorithms using experimental design methods in combination with a novel finite-time confidence interval on an instrumental variable style estimator are presented with sample complexity upper bounds nearly matching a minimax lower bound. Finally, experiments are conducted that demonstrate the efficacy of our approach.
1611.03898
Thai Pham
Derek Farren and Thai Pham and Marco Alban-Hidalgo
Low Latency Anomaly Detection and Bayesian Network Prediction of Anomaly Likelihood
null
null
null
null
cs.LG stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We develop a supervised machine learning model that detects anomalies in systems in real time. Our model processes unbounded streams of data into time series which then form the basis of a low-latency anomaly detection model. Moreover, we extend our preliminary goal of just anomaly detection to simultaneous anomaly prediction. We approach this very challenging problem by developing a Bayesian Network framework that captures the information about the parameters of the lagged regressors calibrated in the first part of our approach and use this structure to learn local conditional probability distributions.
[ { "created": "Fri, 11 Nov 2016 22:20:41 GMT", "version": "v1" } ]
2016-11-16
[ [ "Farren", "Derek", "" ], [ "Pham", "Thai", "" ], [ "Alban-Hidalgo", "Marco", "" ] ]
We develop a supervised machine learning model that detects anomalies in systems in real time. Our model processes unbounded streams of data into time series which then form the basis of a low-latency anomaly detection model. Moreover, we extend our preliminary goal of just anomaly detection to simultaneous anomaly prediction. We approach this very challenging problem by developing a Bayesian Network framework that captures the information about the parameters of the lagged regressors calibrated in the first part of our approach and use this structure to learn local conditional probability distributions.
2211.16958
Prerak Srivastava
Prerak Srivastava, Antoine Deleforge, Archontis Politis, Emmanuel Vincent
How to (virtually) train your speaker localizer
Published in INTERSPEECH 2023
null
null
null
cs.SD eess.AS
http://creativecommons.org/licenses/by/4.0/
Learning-based methods have become ubiquitous in speaker localization. Existing systems rely on simulated training sets for the lack of sufficiently large, diverse and annotated real datasets. Most room acoustics simulators used for this purpose rely on the image source method (ISM) because of its computational efficiency. This paper argues that carefully extending the ISM to incorporate more realistic surface, source and microphone responses into training sets can significantly boost the real-world performance of speaker localization systems. It is shown that increasing the training-set realism of a state-of-the-art direction-of-arrival estimator yields consistent improvements across three different real test sets featuring human speakers in a variety of rooms and various microphone arrays. An ablation study further reveals that every added layer of realism contributes positively to these improvements.
[ { "created": "Wed, 30 Nov 2022 13:01:11 GMT", "version": "v1" }, { "created": "Thu, 25 May 2023 14:51:43 GMT", "version": "v2" } ]
2023-05-26
[ [ "Srivastava", "Prerak", "" ], [ "Deleforge", "Antoine", "" ], [ "Politis", "Archontis", "" ], [ "Vincent", "Emmanuel", "" ] ]
Learning-based methods have become ubiquitous in speaker localization. Existing systems rely on simulated training sets for the lack of sufficiently large, diverse and annotated real datasets. Most room acoustics simulators used for this purpose rely on the image source method (ISM) because of its computational efficiency. This paper argues that carefully extending the ISM to incorporate more realistic surface, source and microphone responses into training sets can significantly boost the real-world performance of speaker localization systems. It is shown that increasing the training-set realism of a state-of-the-art direction-of-arrival estimator yields consistent improvements across three different real test sets featuring human speakers in a variety of rooms and various microphone arrays. An ablation study further reveals that every added layer of realism contributes positively to these improvements.
1808.04337
Samir Chowdhury
Samir Chowdhury and Facundo M\'emoli
The Gromov-Wasserstein distance between networks and stable network invariants
To appear in Information and Inference. Current version is a substantial update from the previous version and includes new computational experiments and also new results on the Gromov-Prokhorov distance between spheres
null
null
null
cs.DM math.MG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We define a metric---the network Gromov-Wasserstein distance---on weighted, directed networks that is sensitive to the presence of outliers. In addition to proving its theoretical properties, we supply network invariants based on optimal transport that approximate this distance by means of lower bounds. We test these methods on a range of simulated network datasets and on a dataset of real-world global bilateral migration. For our simulations, we define a network generative model based on the stochastic block model. This may be of independent interest for benchmarking purposes.
[ { "created": "Mon, 13 Aug 2018 17:24:45 GMT", "version": "v1" }, { "created": "Wed, 4 Sep 2019 14:46:48 GMT", "version": "v2" } ]
2019-09-05
[ [ "Chowdhury", "Samir", "" ], [ "Mémoli", "Facundo", "" ] ]
We define a metric---the network Gromov-Wasserstein distance---on weighted, directed networks that is sensitive to the presence of outliers. In addition to proving its theoretical properties, we supply network invariants based on optimal transport that approximate this distance by means of lower bounds. We test these methods on a range of simulated network datasets and on a dataset of real-world global bilateral migration. For our simulations, we define a network generative model based on the stochastic block model. This may be of independent interest for benchmarking purposes.
2301.10827
Matthew Alan Le Brun
Matthew Alan Le Brun and Ornela Dardha
MAG$\pi$: Types for Failure-Prone Communication
To be published in ESOP'23
null
null
null
cs.PL
http://creativecommons.org/licenses/by/4.0/
Multiparty Session Types (MPST) are a typing discipline for communication-centric systems, guaranteeing communication safety, deadlock freedom and protocol compliance. Several works have emerged which model failures and introduce fault-tolerance techniques. However, such works often make assumptions on the underlying network, e.g., TCP-based communication where messages are guaranteed to be delivered; or adopt centralised reliable nodes and an ad-hoc notion of reliability; or only address a single kind of failure, such as node crash failures. In this work, we develop MAG$\pi$ -- a Multiparty, Asynchronous and Generalised $\pi$-calculus, which is the first language and type system to accommodate in unison: (i) the widest range of non-Byzantine faults, including message loss, delays and reordering; crash failures and link failures; and network partitioning; (ii) a novel and most general notion of reliability, taking into account the viewpoint of each participant in the protocol; (iii) a spectrum of network assumptions from the lowest UDP-based network programming to the TCP-based application level. We prove subject reduction and session fidelity; process properties (deadlock freedom, termination, etc.); failure-handling safety and reliability adherence.
[ { "created": "Wed, 25 Jan 2023 21:04:02 GMT", "version": "v1" } ]
2023-01-27
[ [ "Brun", "Matthew Alan Le", "" ], [ "Dardha", "Ornela", "" ] ]
Multiparty Session Types (MPST) are a typing discipline for communication-centric systems, guaranteeing communication safety, deadlock freedom and protocol compliance. Several works have emerged which model failures and introduce fault-tolerance techniques. However, such works often make assumptions on the underlying network, e.g., TCP-based communication where messages are guaranteed to be delivered; or adopt centralised reliable nodes and an ad-hoc notion of reliability; or only address a single kind of failure, such as node crash failures. In this work, we develop MAG$\pi$ -- a Multiparty, Asynchronous and Generalised $\pi$-calculus, which is the first language and type system to accommodate in unison: (i) the widest range of non-Byzantine faults, including message loss, delays and reordering; crash failures and link failures; and network partitioning; (ii) a novel and most general notion of reliability, taking into account the viewpoint of each participant in the protocol; (iii) a spectrum of network assumptions from the lowest UDP-based network programming to the TCP-based application level. We prove subject reduction and session fidelity; process properties (deadlock freedom, termination, etc.); failure-handling safety and reliability adherence.
2110.09570
Arijit Nag
Arijit Nag, Bidisha Samanta, Animesh Mukherjee, Niloy Ganguly, Soumen Chakrabarti
A Data Bootstrapping Recipe for Low Resource Multilingual Relation Classification
null
null
null
null
cs.CL
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Relation classification (sometimes called 'extraction') requires trustworthy datasets for fine-tuning large language models, as well as for evaluation. Data collection is challenging for Indian languages, because they are syntactically and morphologically diverse, as well as different from resource-rich languages like English. Despite recent interest in deep generative models for Indian languages, relation classification is still not well served by public data sets. In response, we present IndoRE, a dataset with 21K entity and relation tagged gold sentences in three Indian languages, plus English. We start with a multilingual BERT (mBERT) based system that captures entity span positions and type information and provides competitive monolingual relation classification. Using this system, we explore and compare transfer mechanisms between languages. In particular, we study the accuracy efficiency tradeoff between expensive gold instances vs. translated and aligned 'silver' instances. We release the dataset for future research.
[ { "created": "Mon, 18 Oct 2021 18:40:46 GMT", "version": "v1" } ]
2021-10-20
[ [ "Nag", "Arijit", "" ], [ "Samanta", "Bidisha", "" ], [ "Mukherjee", "Animesh", "" ], [ "Ganguly", "Niloy", "" ], [ "Chakrabarti", "Soumen", "" ] ]
Relation classification (sometimes called 'extraction') requires trustworthy datasets for fine-tuning large language models, as well as for evaluation. Data collection is challenging for Indian languages, because they are syntactically and morphologically diverse, as well as different from resource-rich languages like English. Despite recent interest in deep generative models for Indian languages, relation classification is still not well served by public data sets. In response, we present IndoRE, a dataset with 21K entity and relation tagged gold sentences in three Indian languages, plus English. We start with a multilingual BERT (mBERT) based system that captures entity span positions and type information and provides competitive monolingual relation classification. Using this system, we explore and compare transfer mechanisms between languages. In particular, we study the accuracy efficiency tradeoff between expensive gold instances vs. translated and aligned 'silver' instances. We release the dataset for future research.
1805.12081
Md. Mostafa Kamal Sarker
Md. Mostafa Kamal Sarker, Mohammed Jabreel, Hatem A. Rashwan, Syeda Furruka Banu, Antonio Moreno, Petia Radeva, Domenec Puig
CuisineNet: Food Attributes Classification using Multi-scale Convolution Network
8 pages, Submitted in CCIA 2018
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Diversity of food and its attributes represents the culinary habits of peoples from different countries. Thus, this paper addresses the problem of identifying food culture of people around the world and its flavor by classifying two main food attributes, cuisine and flavor. A deep learning model based on multi-scale convotuional networks is proposed for extracting more accurate features from input images. The aggregation of multi-scale convolution layers with different kernel size is also used for weighting the features results from different scales. In addition, a joint loss function based on Negative Log Likelihood (NLL) is used to fit the model probability to multi labeled classes for multi-modal classification task. Furthermore, this work provides a new dataset for food attributes, so-called Yummly48K, extracted from the popular food website, Yummly. Our model is assessed on the constructed Yummly48K dataset. The experimental results show that our proposed method yields 65% and 62% average F1 score on validation and test set which outperforming the state-of-the-art models.
[ { "created": "Wed, 30 May 2018 16:56:32 GMT", "version": "v1" }, { "created": "Fri, 8 Jun 2018 16:44:15 GMT", "version": "v2" } ]
2018-06-11
[ [ "Sarker", "Md. Mostafa Kamal", "" ], [ "Jabreel", "Mohammed", "" ], [ "Rashwan", "Hatem A.", "" ], [ "Banu", "Syeda Furruka", "" ], [ "Moreno", "Antonio", "" ], [ "Radeva", "Petia", "" ], [ "Puig", "Domenec", "" ] ]
Diversity of food and its attributes represents the culinary habits of peoples from different countries. Thus, this paper addresses the problem of identifying food culture of people around the world and its flavor by classifying two main food attributes, cuisine and flavor. A deep learning model based on multi-scale convotuional networks is proposed for extracting more accurate features from input images. The aggregation of multi-scale convolution layers with different kernel size is also used for weighting the features results from different scales. In addition, a joint loss function based on Negative Log Likelihood (NLL) is used to fit the model probability to multi labeled classes for multi-modal classification task. Furthermore, this work provides a new dataset for food attributes, so-called Yummly48K, extracted from the popular food website, Yummly. Our model is assessed on the constructed Yummly48K dataset. The experimental results show that our proposed method yields 65% and 62% average F1 score on validation and test set which outperforming the state-of-the-art models.
1705.04269
Xingqin Lin
Xingqin Lin, Johan Bergman, Fredrik Gunnarsson, Olof Liberg, Sara Modarres Razavi, Hazhir Shokri Razaghi, Henrik Ryd\'en, and Yutao Sui
Positioning for the Internet of Things: A 3GPP Perspective
8 pages; 7 figures; 1 table; submitted for publication
null
null
null
cs.NI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Many use cases in the Internet of Things (IoT) will require or benefit from location information, making positioning a vital dimension of the IoT. The 3rd Generation Partnership Project (3GPP) has dedicated a significant effort during its Release 14 to enhance positioning support for its IoT technologies to further improve the 3GPP-based IoT eco-system. In this article, we identify the design challenges of positioning support in Long-Term Evolution Machine Type Communication (LTE-M) and Narrowband IoT (NB-IoT), and overview the 3GPP's work in enhancing the positioning support for LTE-M and NB-IoT. We focus on Observed Time Difference of Arrival (OTDOA), which is a downlink based positioning method. We provide an overview of the OTDOA architecture and protocols, summarize the designs of OTDOA positioning reference signals, and present simulation results to illustrate the positioning performance.
[ { "created": "Thu, 13 Apr 2017 00:37:56 GMT", "version": "v1" }, { "created": "Fri, 16 Jun 2017 16:03:38 GMT", "version": "v2" } ]
2017-06-19
[ [ "Lin", "Xingqin", "" ], [ "Bergman", "Johan", "" ], [ "Gunnarsson", "Fredrik", "" ], [ "Liberg", "Olof", "" ], [ "Razavi", "Sara Modarres", "" ], [ "Razaghi", "Hazhir Shokri", "" ], [ "Rydén", "Henrik", "" ], [ "Sui", "Yutao", "" ] ]
Many use cases in the Internet of Things (IoT) will require or benefit from location information, making positioning a vital dimension of the IoT. The 3rd Generation Partnership Project (3GPP) has dedicated a significant effort during its Release 14 to enhance positioning support for its IoT technologies to further improve the 3GPP-based IoT eco-system. In this article, we identify the design challenges of positioning support in Long-Term Evolution Machine Type Communication (LTE-M) and Narrowband IoT (NB-IoT), and overview the 3GPP's work in enhancing the positioning support for LTE-M and NB-IoT. We focus on Observed Time Difference of Arrival (OTDOA), which is a downlink based positioning method. We provide an overview of the OTDOA architecture and protocols, summarize the designs of OTDOA positioning reference signals, and present simulation results to illustrate the positioning performance.
1903.12271
Mahmoud El-Haj
Mahmoud El-Haj, Paul Rayson, Martin Walker, Steven Young, Vasiliki Simaki
In Search of Meaning: Lessons, Resources and Next Steps for Computational Analysis of Financial Discourse
70 page, 18 pages of references, Journal Article
null
null
null
cs.CL
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We critically assess mainstream accounting and finance research applying methods from computational linguistics (CL) to study financial discourse. We also review common themes and innovations in the literature and assess the incremental contributions of work applying CL methods over manual content analysis. Key conclusions emerging from our analysis are: (a) accounting and finance research is behind the curve in terms of CL methods generally and word sense disambiguation in particular; (b) implementation issues mean the proposed benefits of CL are often less pronounced than proponents suggest; (c) structural issues limit practical relevance; and (d) CL methods and high quality manual analysis represent complementary approaches to analyzing financial discourse. We describe four CL tools that have yet to gain traction in mainstream AF research but which we believe offer promising ways to enhance the study of meaning in financial discourse. The four tools are named entity recognition (NER), summarization, semantics and corpus linguistics.
[ { "created": "Thu, 28 Mar 2019 21:12:59 GMT", "version": "v1" } ]
2019-04-01
[ [ "El-Haj", "Mahmoud", "" ], [ "Rayson", "Paul", "" ], [ "Walker", "Martin", "" ], [ "Young", "Steven", "" ], [ "Simaki", "Vasiliki", "" ] ]
We critically assess mainstream accounting and finance research applying methods from computational linguistics (CL) to study financial discourse. We also review common themes and innovations in the literature and assess the incremental contributions of work applying CL methods over manual content analysis. Key conclusions emerging from our analysis are: (a) accounting and finance research is behind the curve in terms of CL methods generally and word sense disambiguation in particular; (b) implementation issues mean the proposed benefits of CL are often less pronounced than proponents suggest; (c) structural issues limit practical relevance; and (d) CL methods and high quality manual analysis represent complementary approaches to analyzing financial discourse. We describe four CL tools that have yet to gain traction in mainstream AF research but which we believe offer promising ways to enhance the study of meaning in financial discourse. The four tools are named entity recognition (NER), summarization, semantics and corpus linguistics.
1510.08301
Wonju Lee
Wonju Lee, Osvaldo Simeone, Joonhyuk Kang and Shlomo Shamai
Multivariate Fronthaul Quantization for Downlink C-RAN
Submitted
null
10.1109/TSP.2016.2593682
null
cs.IT math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The Cloud-Radio Access Network (C-RAN) cellular architecture relies on the transfer of complex baseband signals to and from a central unit (CU) over digital fronthaul links to enable the virtualization of the baseband processing functionalities of distributed radio units (RUs). The standard design of digital fronthauling is based on either scalar quantization or on more sophisticated point to-point compression techniques operating on baseband signals. Motivated by network-information theoretic results, techniques for fronthaul quantization and compression that improve over point-to-point solutions by allowing for joint processing across multiple fronthaul links at the CU have been recently proposed for both the uplink and the downlink. For the downlink, a form of joint compression, known in network information theory as multivariate compression, was shown to be advantageous under a non-constructive asymptotic information-theoretic framework. In this paper, instead, the design of a practical symbol-by-symbol fronthaul quantization algorithm that implements the idea of multivariate compression is investigated for the C-RAN downlink. As compared to current standards, the proposed multivariate quantization (MQ) only requires changes in the CU processing while no modification is needed at the RUs. The algorithm is extended to enable the joint optimization of downlink precoding and quantization, reduced-complexity MQ via successive block quantization, and variable-length compression. Numerical results, which include performance evaluations over standard cellular models, demonstrate the advantages of MQ and the merits of a joint optimization with precoding.
[ { "created": "Wed, 28 Oct 2015 13:27:36 GMT", "version": "v1" }, { "created": "Mon, 11 Jul 2016 01:31:28 GMT", "version": "v2" } ]
2016-08-24
[ [ "Lee", "Wonju", "" ], [ "Simeone", "Osvaldo", "" ], [ "Kang", "Joonhyuk", "" ], [ "Shamai", "Shlomo", "" ] ]
The Cloud-Radio Access Network (C-RAN) cellular architecture relies on the transfer of complex baseband signals to and from a central unit (CU) over digital fronthaul links to enable the virtualization of the baseband processing functionalities of distributed radio units (RUs). The standard design of digital fronthauling is based on either scalar quantization or on more sophisticated point to-point compression techniques operating on baseband signals. Motivated by network-information theoretic results, techniques for fronthaul quantization and compression that improve over point-to-point solutions by allowing for joint processing across multiple fronthaul links at the CU have been recently proposed for both the uplink and the downlink. For the downlink, a form of joint compression, known in network information theory as multivariate compression, was shown to be advantageous under a non-constructive asymptotic information-theoretic framework. In this paper, instead, the design of a practical symbol-by-symbol fronthaul quantization algorithm that implements the idea of multivariate compression is investigated for the C-RAN downlink. As compared to current standards, the proposed multivariate quantization (MQ) only requires changes in the CU processing while no modification is needed at the RUs. The algorithm is extended to enable the joint optimization of downlink precoding and quantization, reduced-complexity MQ via successive block quantization, and variable-length compression. Numerical results, which include performance evaluations over standard cellular models, demonstrate the advantages of MQ and the merits of a joint optimization with precoding.
1201.0081
Yuan Liu Yuan Liu
Hao Zhang, Yuan Liu, and Meixia Tao
Resource Allocation with Subcarrier Pairing in OFDMA Two-Way Relay Networks
null
null
null
null
cs.IT math.IT
http://creativecommons.org/licenses/by/3.0/
This study considers an orthogonal frequency-division multiple-access (OFDMA)-based multi-user two-way relay network where multiple mobile stations (MSs) communicate with a common base station (BS) via multiple relay stations (RSs). We study the joint optimization problem of subcarrier-pairing based relay-power allocation, relay selection, and subcarrier assignment. The problem is formulated as a mixed integer programming problem. By using the dual method, we propose an efficient algorithm to solve the problem in an asymptotically optimal manner. Simulation results show that the proposed method can improve system performance significantly over the conventional methods.
[ { "created": "Fri, 30 Dec 2011 08:49:17 GMT", "version": "v1" } ]
2012-01-04
[ [ "Zhang", "Hao", "" ], [ "Liu", "Yuan", "" ], [ "Tao", "Meixia", "" ] ]
This study considers an orthogonal frequency-division multiple-access (OFDMA)-based multi-user two-way relay network where multiple mobile stations (MSs) communicate with a common base station (BS) via multiple relay stations (RSs). We study the joint optimization problem of subcarrier-pairing based relay-power allocation, relay selection, and subcarrier assignment. The problem is formulated as a mixed integer programming problem. By using the dual method, we propose an efficient algorithm to solve the problem in an asymptotically optimal manner. Simulation results show that the proposed method can improve system performance significantly over the conventional methods.
2110.11404
Edgar Du\'e\~nez-Guzm\'an
Edgar A. Du\'e\~nez-Guzm\'an, Kevin R. McKee, Yiran Mao, Ben Coppin, Silvia Chiappa, Alexander Sasha Vezhnevets, Michiel A. Bakker, Yoram Bachrach, Suzanne Sadedin, William Isaac, Karl Tuyls, Joel Z. Leibo
Statistical discrimination in learning agents
29 pages, 10 figures
null
null
null
cs.LG cs.AI cs.GT cs.MA
http://creativecommons.org/licenses/by/4.0/
Undesired bias afflicts both human and algorithmic decision making, and may be especially prevalent when information processing trade-offs incentivize the use of heuristics. One primary example is \textit{statistical discrimination} -- selecting social partners based not on their underlying attributes, but on readily perceptible characteristics that covary with their suitability for the task at hand. We present a theoretical model to examine how information processing influences statistical discrimination and test its predictions using multi-agent reinforcement learning with various agent architectures in a partner choice-based social dilemma. As predicted, statistical discrimination emerges in agent policies as a function of both the bias in the training population and of agent architecture. All agents showed substantial statistical discrimination, defaulting to using the readily available correlates instead of the outcome relevant features. We show that less discrimination emerges with agents that use recurrent neural networks, and when their training environment has less bias. However, all agent algorithms we tried still exhibited substantial bias after learning in biased training populations.
[ { "created": "Thu, 21 Oct 2021 18:28:57 GMT", "version": "v1" } ]
2021-10-25
[ [ "Duéñez-Guzmán", "Edgar A.", "" ], [ "McKee", "Kevin R.", "" ], [ "Mao", "Yiran", "" ], [ "Coppin", "Ben", "" ], [ "Chiappa", "Silvia", "" ], [ "Vezhnevets", "Alexander Sasha", "" ], [ "Bakker", "Michiel A.", "" ], [ "Bachrach", "Yoram", "" ], [ "Sadedin", "Suzanne", "" ], [ "Isaac", "William", "" ], [ "Tuyls", "Karl", "" ], [ "Leibo", "Joel Z.", "" ] ]
Undesired bias afflicts both human and algorithmic decision making, and may be especially prevalent when information processing trade-offs incentivize the use of heuristics. One primary example is \textit{statistical discrimination} -- selecting social partners based not on their underlying attributes, but on readily perceptible characteristics that covary with their suitability for the task at hand. We present a theoretical model to examine how information processing influences statistical discrimination and test its predictions using multi-agent reinforcement learning with various agent architectures in a partner choice-based social dilemma. As predicted, statistical discrimination emerges in agent policies as a function of both the bias in the training population and of agent architecture. All agents showed substantial statistical discrimination, defaulting to using the readily available correlates instead of the outcome relevant features. We show that less discrimination emerges with agents that use recurrent neural networks, and when their training environment has less bias. However, all agent algorithms we tried still exhibited substantial bias after learning in biased training populations.
2009.09719
arXiv Admin
Sagar Verma
A Survey on Machine Learning Applied to Dynamic Physical Systems
arXiv admin note: submission has been withdrawn by arXiv administrators due to inappropriate text overlap with external source
null
null
null
cs.LG cs.NE
http://creativecommons.org/licenses/by/4.0/
This survey is on recent advancements in the intersection of physical modeling and machine learning. We focus on the modeling of nonlinear systems which are closer to electric motors. Survey on motor control and fault detection in operation of electric motors has been done.
[ { "created": "Mon, 21 Sep 2020 09:41:54 GMT", "version": "v1" }, { "created": "Mon, 28 Sep 2020 13:27:14 GMT", "version": "v2" } ]
2020-09-29
[ [ "Verma", "Sagar", "" ] ]
This survey is on recent advancements in the intersection of physical modeling and machine learning. We focus on the modeling of nonlinear systems which are closer to electric motors. Survey on motor control and fault detection in operation of electric motors has been done.
1903.04959
Haotian Fu
Haotian Fu, Hongyao Tang, Jianye Hao, Zihan Lei, Yingfeng Chen, Changjie Fan
Deep Multi-Agent Reinforcement Learning with Discrete-Continuous Hybrid Action Spaces
null
IJCAI 2019
null
null
cs.LG cs.AI cs.MA stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Deep Reinforcement Learning (DRL) has been applied to address a variety of cooperative multi-agent problems with either discrete action spaces or continuous action spaces. However, to the best of our knowledge, no previous work has ever succeeded in applying DRL to multi-agent problems with discrete-continuous hybrid (or parameterized) action spaces which is very common in practice. Our work fills this gap by proposing two novel algorithms: Deep Multi-Agent Parameterized Q-Networks (Deep MAPQN) and Deep Multi-Agent Hierarchical Hybrid Q-Networks (Deep MAHHQN). We follow the centralized training but decentralized execution paradigm: different levels of communication between different agents are used to facilitate the training process, while each agent executes its policy independently based on local observations during execution. Our empirical results on several challenging tasks (simulated RoboCup Soccer and game Ghost Story) show that both Deep MAPQN and Deep MAHHQN are effective and significantly outperform existing independent deep parameterized Q-learning method.
[ { "created": "Tue, 12 Mar 2019 14:40:32 GMT", "version": "v1" } ]
2019-06-04
[ [ "Fu", "Haotian", "" ], [ "Tang", "Hongyao", "" ], [ "Hao", "Jianye", "" ], [ "Lei", "Zihan", "" ], [ "Chen", "Yingfeng", "" ], [ "Fan", "Changjie", "" ] ]
Deep Reinforcement Learning (DRL) has been applied to address a variety of cooperative multi-agent problems with either discrete action spaces or continuous action spaces. However, to the best of our knowledge, no previous work has ever succeeded in applying DRL to multi-agent problems with discrete-continuous hybrid (or parameterized) action spaces which is very common in practice. Our work fills this gap by proposing two novel algorithms: Deep Multi-Agent Parameterized Q-Networks (Deep MAPQN) and Deep Multi-Agent Hierarchical Hybrid Q-Networks (Deep MAHHQN). We follow the centralized training but decentralized execution paradigm: different levels of communication between different agents are used to facilitate the training process, while each agent executes its policy independently based on local observations during execution. Our empirical results on several challenging tasks (simulated RoboCup Soccer and game Ghost Story) show that both Deep MAPQN and Deep MAHHQN are effective and significantly outperform existing independent deep parameterized Q-learning method.
2003.00145
Nupur Patanker
Nupur Patanker, Sanjay Kumar Singh
Generalization of trace codes to places of higher degree
Due to error in Section 6 on the dimension of codes
null
null
null
cs.IT math.AG math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In this note, we give a construction of codes on algebraic function field $F/ \mathbb{F}_{q}$ using places of $F$ (not necessarily of degree one) and trace functions from various extensions of $\mathbb{F}_{q}$. This is a generalization of trace code of geometric Goppa codes to higher degree places. We compute a bound on the dimension of this code. Furthermore, we give a condition under which we get exact dimension of the code. We also determine a bound on the minimum distance of this code in terms of $B_{r}(F)$ ( the number of places of degree $r$ in $F$), $1 \leq r < \infty$. Few quasi-cyclic codes over $\mathbb{F}_{p}$ are also obtained as examples of these codes.
[ { "created": "Sat, 29 Feb 2020 01:19:05 GMT", "version": "v1" }, { "created": "Fri, 2 Oct 2020 16:05:02 GMT", "version": "v2" }, { "created": "Wed, 14 Apr 2021 03:05:04 GMT", "version": "v3" } ]
2021-04-15
[ [ "Patanker", "Nupur", "" ], [ "Singh", "Sanjay Kumar", "" ] ]
In this note, we give a construction of codes on algebraic function field $F/ \mathbb{F}_{q}$ using places of $F$ (not necessarily of degree one) and trace functions from various extensions of $\mathbb{F}_{q}$. This is a generalization of trace code of geometric Goppa codes to higher degree places. We compute a bound on the dimension of this code. Furthermore, we give a condition under which we get exact dimension of the code. We also determine a bound on the minimum distance of this code in terms of $B_{r}(F)$ ( the number of places of degree $r$ in $F$), $1 \leq r < \infty$. Few quasi-cyclic codes over $\mathbb{F}_{p}$ are also obtained as examples of these codes.
2305.10621
Yulin Sun
Yulin Sun, Qingming Qu, Chenxingyu Zhao, Arvind Krishnamurthy, Hong Chang, Ying Xiong
TSoR: TCP Socket over RDMA Container Network for Cloud Native Computing
null
null
null
null
cs.NI cs.DC
http://creativecommons.org/licenses/by/4.0/
Cloud-native containerized applications constantly seek high-performance and easy-to-operate container network solutions. RDMA network is a potential enabler with higher throughput and lower latency than the standard TCP/IP network stack. However, several challenges remain in equipping containerized applications with RDMA network: 1) How to deliver transparent improvements without modifying application code; 2) How to integrate RDMA-based network solutions with container orchestration systems; 3) How to efficiently utilize RDMA for container networks. In this paper, we present an RDMA-based container network solution, TCP Socket over RDMA (TSoR), which addresses all the above challenges. To transparently accelerate applications using POSIX socket interfaces without modifications, we integrate TSoR with a container runtime that can intercept system calls for socket interfaces. To be compatible with orchestration systems like Kubernetes, TSoR implements a container network following the Kubernetes network model and satisfies all requirements of the model. To leverage RDMA benefits, TSoR designs a high-performance network stack that efficiently transfers TCP traffic using RDMA network. Thus, TSoR provides a turn-key solution for existing Kubernetes clusters to adopt the high-performance RDMA network with minimal effort. Our evaluation results show that TSoR provides up to 2.3x higher throughput and 64\% lower latency for existing containerized applications, such as Redis key-value store and Node.js web server, with no code changes. TSoR code will be open-sourced.
[ { "created": "Thu, 18 May 2023 00:20:56 GMT", "version": "v1" } ]
2023-05-19
[ [ "Sun", "Yulin", "" ], [ "Qu", "Qingming", "" ], [ "Zhao", "Chenxingyu", "" ], [ "Krishnamurthy", "Arvind", "" ], [ "Chang", "Hong", "" ], [ "Xiong", "Ying", "" ] ]
Cloud-native containerized applications constantly seek high-performance and easy-to-operate container network solutions. RDMA network is a potential enabler with higher throughput and lower latency than the standard TCP/IP network stack. However, several challenges remain in equipping containerized applications with RDMA network: 1) How to deliver transparent improvements without modifying application code; 2) How to integrate RDMA-based network solutions with container orchestration systems; 3) How to efficiently utilize RDMA for container networks. In this paper, we present an RDMA-based container network solution, TCP Socket over RDMA (TSoR), which addresses all the above challenges. To transparently accelerate applications using POSIX socket interfaces without modifications, we integrate TSoR with a container runtime that can intercept system calls for socket interfaces. To be compatible with orchestration systems like Kubernetes, TSoR implements a container network following the Kubernetes network model and satisfies all requirements of the model. To leverage RDMA benefits, TSoR designs a high-performance network stack that efficiently transfers TCP traffic using RDMA network. Thus, TSoR provides a turn-key solution for existing Kubernetes clusters to adopt the high-performance RDMA network with minimal effort. Our evaluation results show that TSoR provides up to 2.3x higher throughput and 64\% lower latency for existing containerized applications, such as Redis key-value store and Node.js web server, with no code changes. TSoR code will be open-sourced.
2407.02442
Hao Xu
Hao Xu, Kai-Kit Wong, Giuseppe Caire
A New Achievable Region of the $K$-User MAC Wiretap Channel with Confidential and Open Messages Under Strong Secrecy
61 pages, 15 figures. arXiv admin note: text overlap with arXiv:2209.05403
null
null
null
cs.IT math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This paper investigates the achievable region of a $K$-user discrete memoryless (DM) multiple access wiretap (MAC-WT) channel, where each user transmits both secret and open messages. All these messages are intended for Bob, while Eve is only interested in the secret messages. In the achievable coding strategy, the confidential information is protected by open messages and also by the introduction of auxiliary messages. When introducing an auxiliary message, one has to ensure that, on one hand, its rate is large enough for protecting the secret message from Eve and, on the other hand, the resulting sum rate (together with the secret and open message rate) does not exceed Bob's decoding capability. This yields an inequality structure involving the rates of all users' secret, open, and auxiliary messages. To obtain the rate region, the auxiliary message rates must be eliminated from the system of inequalities. A direct application of the Fourier-Motzkin elimination procedure is elusive since a) it requires that the number of users $K$ is explicitly given, and b) even for small $K = 3, 4, \ldots$, the number of inequalities becomes extremely large. We prove the result for general $K$ through the combined use of Fourier-Motzkin elimination procedure and mathematical induction. This paper adopts the strong secrecy metric, characterized by information leakage. To prove the achievability under this criterion, we analyze the resolvability region of a $K$-user DM-MAC channel. In addition, we show that users with zero secrecy rate can play different roles and use different strategies in encoding their messages. These strategies yield non-redundant rate inequalities. By considering all possible coding strategies, we provide a new achievable region for the considered channel, and show that it strictly improves those already known in the existing literature by considering a specific example.
[ { "created": "Tue, 2 Jul 2024 17:17:53 GMT", "version": "v1" } ]
2024-07-03
[ [ "Xu", "Hao", "" ], [ "Wong", "Kai-Kit", "" ], [ "Caire", "Giuseppe", "" ] ]
This paper investigates the achievable region of a $K$-user discrete memoryless (DM) multiple access wiretap (MAC-WT) channel, where each user transmits both secret and open messages. All these messages are intended for Bob, while Eve is only interested in the secret messages. In the achievable coding strategy, the confidential information is protected by open messages and also by the introduction of auxiliary messages. When introducing an auxiliary message, one has to ensure that, on one hand, its rate is large enough for protecting the secret message from Eve and, on the other hand, the resulting sum rate (together with the secret and open message rate) does not exceed Bob's decoding capability. This yields an inequality structure involving the rates of all users' secret, open, and auxiliary messages. To obtain the rate region, the auxiliary message rates must be eliminated from the system of inequalities. A direct application of the Fourier-Motzkin elimination procedure is elusive since a) it requires that the number of users $K$ is explicitly given, and b) even for small $K = 3, 4, \ldots$, the number of inequalities becomes extremely large. We prove the result for general $K$ through the combined use of Fourier-Motzkin elimination procedure and mathematical induction. This paper adopts the strong secrecy metric, characterized by information leakage. To prove the achievability under this criterion, we analyze the resolvability region of a $K$-user DM-MAC channel. In addition, we show that users with zero secrecy rate can play different roles and use different strategies in encoding their messages. These strategies yield non-redundant rate inequalities. By considering all possible coding strategies, we provide a new achievable region for the considered channel, and show that it strictly improves those already known in the existing literature by considering a specific example.
2303.02673
Jiguo Li
Jiguo Li, Tianzi Zhang, Xiaobin Liu, Lirong Zheng
Time-frequency Network for Robust Speaker Recognition
5pages, 3 figures
null
null
null
cs.SD cs.MM eess.AS
http://creativecommons.org/licenses/by-nc-sa/4.0/
The wide deployment of speech-based biometric systems usually demands high-performance speaker recognition algorithms. However, most of the prior works for speaker recognition either process the speech in the frequency domain or time domain, which may produce suboptimal results because both time and frequency domains are important for speaker recognition. In this paper, we attempt to analyze the speech signal in both time and frequency domains and propose the time-frequency network~(TFN) for speaker recognition by extracting and fusing the features in the two domains. Based on the recent advance of deep neural networks, we propose a convolution neural network to encode the raw speech waveform and the frequency spectrum into domain-specific features, which are then fused and transformed into a classification feature space for speaker recognition. Experimental results on the publicly available datasets TIMIT and LibriSpeech show that our framework is effective to combine the information in the two domains and performs better than the state-of-the-art methods for speaker recognition.
[ { "created": "Sun, 5 Mar 2023 13:48:47 GMT", "version": "v1" }, { "created": "Tue, 7 Mar 2023 02:38:16 GMT", "version": "v2" } ]
2023-03-08
[ [ "Li", "Jiguo", "" ], [ "Zhang", "Tianzi", "" ], [ "Liu", "Xiaobin", "" ], [ "Zheng", "Lirong", "" ] ]
The wide deployment of speech-based biometric systems usually demands high-performance speaker recognition algorithms. However, most of the prior works for speaker recognition either process the speech in the frequency domain or time domain, which may produce suboptimal results because both time and frequency domains are important for speaker recognition. In this paper, we attempt to analyze the speech signal in both time and frequency domains and propose the time-frequency network~(TFN) for speaker recognition by extracting and fusing the features in the two domains. Based on the recent advance of deep neural networks, we propose a convolution neural network to encode the raw speech waveform and the frequency spectrum into domain-specific features, which are then fused and transformed into a classification feature space for speaker recognition. Experimental results on the publicly available datasets TIMIT and LibriSpeech show that our framework is effective to combine the information in the two domains and performs better than the state-of-the-art methods for speaker recognition.
2104.04909
Saed Rezayi
Saed Rezayi, Handong Zhao, Sungchul Kim, Ryan A. Rossi, Nedim Lipka, Sheng Li
Edge: Enriching Knowledge Graph Embeddings with External Text
Accepted in NAACL'21
null
null
null
cs.CL cs.LG
http://creativecommons.org/licenses/by-nc-sa/4.0/
Knowledge graphs suffer from sparsity which degrades the quality of representations generated by various methods. While there is an abundance of textual information throughout the web and many existing knowledge bases, aligning information across these diverse data sources remains a challenge in the literature. Previous work has partially addressed this issue by enriching knowledge graph entities based on "hard" co-occurrence of words present in the entities of the knowledge graphs and external text, while we achieve "soft" augmentation by proposing a knowledge graph enrichment and embedding framework named Edge. Given an original knowledge graph, we first generate a rich but noisy augmented graph using external texts in semantic and structural level. To distill the relevant knowledge and suppress the introduced noise, we design a graph alignment term in a shared embedding space between the original graph and augmented graph. To enhance the embedding learning on the augmented graph, we further regularize the locality relationship of target entity based on negative sampling. Experimental results on four benchmark datasets demonstrate the robustness and effectiveness of Edge in link prediction and node classification.
[ { "created": "Sun, 11 Apr 2021 03:47:06 GMT", "version": "v1" } ]
2021-04-13
[ [ "Rezayi", "Saed", "" ], [ "Zhao", "Handong", "" ], [ "Kim", "Sungchul", "" ], [ "Rossi", "Ryan A.", "" ], [ "Lipka", "Nedim", "" ], [ "Li", "Sheng", "" ] ]
Knowledge graphs suffer from sparsity which degrades the quality of representations generated by various methods. While there is an abundance of textual information throughout the web and many existing knowledge bases, aligning information across these diverse data sources remains a challenge in the literature. Previous work has partially addressed this issue by enriching knowledge graph entities based on "hard" co-occurrence of words present in the entities of the knowledge graphs and external text, while we achieve "soft" augmentation by proposing a knowledge graph enrichment and embedding framework named Edge. Given an original knowledge graph, we first generate a rich but noisy augmented graph using external texts in semantic and structural level. To distill the relevant knowledge and suppress the introduced noise, we design a graph alignment term in a shared embedding space between the original graph and augmented graph. To enhance the embedding learning on the augmented graph, we further regularize the locality relationship of target entity based on negative sampling. Experimental results on four benchmark datasets demonstrate the robustness and effectiveness of Edge in link prediction and node classification.
2009.14361
Angus Addlesee
Angus Addlesee and Pierre Albert
Ethically Collecting Multi-Modal Spontaneous Conversations with People that have Cognitive Impairments
Published at LREC's Workshop on Legal and Ethical Issues in Human Language Technologies 2020
LREC Workshop on Legal and Ethical Issues in Human Language Technologies (2020) 15-20
null
null
cs.CL cs.CY
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In order to make spoken dialogue systems (such as Amazon Alexa or Google Assistant) more accessible and naturally interactive for people with cognitive impairments, appropriate data must be obtainable. Recordings of multi-modal spontaneous conversations with vulnerable user groups are scarce however and this valuable data is challenging to collect. Researchers that call for this data are commonly inexperienced in ethical and legal issues around working with vulnerable participants. Additionally, standard recording equipment is insecure and should not be used to capture sensitive data. We spent a year consulting experts on how to ethically capture and share recordings of multi-modal spontaneous conversations with vulnerable user groups. In this paper we provide guidance, collated from these experts, on how to ethically collect such data and we present a new system - "CUSCO" - to capture, transport and exchange sensitive data securely. This framework is intended to be easily followed and implemented to encourage further publications of similar corpora. Using this guide and secure recording system, researchers can review and refine their ethical measures.
[ { "created": "Wed, 30 Sep 2020 00:57:33 GMT", "version": "v1" } ]
2020-10-01
[ [ "Addlesee", "Angus", "" ], [ "Albert", "Pierre", "" ] ]
In order to make spoken dialogue systems (such as Amazon Alexa or Google Assistant) more accessible and naturally interactive for people with cognitive impairments, appropriate data must be obtainable. Recordings of multi-modal spontaneous conversations with vulnerable user groups are scarce however and this valuable data is challenging to collect. Researchers that call for this data are commonly inexperienced in ethical and legal issues around working with vulnerable participants. Additionally, standard recording equipment is insecure and should not be used to capture sensitive data. We spent a year consulting experts on how to ethically capture and share recordings of multi-modal spontaneous conversations with vulnerable user groups. In this paper we provide guidance, collated from these experts, on how to ethically collect such data and we present a new system - "CUSCO" - to capture, transport and exchange sensitive data securely. This framework is intended to be easily followed and implemented to encourage further publications of similar corpora. Using this guide and secure recording system, researchers can review and refine their ethical measures.
2306.01953
Xuandong Zhao
Xuandong Zhao, Kexun Zhang, Zihao Su, Saastha Vasan, Ilya Grishchenko, Christopher Kruegel, Giovanni Vigna, Yu-Xiang Wang, Lei Li
Invisible Image Watermarks Are Provably Removable Using Generative AI
null
null
null
null
cs.CR cs.AI cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Invisible watermarks safeguard images' copyright by embedding hidden messages only detectable by owners. They also prevent people from misusing images, especially those generated by AI models. We propose a family of regeneration attacks to remove these invisible watermarks. The proposed attack method first adds random noise to an image to destroy the watermark and then reconstructs the image. This approach is flexible and can be instantiated with many existing image-denoising algorithms and pre-trained generative models such as diffusion models. Through formal proofs and empirical results, we show that all invisible watermarks are vulnerable to the proposed attack. For a particularly resilient watermark, RivaGAN, regeneration attacks remove 93-99% of the invisible watermarks while the baseline attacks remove no more than 3%. However, if we do not require the watermarked image to look the same as the original one, watermarks that keep the image semantically similar can be an alternative defense against our attack. Our finding underscores the need for a shift in research/industry emphasis from invisible watermarks to semantically similar ones. Code is available at https://github.com/XuandongZhao/WatermarkAttacker.
[ { "created": "Fri, 2 Jun 2023 23:29:28 GMT", "version": "v1" }, { "created": "Sun, 6 Aug 2023 17:17:04 GMT", "version": "v2" } ]
2023-08-08
[ [ "Zhao", "Xuandong", "" ], [ "Zhang", "Kexun", "" ], [ "Su", "Zihao", "" ], [ "Vasan", "Saastha", "" ], [ "Grishchenko", "Ilya", "" ], [ "Kruegel", "Christopher", "" ], [ "Vigna", "Giovanni", "" ], [ "Wang", "Yu-Xiang", "" ], [ "Li", "Lei", "" ] ]
Invisible watermarks safeguard images' copyright by embedding hidden messages only detectable by owners. They also prevent people from misusing images, especially those generated by AI models. We propose a family of regeneration attacks to remove these invisible watermarks. The proposed attack method first adds random noise to an image to destroy the watermark and then reconstructs the image. This approach is flexible and can be instantiated with many existing image-denoising algorithms and pre-trained generative models such as diffusion models. Through formal proofs and empirical results, we show that all invisible watermarks are vulnerable to the proposed attack. For a particularly resilient watermark, RivaGAN, regeneration attacks remove 93-99% of the invisible watermarks while the baseline attacks remove no more than 3%. However, if we do not require the watermarked image to look the same as the original one, watermarks that keep the image semantically similar can be an alternative defense against our attack. Our finding underscores the need for a shift in research/industry emphasis from invisible watermarks to semantically similar ones. Code is available at https://github.com/XuandongZhao/WatermarkAttacker.
2104.08894
Ahmed Abdelkader
Phillip Pope, Chen Zhu, Ahmed Abdelkader, Micah Goldblum, Tom Goldstein
The Intrinsic Dimension of Images and Its Impact on Learning
To appear at ICLR 2021 (spotlight), 17 pages with appendix, 15 figures
null
null
null
cs.CV cs.LG stat.ML
http://creativecommons.org/licenses/by-sa/4.0/
It is widely believed that natural image data exhibits low-dimensional structure despite the high dimensionality of conventional pixel representations. This idea underlies a common intuition for the remarkable success of deep learning in computer vision. In this work, we apply dimension estimation tools to popular datasets and investigate the role of low-dimensional structure in deep learning. We find that common natural image datasets indeed have very low intrinsic dimension relative to the high number of pixels in the images. Additionally, we find that low dimensional datasets are easier for neural networks to learn, and models solving these tasks generalize better from training to test data. Along the way, we develop a technique for validating our dimension estimation tools on synthetic data generated by GANs allowing us to actively manipulate the intrinsic dimension by controlling the image generation process. Code for our experiments may be found here https://github.com/ppope/dimensions.
[ { "created": "Sun, 18 Apr 2021 16:29:23 GMT", "version": "v1" } ]
2021-04-20
[ [ "Pope", "Phillip", "" ], [ "Zhu", "Chen", "" ], [ "Abdelkader", "Ahmed", "" ], [ "Goldblum", "Micah", "" ], [ "Goldstein", "Tom", "" ] ]
It is widely believed that natural image data exhibits low-dimensional structure despite the high dimensionality of conventional pixel representations. This idea underlies a common intuition for the remarkable success of deep learning in computer vision. In this work, we apply dimension estimation tools to popular datasets and investigate the role of low-dimensional structure in deep learning. We find that common natural image datasets indeed have very low intrinsic dimension relative to the high number of pixels in the images. Additionally, we find that low dimensional datasets are easier for neural networks to learn, and models solving these tasks generalize better from training to test data. Along the way, we develop a technique for validating our dimension estimation tools on synthetic data generated by GANs allowing us to actively manipulate the intrinsic dimension by controlling the image generation process. Code for our experiments may be found here https://github.com/ppope/dimensions.
1603.06679
Wenya Wang
Wenya Wang, Sinno Jialin Pan, Daniel Dahlmeier and Xiaokui Xiao
Recursive Neural Conditional Random Fields for Aspect-based Sentiment Analysis
null
null
null
null
cs.CL cs.IR cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In aspect-based sentiment analysis, extracting aspect terms along with the opinions being expressed from user-generated content is one of the most important subtasks. Previous studies have shown that exploiting connections between aspect and opinion terms is promising for this task. In this paper, we propose a novel joint model that integrates recursive neural networks and conditional random fields into a unified framework for explicit aspect and opinion terms co-extraction. The proposed model learns high-level discriminative features and double propagate information between aspect and opinion terms, simultaneously. Moreover, it is flexible to incorporate hand-crafted features into the proposed model to further boost its information extraction performance. Experimental results on the SemEval Challenge 2014 dataset show the superiority of our proposed model over several baseline methods as well as the winning systems of the challenge.
[ { "created": "Tue, 22 Mar 2016 05:59:00 GMT", "version": "v1" }, { "created": "Wed, 8 Jun 2016 06:24:06 GMT", "version": "v2" }, { "created": "Mon, 19 Sep 2016 14:00:43 GMT", "version": "v3" } ]
2016-09-20
[ [ "Wang", "Wenya", "" ], [ "Pan", "Sinno Jialin", "" ], [ "Dahlmeier", "Daniel", "" ], [ "Xiao", "Xiaokui", "" ] ]
In aspect-based sentiment analysis, extracting aspect terms along with the opinions being expressed from user-generated content is one of the most important subtasks. Previous studies have shown that exploiting connections between aspect and opinion terms is promising for this task. In this paper, we propose a novel joint model that integrates recursive neural networks and conditional random fields into a unified framework for explicit aspect and opinion terms co-extraction. The proposed model learns high-level discriminative features and double propagate information between aspect and opinion terms, simultaneously. Moreover, it is flexible to incorporate hand-crafted features into the proposed model to further boost its information extraction performance. Experimental results on the SemEval Challenge 2014 dataset show the superiority of our proposed model over several baseline methods as well as the winning systems of the challenge.
2105.10440
Prajwol Kumar Nakarmi
John Preu{\ss} Mattsson, Prajwol Kumar Nakarmi
Nori: Concealing the Concealed Identifier in 5G
9 pages, 8 figures, 1 table
2021
10.1145/3465481
ARES '21: Proceedings of the 16th International Conference on Availability, Reliability and Security
cs.CR
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
IMSI catchers have been a long standing and serious privacy problem in pre-5G mobile networks. To tackle this, 3GPP introduced the Subscription Concealed Identifier (SUCI) and other countermeasures in 5G. In this paper, we analyze the new SUCI mechanism and discover that it provides very poor anonymity when used with the variable length Network Specific Identifiers (NSI), which are part of the 5G standard. When applied to real-world name length data, we see that SUCI only provides 1-anonymity, meaning that individual subscribers can easily be identified and tracked. We strongly recommend 3GPP and GSMA to standardize and recommend the use of a padding mechanism for SUCI before variable length identifiers get more commonly used. We further show that the padding schemes, commonly used for network traffic, are not optimal for padding of identifiers based on real names. We propose a new improved padding scheme that achieves much less message expansion for a given $k$-anonymity.
[ { "created": "Fri, 21 May 2021 16:20:16 GMT", "version": "v1" }, { "created": "Mon, 14 Jun 2021 16:00:17 GMT", "version": "v2" } ]
2023-09-12
[ [ "Mattsson", "John Preuß", "" ], [ "Nakarmi", "Prajwol Kumar", "" ] ]
IMSI catchers have been a long standing and serious privacy problem in pre-5G mobile networks. To tackle this, 3GPP introduced the Subscription Concealed Identifier (SUCI) and other countermeasures in 5G. In this paper, we analyze the new SUCI mechanism and discover that it provides very poor anonymity when used with the variable length Network Specific Identifiers (NSI), which are part of the 5G standard. When applied to real-world name length data, we see that SUCI only provides 1-anonymity, meaning that individual subscribers can easily be identified and tracked. We strongly recommend 3GPP and GSMA to standardize and recommend the use of a padding mechanism for SUCI before variable length identifiers get more commonly used. We further show that the padding schemes, commonly used for network traffic, are not optimal for padding of identifiers based on real names. We propose a new improved padding scheme that achieves much less message expansion for a given $k$-anonymity.
1905.13340
Hsin-Po Wang
Hsin-Po Wang and Iwan Duursma
Log-logarithmic Time Pruned Polar Coding
13 pages, 13 figures; we extend arXiv:1812.08106 and remove "BEC" from title
null
10.1109/TIT.2020.3041523
null
cs.IT math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
A pruned variant of polar coding is proposed for binary erasure channels. For sufficiently small $\varepsilon>0$, we construct a series of capacity achieving codes with block length $N=\varepsilon^{-5}$, code rate $R=\text{Capacity}-\varepsilon$, error probability $P=\varepsilon$, and encoding and decoding time complexity $\text{bC}=O(\log\left|\log\varepsilon\right|)$ per information bit. The given per-bit complexity $\text{bC}$ is log-logarithmic in $N$, in $\text{Capacity}-R$, and in $P$; no known family of codes possesses this property. It is also the second lowest $\text{bC}$ after repeat-accumulate codes and their variants. While random codes and classical polar codes are the only two families of capacity-achieving codes whose $N$, $R$, $P$, and $\text{bC}$ were written down as explicit functions, our construction gives the third family. Then we generalize the result to: Fix a prime $q$ and fix a $q$-ary-input discrete symmetric memoryless channel. For sufficiently small $\varepsilon>0$, we construct a series of capacity achieving codes with block length $N=\varepsilon^{-O(1)}$, code rate $R=\text{Capacity}-\varepsilon$, error probability $P=\varepsilon$, and encoding and decoding time complexity $\text{bC}=O(\log\left|\log\varepsilon\right|)$ per information bit. The later construction gives the fastest family of capacity-achieving codes to date on those channels.
[ { "created": "Thu, 30 May 2019 22:37:41 GMT", "version": "v1" } ]
2020-12-14
[ [ "Wang", "Hsin-Po", "" ], [ "Duursma", "Iwan", "" ] ]
A pruned variant of polar coding is proposed for binary erasure channels. For sufficiently small $\varepsilon>0$, we construct a series of capacity achieving codes with block length $N=\varepsilon^{-5}$, code rate $R=\text{Capacity}-\varepsilon$, error probability $P=\varepsilon$, and encoding and decoding time complexity $\text{bC}=O(\log\left|\log\varepsilon\right|)$ per information bit. The given per-bit complexity $\text{bC}$ is log-logarithmic in $N$, in $\text{Capacity}-R$, and in $P$; no known family of codes possesses this property. It is also the second lowest $\text{bC}$ after repeat-accumulate codes and their variants. While random codes and classical polar codes are the only two families of capacity-achieving codes whose $N$, $R$, $P$, and $\text{bC}$ were written down as explicit functions, our construction gives the third family. Then we generalize the result to: Fix a prime $q$ and fix a $q$-ary-input discrete symmetric memoryless channel. For sufficiently small $\varepsilon>0$, we construct a series of capacity achieving codes with block length $N=\varepsilon^{-O(1)}$, code rate $R=\text{Capacity}-\varepsilon$, error probability $P=\varepsilon$, and encoding and decoding time complexity $\text{bC}=O(\log\left|\log\varepsilon\right|)$ per information bit. The later construction gives the fastest family of capacity-achieving codes to date on those channels.
2406.01080
Zhibo Xing
Zhibo Xing, Zijian Zhang, Zi'ang Zhang, Jiamou Liu, Liehuang Zhu, Giovanni Russello
No Vandalism: Privacy-Preserving and Byzantine-Robust Federated Learning
null
null
null
null
cs.CR cs.DC cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Federated learning allows several clients to train one machine learning model jointly without sharing private data, providing privacy protection. However, traditional federated learning is vulnerable to poisoning attacks, which can not only decrease the model performance, but also implant malicious backdoors. In addition, direct submission of local model parameters can also lead to the privacy leakage of the training dataset. In this paper, we aim to build a privacy-preserving and Byzantine-robust federated learning scheme to provide an environment with no vandalism (NoV) against attacks from malicious participants. Specifically, we construct a model filter for poisoned local models, protecting the global model from data and model poisoning attacks. This model filter combines zero-knowledge proofs to provide further privacy protection. Then, we adopt secret sharing to provide verifiable secure aggregation, removing malicious clients that disrupting the aggregation process. Our formal analysis proves that NoV can protect data privacy and weed out Byzantine attackers. Our experiments illustrate that NoV can effectively address data and model poisoning attacks, including PGD, and outperforms other related schemes.
[ { "created": "Mon, 3 Jun 2024 07:59:10 GMT", "version": "v1" } ]
2024-06-05
[ [ "Xing", "Zhibo", "" ], [ "Zhang", "Zijian", "" ], [ "Zhang", "Zi'ang", "" ], [ "Liu", "Jiamou", "" ], [ "Zhu", "Liehuang", "" ], [ "Russello", "Giovanni", "" ] ]
Federated learning allows several clients to train one machine learning model jointly without sharing private data, providing privacy protection. However, traditional federated learning is vulnerable to poisoning attacks, which can not only decrease the model performance, but also implant malicious backdoors. In addition, direct submission of local model parameters can also lead to the privacy leakage of the training dataset. In this paper, we aim to build a privacy-preserving and Byzantine-robust federated learning scheme to provide an environment with no vandalism (NoV) against attacks from malicious participants. Specifically, we construct a model filter for poisoned local models, protecting the global model from data and model poisoning attacks. This model filter combines zero-knowledge proofs to provide further privacy protection. Then, we adopt secret sharing to provide verifiable secure aggregation, removing malicious clients that disrupting the aggregation process. Our formal analysis proves that NoV can protect data privacy and weed out Byzantine attackers. Our experiments illustrate that NoV can effectively address data and model poisoning attacks, including PGD, and outperforms other related schemes.
1604.04586
Mouhacine Benosman
Mouhacine Benosman, Jeff Borggaard, Boris Kramer
Robust Reduced-Order Model Stabilization for Partial Differential Equations Based on Lyapunov Theory and Extremum Seeking with Application to the 3D Boussinesq Equations
arXiv admin note: text overlap with arXiv:1510.01728
null
null
null
cs.SY math.NA
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We present some results on stabilization for reduced-order models (ROMs) of partial differential equations. The stabilization is achieved using Lyapunov theory to design a new closure model that is robust to parametric uncertainties. The free parameters in the proposed ROM stabilization method are optimized using a model-free multi-parametric extremum seeking (MES) algorithm. The 3D Boussinesq equations provide a challenging numerical test-problem that is used to demonstrate the advantages of the proposed method.
[ { "created": "Fri, 15 Apr 2016 18:04:32 GMT", "version": "v1" } ]
2016-04-18
[ [ "Benosman", "Mouhacine", "" ], [ "Borggaard", "Jeff", "" ], [ "Kramer", "Boris", "" ] ]
We present some results on stabilization for reduced-order models (ROMs) of partial differential equations. The stabilization is achieved using Lyapunov theory to design a new closure model that is robust to parametric uncertainties. The free parameters in the proposed ROM stabilization method are optimized using a model-free multi-parametric extremum seeking (MES) algorithm. The 3D Boussinesq equations provide a challenging numerical test-problem that is used to demonstrate the advantages of the proposed method.
2205.10123
Stefano Teso
Andrea Bontempelli, Marcelo Rodas Britez, Xiaoyue Li, Haonan Zhao, Luca Erculiani, Stefano Teso, Andrea Passerini, Fausto Giunchiglia
Lifelong Personal Context Recognition
8 pages
null
null
null
cs.AI cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We focus on the development of AIs which live in lifelong symbiosis with a human. The key prerequisite for this task is that the AI understands - at any moment in time - the personal situational context that the human is in. We outline the key challenges that this task brings forth, namely (i) handling the human-like and ego-centric nature of the the user's context, necessary for understanding and providing useful suggestions, (ii) performing lifelong context recognition using machine learning in a way that is robust to change, and (iii) maintaining alignment between the AI's and human's representations of the world through continual bidirectional interaction. In this short paper, we summarize our recent attempts at tackling these challenges, discuss the lessons learned, and highlight directions of future research. The main take-away message is that pursuing this project requires research which lies at the intersection of knowledge representation and machine learning. Neither technology can achieve this goal without the other.
[ { "created": "Tue, 10 May 2022 13:24:47 GMT", "version": "v1" } ]
2022-05-23
[ [ "Bontempelli", "Andrea", "" ], [ "Britez", "Marcelo Rodas", "" ], [ "Li", "Xiaoyue", "" ], [ "Zhao", "Haonan", "" ], [ "Erculiani", "Luca", "" ], [ "Teso", "Stefano", "" ], [ "Passerini", "Andrea", "" ], [ "Giunchiglia", "Fausto", "" ] ]
We focus on the development of AIs which live in lifelong symbiosis with a human. The key prerequisite for this task is that the AI understands - at any moment in time - the personal situational context that the human is in. We outline the key challenges that this task brings forth, namely (i) handling the human-like and ego-centric nature of the the user's context, necessary for understanding and providing useful suggestions, (ii) performing lifelong context recognition using machine learning in a way that is robust to change, and (iii) maintaining alignment between the AI's and human's representations of the world through continual bidirectional interaction. In this short paper, we summarize our recent attempts at tackling these challenges, discuss the lessons learned, and highlight directions of future research. The main take-away message is that pursuing this project requires research which lies at the intersection of knowledge representation and machine learning. Neither technology can achieve this goal without the other.
2310.01882
Nick Brown
Nick Brown, Maurice Jamieson, Anton Lydike, Emilien Bauer, Tobias Grosser
Fortran performance optimisation and auto-parallelisation by leveraging MLIR-based domain specific abstractions in Flang
Author accepted version of paper in ACM Workshops of The International Conference on High Performance Computing, Network, Storage, and Analysis (SC-W 2023)
null
10.1145/3624062.3624167
null
cs.DC cs.PL
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
MLIR has become popular since it was open sourced in 2019. A sub-project of LLVM, the flexibility provided by MLIR to represent Intermediate Representations (IR) as dialects at different abstraction levels, to mix these, and to leverage transformations between dialects provides opportunities for automated program optimisation and parallelisation. In addition to general purpose compilers built upon MLIR, domain specific abstractions have also been developed. In this paper we explore complimenting the Flang MLIR general purpose compiler by combining with the domain specific Open Earth Compiler's MLIR stencil dialect. Developing transformations to discover and extracts stencils from Fortran, this specialisation delivers between a 2 and 10 times performance improvement for our benchmarks on a Cray supercomputer compared to using Flang alone. Furthermore, by leveraging existing MLIR transformations we develop an auto-parallelisation approach targeting multi-threaded and distributed memory parallelism, and optimised execution on GPUs, without any modifications to the serial Fortran source code.
[ { "created": "Tue, 3 Oct 2023 08:36:26 GMT", "version": "v1" } ]
2023-10-04
[ [ "Brown", "Nick", "" ], [ "Jamieson", "Maurice", "" ], [ "Lydike", "Anton", "" ], [ "Bauer", "Emilien", "" ], [ "Grosser", "Tobias", "" ] ]
MLIR has become popular since it was open sourced in 2019. A sub-project of LLVM, the flexibility provided by MLIR to represent Intermediate Representations (IR) as dialects at different abstraction levels, to mix these, and to leverage transformations between dialects provides opportunities for automated program optimisation and parallelisation. In addition to general purpose compilers built upon MLIR, domain specific abstractions have also been developed. In this paper we explore complimenting the Flang MLIR general purpose compiler by combining with the domain specific Open Earth Compiler's MLIR stencil dialect. Developing transformations to discover and extracts stencils from Fortran, this specialisation delivers between a 2 and 10 times performance improvement for our benchmarks on a Cray supercomputer compared to using Flang alone. Furthermore, by leveraging existing MLIR transformations we develop an auto-parallelisation approach targeting multi-threaded and distributed memory parallelism, and optimised execution on GPUs, without any modifications to the serial Fortran source code.
2309.11119
Minsu Kim
Minsu Kim, Giseop Kim, Kyong Hwan Jin, Sunwook Choi
BroadBEV: Collaborative LiDAR-camera Fusion for Broad-sighted Bird's Eye View Map Construction
null
null
null
null
cs.CV
http://creativecommons.org/licenses/by-nc-nd/4.0/
A recent sensor fusion in a Bird's Eye View (BEV) space has shown its utility in various tasks such as 3D detection, map segmentation, etc. However, the approach struggles with inaccurate camera BEV estimation, and a perception of distant areas due to the sparsity of LiDAR points. In this paper, we propose a broad BEV fusion (BroadBEV) that addresses the problems with a spatial synchronization approach of cross-modality. Our strategy aims to enhance camera BEV estimation for a broad-sighted perception while simultaneously improving the completion of LiDAR's sparsity in the entire BEV space. Toward that end, we devise Point-scattering that scatters LiDAR BEV distribution to camera depth distribution. The method boosts the learning of depth estimation of the camera branch and induces accurate location of dense camera features in BEV space. For an effective BEV fusion between the spatially synchronized features, we suggest ColFusion that applies self-attention weights of LiDAR and camera BEV features to each other. Our extensive experiments demonstrate that BroadBEV provides a broad-sighted BEV perception with remarkable performance gains.
[ { "created": "Wed, 20 Sep 2023 07:55:57 GMT", "version": "v1" }, { "created": "Thu, 21 Sep 2023 01:14:02 GMT", "version": "v2" }, { "created": "Mon, 25 Sep 2023 06:46:12 GMT", "version": "v3" }, { "created": "Wed, 8 Nov 2023 11:18:24 GMT", "version": "v4" } ]
2023-11-09
[ [ "Kim", "Minsu", "" ], [ "Kim", "Giseop", "" ], [ "Jin", "Kyong Hwan", "" ], [ "Choi", "Sunwook", "" ] ]
A recent sensor fusion in a Bird's Eye View (BEV) space has shown its utility in various tasks such as 3D detection, map segmentation, etc. However, the approach struggles with inaccurate camera BEV estimation, and a perception of distant areas due to the sparsity of LiDAR points. In this paper, we propose a broad BEV fusion (BroadBEV) that addresses the problems with a spatial synchronization approach of cross-modality. Our strategy aims to enhance camera BEV estimation for a broad-sighted perception while simultaneously improving the completion of LiDAR's sparsity in the entire BEV space. Toward that end, we devise Point-scattering that scatters LiDAR BEV distribution to camera depth distribution. The method boosts the learning of depth estimation of the camera branch and induces accurate location of dense camera features in BEV space. For an effective BEV fusion between the spatially synchronized features, we suggest ColFusion that applies self-attention weights of LiDAR and camera BEV features to each other. Our extensive experiments demonstrate that BroadBEV provides a broad-sighted BEV perception with remarkable performance gains.
1611.04145
Zijie Zheng
Zijie Zheng, Lingyang Song, Dusit Niyato, and Zhu Han
Resource Allocation in Wireless Powered Relay Networks: A Bargaining Game Approach
14 pages, 7 figures, journal paper
null
null
null
cs.IT cs.GT math.IT
http://creativecommons.org/licenses/by/4.0/
Simultaneously information and power transfer in mobile relay networks have recently emerged, where the relay can harvest the radio frequency (RF) energy and then use this energy for data forwarding and system operation. Most of the previous works do not consider that the relay may have its own objectives, such as using the harvested energy for its own transmission instead of maximizing transmission of the network. Therefore, in this paper, we propose a Nash bargaining approach to balance the information transmission efficiency of source-destination pairs and the harvested energy of the relay in a wireless powered relay network with multiple source-destination pairs and one relay. We analyze and prove that the Nash bargaining problem has several desirable properties such as the discreteness and quasi-concavity, when it is decomposed into three sub-problems: the energy transmission power optimization, the power control for data transmission and the time division between energy transmission and data transmission. Based on the theoretical analysis, we propose an alternating power control and time division algorithm to find a suboptimal solution. Simulation results clearly show and demonstrate the properties of the problem and the convergence of our algorithm.
[ { "created": "Sun, 13 Nov 2016 15:34:14 GMT", "version": "v1" } ]
2016-11-15
[ [ "Zheng", "Zijie", "" ], [ "Song", "Lingyang", "" ], [ "Niyato", "Dusit", "" ], [ "Han", "Zhu", "" ] ]
Simultaneously information and power transfer in mobile relay networks have recently emerged, where the relay can harvest the radio frequency (RF) energy and then use this energy for data forwarding and system operation. Most of the previous works do not consider that the relay may have its own objectives, such as using the harvested energy for its own transmission instead of maximizing transmission of the network. Therefore, in this paper, we propose a Nash bargaining approach to balance the information transmission efficiency of source-destination pairs and the harvested energy of the relay in a wireless powered relay network with multiple source-destination pairs and one relay. We analyze and prove that the Nash bargaining problem has several desirable properties such as the discreteness and quasi-concavity, when it is decomposed into three sub-problems: the energy transmission power optimization, the power control for data transmission and the time division between energy transmission and data transmission. Based on the theoretical analysis, we propose an alternating power control and time division algorithm to find a suboptimal solution. Simulation results clearly show and demonstrate the properties of the problem and the convergence of our algorithm.
2307.13900
Hyunjong Ok
Hyunjong Ok
FinTree: Financial Dataset Pretrain Transformer Encoder for Relation Extraction
4pages, 2 figures, The SIGIR'23 Workshop on Knowledge Discovery from Unstructured Data in Financial Services
null
null
null
cs.CL cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We present FinTree, Financial Dataset Pretrain Transformer Encoder for Relation Extraction. Utilizing an encoder language model, we further pretrain FinTree on the financial dataset, adapting the model in financial domain tasks. FinTree stands out with its novel structure that predicts a masked token instead of the conventional [CLS] token, inspired by the Pattern Exploiting Training methodology. This structure allows for more accurate relation predictions between two given entities. The model is trained with a unique input pattern to provide contextual and positional information about the entities of interest, and a post-processing step ensures accurate predictions in line with the entity types. Our experiments demonstrate that FinTree outperforms on the REFinD, a large-scale financial relation extraction dataset. The code and pretrained models are available at https://github.com/HJ-Ok/FinTree.
[ { "created": "Wed, 26 Jul 2023 01:48:52 GMT", "version": "v1" } ]
2023-07-27
[ [ "Ok", "Hyunjong", "" ] ]
We present FinTree, Financial Dataset Pretrain Transformer Encoder for Relation Extraction. Utilizing an encoder language model, we further pretrain FinTree on the financial dataset, adapting the model in financial domain tasks. FinTree stands out with its novel structure that predicts a masked token instead of the conventional [CLS] token, inspired by the Pattern Exploiting Training methodology. This structure allows for more accurate relation predictions between two given entities. The model is trained with a unique input pattern to provide contextual and positional information about the entities of interest, and a post-processing step ensures accurate predictions in line with the entity types. Our experiments demonstrate that FinTree outperforms on the REFinD, a large-scale financial relation extraction dataset. The code and pretrained models are available at https://github.com/HJ-Ok/FinTree.
2403.13249
Zhenyi Wang
Zhenyi Wang, Yan Li, Li Shen, Heng Huang
A Unified and General Framework for Continual Learning
ICLR 2024
null
null
null
cs.LG cs.AI cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Continual Learning (CL) focuses on learning from dynamic and changing data distributions while retaining previously acquired knowledge. Various methods have been developed to address the challenge of catastrophic forgetting, including regularization-based, Bayesian-based, and memory-replay-based techniques. However, these methods lack a unified framework and common terminology for describing their approaches. This research aims to bridge this gap by introducing a comprehensive and overarching framework that encompasses and reconciles these existing methodologies. Notably, this new framework is capable of encompassing established CL approaches as special instances within a unified and general optimization objective. An intriguing finding is that despite their diverse origins, these methods share common mathematical structures. This observation highlights the compatibility of these seemingly distinct techniques, revealing their interconnectedness through a shared underlying optimization objective. Moreover, the proposed general framework introduces an innovative concept called refresh learning, specifically designed to enhance the CL performance. This novel approach draws inspiration from neuroscience, where the human brain often sheds outdated information to improve the retention of crucial knowledge and facilitate the acquisition of new information. In essence, refresh learning operates by initially unlearning current data and subsequently relearning it. It serves as a versatile plug-in that seamlessly integrates with existing CL methods, offering an adaptable and effective enhancement to the learning process. Extensive experiments on CL benchmarks and theoretical analysis demonstrate the effectiveness of the proposed refresh learning. Code is available at \url{https://github.com/joey-wang123/CL-refresh-learning}.
[ { "created": "Wed, 20 Mar 2024 02:21:44 GMT", "version": "v1" } ]
2024-03-21
[ [ "Wang", "Zhenyi", "" ], [ "Li", "Yan", "" ], [ "Shen", "Li", "" ], [ "Huang", "Heng", "" ] ]
Continual Learning (CL) focuses on learning from dynamic and changing data distributions while retaining previously acquired knowledge. Various methods have been developed to address the challenge of catastrophic forgetting, including regularization-based, Bayesian-based, and memory-replay-based techniques. However, these methods lack a unified framework and common terminology for describing their approaches. This research aims to bridge this gap by introducing a comprehensive and overarching framework that encompasses and reconciles these existing methodologies. Notably, this new framework is capable of encompassing established CL approaches as special instances within a unified and general optimization objective. An intriguing finding is that despite their diverse origins, these methods share common mathematical structures. This observation highlights the compatibility of these seemingly distinct techniques, revealing their interconnectedness through a shared underlying optimization objective. Moreover, the proposed general framework introduces an innovative concept called refresh learning, specifically designed to enhance the CL performance. This novel approach draws inspiration from neuroscience, where the human brain often sheds outdated information to improve the retention of crucial knowledge and facilitate the acquisition of new information. In essence, refresh learning operates by initially unlearning current data and subsequently relearning it. It serves as a versatile plug-in that seamlessly integrates with existing CL methods, offering an adaptable and effective enhancement to the learning process. Extensive experiments on CL benchmarks and theoretical analysis demonstrate the effectiveness of the proposed refresh learning. Code is available at \url{https://github.com/joey-wang123/CL-refresh-learning}.
2012.06354
Alexander Ziller
Alexander Ziller, Jonathan Passerat-Palmbach, Th\'eo Ryffel, Dmitrii Usynin, Andrew Trask, Ion\'esio Da Lima Costa Junior, Jason Mancuso, Marcus Makowski, Daniel Rueckert, Rickmer Braren, Georgios Kaissis
Privacy-preserving medical image analysis
Accepted at the workshop for Medical Imaging meets NeurIPS, 34th Conference on Neural Information Processing Systems (NeurIPS) December 11, 2020
null
null
null
cs.CR cs.CV cs.LG
http://creativecommons.org/licenses/by/4.0/
The utilisation of artificial intelligence in medicine and healthcare has led to successful clinical applications in several domains. The conflict between data usage and privacy protection requirements in such systems must be resolved for optimal results as well as ethical and legal compliance. This calls for innovative solutions such as privacy-preserving machine learning (PPML). We present PriMIA (Privacy-preserving Medical Image Analysis), a software framework designed for PPML in medical imaging. In a real-life case study we demonstrate significantly better classification performance of a securely aggregated federated learning model compared to human experts on unseen datasets. Furthermore, we show an inference-as-a-service scenario for end-to-end encrypted diagnosis, where neither the data nor the model are revealed. Lastly, we empirically evaluate the framework's security against a gradient-based model inversion attack and demonstrate that no usable information can be recovered from the model.
[ { "created": "Thu, 10 Dec 2020 13:56:00 GMT", "version": "v1" } ]
2020-12-14
[ [ "Ziller", "Alexander", "" ], [ "Passerat-Palmbach", "Jonathan", "" ], [ "Ryffel", "Théo", "" ], [ "Usynin", "Dmitrii", "" ], [ "Trask", "Andrew", "" ], [ "Junior", "Ionésio Da Lima Costa", "" ], [ "Mancuso", "Jason", "" ], [ "Makowski", "Marcus", "" ], [ "Rueckert", "Daniel", "" ], [ "Braren", "Rickmer", "" ], [ "Kaissis", "Georgios", "" ] ]
The utilisation of artificial intelligence in medicine and healthcare has led to successful clinical applications in several domains. The conflict between data usage and privacy protection requirements in such systems must be resolved for optimal results as well as ethical and legal compliance. This calls for innovative solutions such as privacy-preserving machine learning (PPML). We present PriMIA (Privacy-preserving Medical Image Analysis), a software framework designed for PPML in medical imaging. In a real-life case study we demonstrate significantly better classification performance of a securely aggregated federated learning model compared to human experts on unseen datasets. Furthermore, we show an inference-as-a-service scenario for end-to-end encrypted diagnosis, where neither the data nor the model are revealed. Lastly, we empirically evaluate the framework's security against a gradient-based model inversion attack and demonstrate that no usable information can be recovered from the model.
2205.13741
Ali Seyfi
Ali Seyfi, Jean-Francois Rajotte, Raymond T. Ng
Generating multivariate time series with COmmon Source CoordInated GAN (COSCI-GAN)
19 pages, 16 figures
null
null
null
cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Generating multivariate time series is a promising approach for sharing sensitive data in many medical, financial, and IoT applications. A common type of multivariate time series originates from a single source such as the biometric measurements from a medical patient. This leads to complex dynamical patterns between individual time series that are hard to learn by typical generation models such as GANs. There is valuable information in those patterns that machine learning models can use to better classify, predict or perform other downstream tasks. We propose a novel framework that takes time series' common origin into account and favors channel/feature relationships preservation. The two key points of our method are: 1) the individual time series are generated from a common point in latent space and 2) a central discriminator favors the preservation of inter-channel/feature dynamics. We demonstrate empirically that our method helps preserve channel/feature correlations and that our synthetic data performs very well in downstream tasks with medical and financial data.
[ { "created": "Fri, 27 May 2022 03:09:55 GMT", "version": "v1" }, { "created": "Thu, 15 Dec 2022 00:19:16 GMT", "version": "v2" } ]
2022-12-16
[ [ "Seyfi", "Ali", "" ], [ "Rajotte", "Jean-Francois", "" ], [ "Ng", "Raymond T.", "" ] ]
Generating multivariate time series is a promising approach for sharing sensitive data in many medical, financial, and IoT applications. A common type of multivariate time series originates from a single source such as the biometric measurements from a medical patient. This leads to complex dynamical patterns between individual time series that are hard to learn by typical generation models such as GANs. There is valuable information in those patterns that machine learning models can use to better classify, predict or perform other downstream tasks. We propose a novel framework that takes time series' common origin into account and favors channel/feature relationships preservation. The two key points of our method are: 1) the individual time series are generated from a common point in latent space and 2) a central discriminator favors the preservation of inter-channel/feature dynamics. We demonstrate empirically that our method helps preserve channel/feature correlations and that our synthetic data performs very well in downstream tasks with medical and financial data.
1401.4802
J\"urgen M\"unch
Alexis Ocampo, J\"urgen M\"unch
Process Evolution Supported by Rationale: An Empirical Investigation of Process Changes
8 pages. The final publication is available at http://link.springer.com/chapter/10.1007%2F11754305_36
Software Process Change, volume 3966 of Lecture Notes in Computer Science, pages 334-341, Springer Berlin Heidelberg, 2006
10.1007/11754305_36
null
cs.SE
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Evolving a software process model without a retrospective and, in consequence, without an understanding of the process evolution, can lead to severe problems for the software development organization, e.g., inefficient performance as a consequence of the arbitrary introduction of changes or difficulty in demonstrating compliance to a given standard. Capturing information on the rationale behind changes can provide a means for better understanding process evolution. This article presents the results of an exploratory study with the goal of understanding the nature of process changes in a given context. It presents the most important issues that motivated process engineers changing important aerospace software process standards during an industrial project. The study is part of research work intended to incrementally define a systematic mechanism for process evolution supported by rationale information.
[ { "created": "Mon, 20 Jan 2014 06:41:59 GMT", "version": "v1" } ]
2014-01-21
[ [ "Ocampo", "Alexis", "" ], [ "Münch", "Jürgen", "" ] ]
Evolving a software process model without a retrospective and, in consequence, without an understanding of the process evolution, can lead to severe problems for the software development organization, e.g., inefficient performance as a consequence of the arbitrary introduction of changes or difficulty in demonstrating compliance to a given standard. Capturing information on the rationale behind changes can provide a means for better understanding process evolution. This article presents the results of an exploratory study with the goal of understanding the nature of process changes in a given context. It presents the most important issues that motivated process engineers changing important aerospace software process standards during an industrial project. The study is part of research work intended to incrementally define a systematic mechanism for process evolution supported by rationale information.
2102.05802
Leighton Barnes
Leighton Pate Barnes and Ayfer Ozgur
Fisher Information and Mutual Information Constraints
null
null
null
null
cs.IT math.IT math.ST stat.TH
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We consider the processing of statistical samples $X\sim P_\theta$ by a channel $p(y|x)$, and characterize how the statistical information from the samples for estimating the parameter $\theta\in\mathbb{R}^d$ can scale with the mutual information or capacity of the channel. We show that if the statistical model has a sub-Gaussian score function, then the trace of the Fisher information matrix for estimating $\theta$ from $Y$ can scale at most linearly with the mutual information between $X$ and $Y$. We apply this result to obtain minimax lower bounds in distributed statistical estimation problems, and obtain a tight preconstant for Gaussian mean estimation. We then show how our Fisher information bound can also imply mutual information or Jensen-Shannon divergence based distributed strong data processing inequalities.
[ { "created": "Thu, 11 Feb 2021 01:53:09 GMT", "version": "v1" }, { "created": "Thu, 8 Jul 2021 22:28:26 GMT", "version": "v2" } ]
2021-07-12
[ [ "Barnes", "Leighton Pate", "" ], [ "Ozgur", "Ayfer", "" ] ]
We consider the processing of statistical samples $X\sim P_\theta$ by a channel $p(y|x)$, and characterize how the statistical information from the samples for estimating the parameter $\theta\in\mathbb{R}^d$ can scale with the mutual information or capacity of the channel. We show that if the statistical model has a sub-Gaussian score function, then the trace of the Fisher information matrix for estimating $\theta$ from $Y$ can scale at most linearly with the mutual information between $X$ and $Y$. We apply this result to obtain minimax lower bounds in distributed statistical estimation problems, and obtain a tight preconstant for Gaussian mean estimation. We then show how our Fisher information bound can also imply mutual information or Jensen-Shannon divergence based distributed strong data processing inequalities.