摘要 :
Program synthesis is now a reality, and we are approaching the point where domain-specific synthesizers can now handle problems of practical sizes. Moreover, some of these tools are finding adoption in industry. However, for synth...
展开
Program synthesis is now a reality, and we are approaching the point where domain-specific synthesizers can now handle problems of practical sizes. Moreover, some of these tools are finding adoption in industry. However, for synthesis to become a mainstream technique adopted at large by programmers as well as by end-users, we need to design programmable synthesis frameworks that (i) are not tailored to specific domains or languages, (ii) enable one to specify synthesis problems with a variety of qualitative and quantitative objectives in mind, and (iii) come equipped with theoretical as well as practical guarantees. We report on our work on designing such frameworks and on building synthesis engines that can handle program-synthesis problems describ-able in such frameworks, and describe open challenges and opportunities.
收起
摘要 :
Program synthesis is now a reality, and we are approaching the point where domain-specific synthesizers can now handle problems of practical sizes. Moreover, some of these tools are finding adoption in industry. However, for synth...
展开
Program synthesis is now a reality, and we are approaching the point where domain-specific synthesizers can now handle problems of practical sizes. Moreover, some of these tools are finding adoption in industry. However, for synthesis to become a mainstream technique adopted at large by programmers as well as by end-users, we need to design programmable synthesis frameworks that (ⅰ) are not tailored to specific domains or languages, (ⅱ) enable one to specify synthesis problems with a variety of qualitative and quantitative objectives in mind, and (ⅲ) come equipped with theoretical as well as practical guarantees. We report on our work on designing such frameworks and on building synthesis engines that can handle program-synthesis problems describ-able in such frameworks, and describe open challenges and opportunities.
收起
摘要 :
Many problems in computer vision and machine learning can be cast as learning on hypergraphs that represent higher-order relations. Recent approaches for hypergraph learning extend graph neural networks based on message passing, w...
展开
Many problems in computer vision and machine learning can be cast as learning on hypergraphs that represent higher-order relations. Recent approaches for hypergraph learning extend graph neural networks based on message passing, which is simple yet fundamentally limited in modeling long-range dependencies and expressive power. On the other hand, tensor-based equivariant neural networks enjoy maximal expressiveness, but their application has been limited in hypergraphs due to heavy computation and strict assumptions on fixed-order hyperedges. We resolve these problems and present Equivariant Hypergraph Neural Network (EHNN), the first attempt to realize maximally expressive equivariant layers for general hypergraph learning. We also present two practical realizations of our framework based on hypernetworks (EHNN-MLP) and self-attention (EHNN-Transformer), which are easy to implement and theoretically more expressive than most message passing approaches. We demonstrate their capability in a range of hypergraph learning problems, including synthetic k-edge identification, semi-supervised classification, and visual keypoint matching, and report improved performances over strong message passing baselines.
收起
摘要 :
Incorporating spatio-temporal human visual perception into video quality assessment (VQA) remains a formidable issue. Previous statistical or computational models of spatio-temporal perception have limitations to be applied to the...
展开
Incorporating spatio-temporal human visual perception into video quality assessment (VQA) remains a formidable issue. Previous statistical or computational models of spatio-temporal perception have limitations to be applied to the general VQA algorithms. In this paper, we propose a novel full-reference (FR) VQA framework named Deep Video Quality Assessor (DeepVQA) to quantify the spatio-temporal visual perception via a convolutional neural network (CNN) and a convolutional neural aggregation network (CNAN). Our framework enables to figure out the spatio-temporal sensitivity behavior through learning in accordance with the subjective score. In addition, to manipulate the temporal variation of distortions, we propose a novel temporal pooling method using an attention model. In the experiment, we show DeepVQA remarkably achieves the state-of-the-art prediction accuracy of more than 0.9 correlation, which is ~5% higher than those of conventional methods on the LIVE and CSIQ video databases.
收起
摘要 :
Incorporating spatio-temporal human visual perception into video quality assessment (VQA) remains a formidable issue. Previous statistical or computational models of spatio-temporal perception have limitations to be applied to the...
展开
Incorporating spatio-temporal human visual perception into video quality assessment (VQA) remains a formidable issue. Previous statistical or computational models of spatio-temporal perception have limitations to be applied to the general VQA algorithms. In this paper, we propose a novel full-reference (FR) VQA framework named Deep Video Quality Assessor (DeepVQA) to quantify the spatio-temporal visual perception via a convolutional neural network (CNN) and a convolutional neural aggregation network (CNAN). Our framework enables to figure out the spatio-temporal sensitivity behavior through learning in accordance with the subjective score. In addition, to manipulate the temporal variation of distortions, we propose a novel temporal pooling method using an attention model. In the experiment, we show DeepVQA remarkably achieves the state-of-the-art prediction accuracy of more than 0.9 correlation, which is ~5% higher than those of conventional methods on the LIVE and CSIQ video databases.
收起
摘要 :
We introduce a new privacy issue on Facebook. We were motivated by the Facebook's search option, which exposes a user profile with his or her phone number. Based on this search option, we developed a method to automatically collec...
展开
We introduce a new privacy issue on Facebook. We were motivated by the Facebook's search option, which exposes a user profile with his or her phone number. Based on this search option, we developed a method to automatically collect Facebook users' personal data (e.g., phone number, location and birthday) by enumerating the possibly almost entire phone number range for the target area. To show the feasibility, we launched attacks for targeting the users who live in two specific regions (United States and South Korea) by mimicking real users' search activities with three sybil accounts. Despite Facebook's best efforts to stop such attempts from crawling users' data with several security practices, 214,705 phone numbers were successfully tested and 25,518 actual users' personal data were obtained within 15 days in California, United States; 215,679 phone numbers were also tested and 56,564 actual users' personal data were obtained in South Korea. To prevent such attacks, we recommend several practical defense mechanisms.
收起
摘要 :
We introduce a new privacy issue on Facebook. We were motivated by the Facebook's search option, which exposes a user profile with his or her phone number. Based on this search option, we developed a method to automatically collec...
展开
We introduce a new privacy issue on Facebook. We were motivated by the Facebook's search option, which exposes a user profile with his or her phone number. Based on this search option, we developed a method to automatically collect Facebook users' personal data (e.g., phone number, location and birthday) by enumerating the possibly almost entire phone number range for the target area. To show the feasibility, we launched attacks for targeting the users who live in two specific regions (United States and South Korea) by mimicking real users' search activities with three sybil accounts. Despite Facebook's best efforts to stop such attempts from crawling users' data with several security practices, 214,705 phone numbers were successfully tested and 25,518 actual users' personal data were obtained within 15 days in California, United States; 215,679 phone numbers were also tested and 56,564 actual users' personal data were obtained in South Korea. To prevent such attacks, we recommend several practical defense mechanisms.
收起
摘要 :
Graphical representations are used heavily in systems analysis and design without proper verification of their usability. However, different representations have varying impacts on the effectiveness of systems analysts. The object...
展开
Graphical representations are used heavily in systems analysis and design without proper verification of their usability. However, different representations have varying impacts on the effectiveness of systems analysts. The objective of this study was to explore the effects of diagrammatic representations on task behavior and performance. A cognitive model of an analysis and design task using diagrams was proposed, from which hypotheses about the effects of diagrammatic representation on task behavior and performance were derived. A laboratory experiment was conducted to test the hypotheses. Results of the experiment show that the representational features of the diagrams induced subjects to adopt different problem solving strategies, which resulted in different task performances.
收起
摘要 :
This study proposes a novel module in the convolutional neural networks (CNN) framework named permutation layer. With the new layer, we are particularly targeting time-series tasks where 2-dimensional CNN kernel loses its ability ...
展开
This study proposes a novel module in the convolutional neural networks (CNN) framework named permutation layer. With the new layer, we are particularly targeting time-series tasks where 2-dimensional CNN kernel loses its ability to capture the spatially co-related features. Multivariate time-series analysis consists of stacked input channels without considering the order of the channels resulting in an unsorted "2D-image". 2D convolution kernels are not efficient at capturing features from these distorted as the time-series lacks spatial information between the sensor channels. To overcome this weakness, we propose learnable permutation layers as an extension of vanilla convolution layers which allow to interchange different sensor channels such that sensor channels with similar information content are brought together to enable a more effective 2D convolution operation. We test the approach on a benchmark time-series classification task and report the superior performance and applicability of the proposed method.
收起
摘要 :
This study proposes a novel module in the convolutional neural networks (CNN) framework named permutation layer. With the new layer, we are particularly targeting time-series tasks where 2-dimensional CNN kernel loses its ability ...
展开
This study proposes a novel module in the convolutional neural networks (CNN) framework named permutation layer. With the new layer, we are particularly targeting time-series tasks where 2-dimensional CNN kernel loses its ability to capture the spatially co-related features. Multivariate time-series analysis consists of stacked input channels without considering the order of the channels resulting in an unsorted "2D-image". 2D convolution kernels are not efficient at capturing features from these distorted as the time-series lacks spatial information between the sensor channels. To overcome this weakness, we propose learnable permutation layers as an extension of vanilla convolution layers which allow to interchange different sensor channels such that sensor channels with similar information content are brought together to enable a more effective 2D convolution operation. We test the approach on a benchmark time-series classification task and report the superior performance and applicability of the proposed method.
收起