检索结果-内蒙古大学图书馆

Proceedings of the 38th International Conference on Neural Information Processing Systems

作者： Chenhao Zhou Zebang Shen Chao Zhang Hanbin Zhao Hui Qian College of Computer Science and Technology Zhejiang University Department of Computer Science ETH Zurich College of Computer Science and Technology Zhejiang University and State Key Lab of CAD & CG Zhejiang University

ISBN: (纸本)9798331314385

In this paper, we propose a provably efficient natural policy gradient algorithm called Spectral Dynamic Embedding Policy Optimization (SDEPO) for two-player zero-sum stochastic Markov games with continuous state space and finite action space. In the policy evaluation procedure of our algorithm, a novel kernel embedding method is employed to construct a finite-dimensional linear approximations to the state-action value function. We explicitly analyze the approximation error in policy evaluation, and show that SDEPO achieves an Õ(1/(1-γ)3ε) last-iterate convergence to the ε-optimal Nash equilibrium, which is independent of the cardinality of the state space. The complexity result matches the best-known results for global convergence of policy gradient algorithms for single agent setting. Moreover, we also propose a practical variant of SDEPO to deal with continuous action space and empirical results demonstrate the practical superiority of the proposed method.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Computable one-way functions on the reals

arXiv

引用

arXiv 2024年

作者： Barmpalias, George Zhang, Xiaoyan State Key Lab of Computer Science Institute of Software Chinese Academy of Sciences Beijing China

A major open problem in computational complexity is the existence of a one-way function, namely a function from strings to strings which is computationally easy to compute but hard to invert. Levin (2023) formulated the notion of one-way functions from reals (infinite bit-sequences) to reals in terms of computability, and asked whether partial computable one-way functions exist. We give a strong positive answer using the hardness of the halting problem and exhibiting a total computable one-way function. Copyright © 2024, The Authors. All rights reserved.

关键词： Computational complexity

来源：评论

学校读者我要写书评

暂无评论

Unleashing Unlabeled Data: A Paradigm for Cross-View Geo-Localization

Unleashing Unlabeled Data: A Paradigm for Cross-View Geo-Loc...

引用

Conference on computer Vision and Pattern Recognition (CVPR)

作者： Guopeng Li Ming Qian Gui-Song Xia School of Computer Science Wuhan University State Key Lab. LIESMARS Wuhan University

ISBN: (数字)9798350353006

ISBN: (纸本)9798350353013

This paper investigates the effective utilization of unlabeled data for large-area cross-view gee-localization (CVGL), encompassing both unsupervised and semi-supervised settings. Common approaches to CVGL rely on ground-satellite image pairs and employ label-driven supervised training. However, the cost of collecting precise cross-view image pairs hinders the deployment of CVGL in real-life scenarios. Without the pairs, CVGL will be more challenging to handle the significant imaging and spatial gaps between ground and satellite images. To this end, we propose an unsupervised framework including a cross-view projection to guide the model for retrieving initial pseudo-labels and a fast re-ranking mechanism to refine the pseudo-labels by leveraging the fact that “the perfectly paired ground-satellite image is located in a unique and identical scene”. The framework exhibits competitive performance compared with supervised works on three open-source benchmarks. Our code and models will be released on https://***/liguopeng0923/UCVGL.

关键词： Training computer vision Costs Codes Imaging Benchmark testing Satellite images

来源：评论

学校读者我要写书评

暂无评论

Dimensionality and randomness

arXiv

引用

arXiv 2024年

作者： Barmpalias, George Zhang, Xiaoyan State Key Lab of Computer Science Institute of Software Chinese Academy of Sciences Beijing China

Arranging the bits of a random string or real into k columns of a two-dimensional array or higher dimensional structure is typically accompanied with loss in the Kolmogorov complexity of the columns, which depends on k. We quantify and characterize this phenomenon for arrays and trees and its relationship to negligible classes. Copyright © 2024, The Authors. All rights reserved.

关键词： Computational complexity

来源：评论

学校读者我要写书评

暂无评论

Unleashing Unlabeled Data: A Paradigm for Cross-View Geo-Localization

arXiv

引用

arXiv 2024年

作者： Li, Guopeng Qian, Ming Xia, Gui-Song School of Computer Science China State Key Lab. LIESMARS Wuhan University China

This paper investigates the effective utilization of unlabeled data for large-area cross-view geo-localization (CVGL), encompassing both unsupervised and semi-supervised settings. Common approaches to CVGL rely on ground-satellite image pairs and employ label-driven supervised training. However, the cost of collecting precise cross-view image pairs hinders the deployment of CVGL in real-life scenarios. Without the pairs, CVGL will be more challenging to handle the significant imaging and spatial gaps between ground and satellite images. To this end, we propose an unsupervised framework including a cross-view projection to guide the model for retrieving initial pseudo-labels and a fast re-ranking mechanism to refine the pseudo-labels by leveraging the fact that "the perfectly paired ground-satellite image is located in a unique and identical scene". The framework exhibits competitive performance compared with supervised works on three open-source benchmarks. Our code and models will be released on https://***/liguopeng0923/UCVGL. Copyright © 2024, The Authors. All rights reserved.

关键词： Benchmarking

来源：评论

学校读者我要写书评

暂无评论

Transformer in transformer 21

Transformer in transformer

引用

Proceedings of the 35th International Conference on Neural Information Processing Systems

作者： Kai Han An Xiao Enhua Wu Jianyuan Guo Chunjing Xu Yunhe Wang State Key Lab of Computer Science ISCAS & UCAS and Huawei Noah's Ark Lab Huawei Noah's Ark Lab State Key Lab of Computer Science ISCAS & UCAS and University of Macau

ISBN: (纸本)9781713845393

Transformer is a new kind of neural architecture which encodes the input data as powerful features via the attention mechanism. Basically, the visual transformers first divide the input images into several local patches and then calculate both representations and their relationship. Since natural images are of high complexity with abundant detail and color information, the granularity of the patch dividing is not fine enough for excavating features of objects in different scales and locations. In this paper, we point out that the attention inside these local patches are also essential for building visual transformers with high performance and we explore a new architecture, namely, Transformer iN Transformer (TNT). Specifically, we regard the local patches (e.g., 16×16) as "visual sentences" and present to further divide them into smaller patches (e.g., 4×4) as "visual words". The attention of each word will be calculated with other words in the given visual sentence with negligible computational costs. Features of both words and sentences will be aggregated to enhance the representation ability. Experiments on several benchmarks demonstrate the effectiveness of the proposed TNT architecture, e.g., we achieve an 81.5% top-1 accuracy on the ImageNet, which is about 1.7% higher than that of the state-of-the-art visual transformer with similar computational cost. The PyTorch code is available at https://***/huawei-noah/CV-Backbones, and the MindSpore code is available at https://***/mindspore/models/tree/master/research/cv/TNT.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Energy consumption forecasting for laser manufacturing of large artifacts based on fusionable transfer learning

引用

Visual Computing for Industry,Biomedicine,and Art 2024年第1期7卷 19-32页

作者： Linxuan Wang Jinghua Xu Shuyou Zhang Jianrong Tan Shaomei Fei Xuezhi Shi Jihong Pang Sheng Luo State Key Lab of Fluid Power and Mechatronic Systems Zhejiang UniversityHangzhouZhejiang 310058China Engineering Research Center for Design Engineering and Digital Twin of Zhejiang Province Zhejiang UniversityHangzhouZhejiang 310058China State Key Lab of Materials Processing and Die&Mould Technology Huazhong University of Science and TechnologyWuhanHubei 430074China School of Marine Engineering Equipment Zhejiang Ocean UniversityZhoushanZhejiang 316022China College of Business Shaoxing UniversityShaoxingZhejiang 312000China College of Computer Science and Artificial Intelligence Wenzhou UniversityWenzhouZhejiang 325000China

This study presents an energy consumption(EC)forecasting method for laser melting manufacturing of metal artifacts based on fusionable transfer learning(FTL).To predict the EC of manufacturing products,particularly from scale-down to scale-up,a general paradigm was first developed by categorizing the overall process into three main *** operating electrical power was further formulated as a combinatorial function,based on which an operator learning network was adopted to fit the nonlinear relations between the fabricating arguments and ***-arranged networks were constructed to investigate the impacts of fabrication variables and devices on *** the interconnections among these factors,the outputs of the neural networks were blended and fused to jointly predict the electrical *** innovatively,large artifacts can be decomposed into timedependent laser-scanning trajectories,which can be further transformed into fusionable information via neural networks,inspired by large language ***,transfer learning can deal with either scale-down or scale-up forecasting,namely,FTL with scalability within artifact *** effectiveness of the proposed FTL was verified through physical fabrication experiments via laser powder bed *** relative error of the average and overall EC predictions based on FTL was maintained below 0.83%.The melting fusion quality was examined using metallographic *** proposed FTL framework can forecast the EC of scaled structures,which is particularly helpful in price estimation and quotation of large metal products towards carbon peaking and carbon neutrality.

关键词： Energy consumption forecasting Large metal artifacts Carbon peaking and carbon neutrality Laser powder bed fusion Fusionable transfer learning

来源：评论

学校读者我要写书评

暂无评论

Make It Easy: Action Quality Assessment of Cyborg Animals Based on Spatial-Temporal Pose Inference 10

Make It Easy: Action Quality Assessment of Cyborg Animals Ba...

引用

10th IEEE Smart World Congress, SWC 2024

作者： Li, Qiqi Han, Le Wang, Pengfei Zheng, Nenggan Zhejiang University Qiushi Academy for Advanced Studies College of Computer Science and Technology Hangzhou China Zhejiang University School of Software Technology Hangzhou China Zhejiang University Qiushi Academy for Advanced Studies The State Key Lab of Brain-Machine Intelligence College of Computer Science and Technology Hangzhou China Bengbu University Bengbu China

ISBN: (纸本)9798331520861

Assessing the action quality of cyborg animals helps to adjust control strategies, guide the development of control algorithm, and enhance the efficiency of navigation and military applications. However, existing research is limited in control algorithm and ignores the assessment of movement smoothness. In this paper, we first propose EASY, an effective assessment framework, which is able to (i) evaluate the action quality of cyborg animals of various sizes and motion patterns;(ii) obtain assessment results directly based on keypoints. Specifically, EASY first captures keypoints sequences through a small object pose estimation module. It employs the convolutional block attention module (CBAM) and depthwise over parameterized convolution (DOConv), allowing more accurate feature recognition with fewer parameters. Then, EASY employs multi-field fusion strategy that merges commonalities and differences, enabling a comprehensive and accurate assessment of fluency. We tested EASY on both CyborgRat-MFA dataset and CyborgBee-MFA dataset. Extensive experiments demonstrate that the assessment results are consistent with human expert scores, which prove that our proposed EASY is reasonable, effective, and interpretable. © 2024 IEEE.

关键词： action quality assessment cyborg animals electrical stimulation control motion fluency

来源：评论

学校读者我要写书评

暂无评论

Efficient Model-Based OPC via Graph Neural Network

Efficient Model-Based OPC via Graph Neural Network

引用

2023 International Symposium of Electronics Design Automation, ISEDA 2023

作者： Sun, Shuyuan Chen, Xuelian Yang, Fan Yu, Bei Li, Shang Zeng, Xuan School of Microelectronics Fudan University State Key Lab of ASIC and System China Cogenda Inc China The Chinese University of Hong Kong Department of Computer Science and Engineering Hong Kong

ISBN: (纸本)9798350304510

As feature size continues to shrink and light source wavelengths remain unchanged, the optical diffraction effects seriously degrade chip yield. Optical proximity correction (OPC) has become an essential step for chip manufacturability. However, OPC cannot achieve satisfactory mask correction results in an affordable number of iterations on some layouts, delaying the chip design cycle. In this paper, we propose to use the Graph Neural Network (GNN) to predict the initial mask correction and speed up model-based OPC. We design a graph model to represent the chip layout, with segments represented as graph nodes and the diffraction effects between them represented as graph edges. The GNN is used to obtain the embedding of the segment, and we predict the shift value based on its output. The proposed GNN-based predictor enhances the original OPC to find a good correction solution for full-chip-size masks in fewer iterations. compared with the original OPC method, experimental results on 55nm and 32nm full chip designs show that the proposed GNN-based OPC method reduces the number of iterations up to 75% and the acquired mask of comparable quality. © 2023 IEEE.

关键词： Diffraction

来源：评论

学校读者我要写书评

暂无评论

Comprehensive Semantic Repair of Obsolete GUI Test Scripts for Mobile Applications 24

Comprehensive Semantic Repair of Obsolete GUI Test Scripts f...

引用

44th ACM/IEEE International Conference on Software Engineering, ICSE 2024

作者： Cao, Shaoheng Pan, Minxue Pei, Yu Yang, Wenhua Zhang, Tian Wang, Linzhang Li, Xuandong Nanjing University State Key Lab for Novel Software Technology Nanjing China The Hong Kong Polytechnic University Department of Computing Hong Kong College of Computer Science and Technology Nanjing University of Aeronautics and Astronautics Nanjing China

ISBN: (纸本)9798400702174

Graphical User Interface (GUI) testing is one of the primary approaches for testing mobile apps. Test scripts serve as the main carrier of GUI testing, yet they are prone to obsolescence when the GUIs change with the apps' evolution. Existing repair approaches based on GUI layouts or images prove effective when the GUI changes between the base and updated versions are minor, however, they may struggle with substantial changes. In this paper, a novel approach named COSER is introduced as a solution to re-pairing broken scripts, which is capable of addressing larger GUI changes compared to existing methods. COSER incorporates both external semantic information from the GUI elements and internal semantic information from the source code to provide a unique and comprehensive solution. The efficacy of COSER was demonstrated through experiments conducted on 20 Android apps, resulting in superior performance when compared to the state-of-the-art tools METER and GUIDER. In addition, a tool that implements the COSER approach is available for practical use and future research. © 2024 ACM.

关键词： Graphical user interfaces

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：