检索结果-内蒙古大学图书馆

16th Annual International Conference on Combinatorial Optimization and Applications, COCOA 2023

作者： Ning, Quancheng Zhang, Nan Institute of Computing Theory and Technology Xidian University Xi’an710071 China

ISBN: (纸本)9783031496134

In this paper, an end-to-end multi-stage dehazing network based on convolution and Transformer is proposed. The network design is divided into three parts: encoding network, feature fusion network and decoding network. The encoding network extracts primary features of the haze image, the feature fusion network uses the serial Transformer module and the dynamic convolution module to make the information extracted from the features richer, and the decoding network is used to recover the image resolution. In the resolution restoration stage of the decoding module, the MixUp module is used to restore the high resolution image by combining the extracted primary features to reduce the loss of information. Extensive experiments were conducted on synthetic and real datasets to validate the role of Transformer module and Dynamic Convolution module in dehazing respectively. The results show that the proposed method achieves a good objective evaluation score and reconstructs a subjectively better dehazing image with a PSNR of 35.37 and a SSIM of 0.9849 on the SOTS test dataset. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.

关键词： Convolution

来源：评论

学校读者我要写书评

暂无评论

An Approach to Agent Path Planning Under Temporal Logic Constraints 29th

An Approach to Agent Path Planning Under Temporal Logic Cons...

引用

29th International computing and Combinatorics Conference, COCOON 2023

作者： Yu, Chaofeng Zhang, Nan Duan, Zhenhua Tian, Cong Institute of Computing Theory and Technology and ISN Laboratory Xidian University Xi’an710071 China

ISBN: (纸本)9783031491924

The capability of path planning is a necessity for an agent to accomplish tasks autonomously. Traditional path planning methods fail to complete tasks that are constrained by temporal properties, such as conditional reachability, safety, and liveness. Our work presents an integrated approach that combines reinforcement learning (RL) with multi-objective optimization to address path planning problems with the consideration of temporal logic constraints. The main contributions of this paper are as follows. (1) We propose an algorithm LCAP 2 to design extra rewards and accelerate training by tackling a multi-objective optimization problem. The experimental results show that the method effectively accelerates the convergence of the path lengths traversed during the agent’s training. (2) We provide a convergence theorem based on the fixed-point theory and contraction mapping theorem. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.

关键词： Temporal logic

来源：评论

学校读者我要写书评

暂无评论

A Dynamic Parameter Adaptive Path Planning Algorithm 16th

A Dynamic Parameter Adaptive Path Planning Algorithm

引用

16th Annual International Conference on Combinatorial Optimization and Applications, COCOA 2023

作者： Yao, Guangyu Zhang, Nan Duan, Zhenhua Tian, Cong Institute of Computing Theory and Technology and ISN Laboratory Xidian University Xi’an710071 China

ISBN: (纸本)9783031496134

Path planning in complex environments has always been a focus of research for scholars both domestically and internationally. This study addresses the challenge of path planning that combines obstacle avoidance and optimal path searching in scenarios lacking prior knowledge. The proposed approach introduces a parameter dynamic adaptation strategy for path planning. Experimental investigations are conducted using grid-based maps, and the results demonstrate that the method presented in this paper surpasses Q-learning and Sarsa algorithms in terms of comprehensive exploration, enhanced stability, and quicker convergence speed. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.

关键词： Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

Formal Modeling and Verification of Convolutional Neural Networks based on MSVL 9

Formal Modeling and Verification of Convolutional Neural Net...

引用

9th International Conference on Dependable Systems and Their Applications, DSA 2022

作者： Zhao, Liang Wu, Leping Gao, Yu Wang, Xiaobing Yu, Bin Institute of Computing Theory and Technology Xidian University Shaanxi Province Xian China

ISBN: (数字)9781665488778

ISBN: (纸本)9781665488778

With the rapid development and wide application of neural networks, it is more and more important to use formal methods to verify and ensure their security. In this paper, we propose a comprehensive formal framework for the modeling and verification of convolutional neural networks (CNN). The framework is developed based on Modeling, Simulation and Verification Language (MSVL), a formal language with temporal-logic basis. First, the structure and basic behavior of a CNN are characterized hierarchically as MSVL specifications. On this basis, the prediction model, training model and verification module are developed. Experimental results show that the framework constructs formal models of CNNs effectively and supports the verification of various network properties. © 2022 IEEE.

关键词： Formal verification

来源：评论

学校读者我要写书评

暂无评论

DUAL SPARSE TRAINING FRAMEWORK: INDUCING ACTIVATION MAP SPARSITY VIA TRANSFORMED 1 REGULARIZATION

arXiv

引用

arXiv 2024年

作者： Yu, Xiaolong Tian, Cong Institute of Computing Theory and Technology Xidian University Xi’an China

Although deep convolutional neural networks have achieved rapid development, it is challenging to widely promote and apply these models on low-power devices, due to computational and storage limitations. To address this issue, researchers have proposed techniques such as model compression, activation sparsity induction, and hardware accelerators. This paper presents a method to induce the sparsity of activation maps based on Transformed 1 regularization, so as to improve the research in the field of activation sparsity induction. Further, the method is innovatively combined with traditional pruning, constituting a dual sparse training framework. Compared to previous methods, Transformed 1 can achieve higher sparsity1 and better adapt to different network structures. Experimental results show that the method achieves improvements by more than 20% in activation map sparsity on most models and corresponding datasets without compromising the accuracy. Specifically, it achieves a 27.52% improvement for ResNet18 on the ImageNet dataset, and a 44.04% improvement for LeNet5 on the MNIST dataset. In addition, the dual sparse training framework can greatly reduce the computational load and provide potential for reducing the required storage during runtime. Specifically, the ResNet18 and ResNet50 models obtained by the dual sparse training framework respectively reduce 81.7% and 84.13% of multiplicative floating-point operations, while maintaining accuracy and a low pruning rate. Copyright © 2024, The Authors. All rights reserved.

关键词： Chemical activation

来源：评论

学校读者我要写书评

暂无评论

Three Algorithms for Converting Control Flow Statements from Python to XD-M 16th

Three Algorithms for Converting Control Flow Statements fro...

引用

16th International Conference on Algorithmic Aspects in Information and Management, AAIM 2022

作者： Wang, Jiarui Zhang, Nan Duan, Zhenhua Institute of Computing Theory and Technology and ISN Laboratory Xidian University Xi’an710071 China

ISBN: (纸本)9783031160806

This paper presents an approach to show how to implement three complex statements: continue, break, return of Python to XD-M language. To this end, three algorithms for implementing three complex statements are given in detail. Further, three complex statements are implemented in XD-M. Finally, a program example is presented to illustrate how to use the proposed approach to build an XD-M program equivalent to the original Python program. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.

关键词： Python

来源：评论

学校读者我要写书评

暂无评论

An Approach for Automatically Generating Traces for Python Programs 9

An Approach for Automatically Generating Traces for Python P...

引用

9th International Conference on Dependable Systems and Their Applications, DSA 2022

作者： Zhong, Xiaoting Zhang, Nan Duan, Zhenhua Institute of Computing Theory and Technology Xidian University Shaanxi Province Xi'an710071 China

ISBN: (数字)9781665488778

ISBN: (纸本)9781665488778

Dynamic specification mining technology depends on traces, and the coverage of traces determine the effectiveness of specification. With the development of Python programs, mining specification of Python programs has become a big challenge. However, there are no tools that are enable to generate traces automatically. In order to improving the coverage and availability of traces, we propose an approach for automatically generating traces for Python programs. Firstly, we preprocess a Python program to get the annotation information. Next, using a random generation method and symbolic execution technology, we get test data automatically for the program. Then, we generate test cases by coverage guidance. Finally, we execute the test cases to extract traces of the program. Taking the generated traces as the input of specification mining tool, we get a specification model represented by finite state automata, which directly proves the availability of traces. And the experiment shows that our method has more advantages in statement coverage and branch coverage than the existing Python test case generation tool Pynguin. © 2022 IEEE.

关键词： Python

来源：评论

学校读者我要写书评

暂无评论

First-principles electron-phonon interactions and polarons in the parent cuprate La2CuO4

引用

Physical Review Research 2025年第1期7卷 L012073-L012073页

作者： Benjamin K. Chang Iurii Timrov Jinsoo Park Jin-Jian Zhou Nicola Marzari Marco Bernardi Department of Applied Physics and Materials Science and Department of Physics California Institute of Technology Pasadena California 91125 USA PSI Center for Scientific Computing Theory and Data Paul Scherrer Institut 5232 Villigen PSI Switzerland Department of Physics Pohang University of Science and Technology Pohang 37673 Korea School of Physics Beijing Institute of Technology Beijing 100081 China Theory and Simulation of Materials (THEOS) and National Centre for Computational Design and Discovery of Novel Materials (MARVEL) École Polytechnique Fédérale de Lausanne (EPFL) 1015 Lausanne Switzerland

Understanding electronic interactions in high-temperature superconductors is an outstanding challenge. In the widely studied cuprate materials, experimental evidence points to strong electron-phonon (e-ph) coupling and broad photoemission spectra. Yet, the microscopic origin of this behavior is not fully understood. Here, we study e-ph interactions and polarons in a prototypical parent (undoped) cuprate, La2CuO4 (LCO), by means of first-principles calculations. Leveraging parameter-free Hubbard-corrected density functional theory, we obtain a ground state with the band gap and Cu magnetic moment in nearly exact agreement with experiments. This enables a quantitative characterization of e-ph interactions. Our calculations reveal two classes of longitudinal optical (LO) phonons with strong e-ph coupling to hole states. These modes consist of bond stretching and bond bending in the Cu-O plane as well as vibrations of apical O atoms. The hole spectral functions, obtained with a cumulant method that can capture strong e-ph coupling, exhibit broad quasiparticle peaks with a small spectral weight (Z≈0.25) and pronounced LO-phonon sidebands characteristic of polaron effects. Our calculations predict features observed in photoemission spectra, including a 40-meV peak in the e-ph coupling distribution function not explained by existing models. These results show that the universal strong e-ph coupling found experimentally in doped lanthanum cuprates is also present in the parent compound, and elucidate its microscopic origin.

关键词： High temperature superconductors

来源：评论

学校读者我要写书评

暂无评论

AutoSketch: Automatic Sketch-Oriented Compiler for Query-driven Network Telemetry 21

AutoSketch: Automatic Sketch-Oriented Compiler for Query-dri...

引用

21st USENIX Symposium on Networked Systems Design and Implementation, NSDI 2024

作者： Sun, Haifeng Huang, Qun Sun, Jinbo Wang, Wei Li, Jiaheng Li, Fuliang Bao, Yungang Yao, Xin Zhang, Gong National Key Laboratory for Multimedia Information Processing School of Computer Science Peking University China Institute of Computing Technology CAS China Northeastern University China Huawei Theory Department

ISBN: (纸本)9781939133397

Recent network telemetry witnesses tremendous progress in two directions: query-driven telemetry that targets expressiveness as the primary goal, and sketch-based algorithms that address resource-accuracy trade-offs. In this paper, we propose AutoSketch that aims to integrate the advantages of both classes. In a nutshell, AutoSketch automatically compiles high-level operators into sketch instances that can be readily deployed with low resource usage and incur limited accuracy drop. However, there remains a gap between the expressiveness of high-level operators and the underlying realization of sketch algorithms. AutoSketch bridges this gap in three aspects. First, AutoSketch extends its interface derived from existing query-driven telemetry such that users can specify the desired telemetry accuracy. The specified accuracy intent will be utilized to guide the compiling procedure. Second, AutoSketch leverages various techniques, such as syntax analysis and performance estimation, to construct efficient sketch instances. Finally, AutoSketch automatically searches for the most suitable parameter configurations that fulfill the accuracy intent with minimum resource usage. Our experiments demonstrate that AutoSketch can achieve high expressiveness, high accuracy, and low resource usage compared to state-of-the-art telemetry solutions. © 2024 Proceedings of the 21st USENIX Symposium on Networked Systems Design and Implementation, NSDI 2024. All rights reserved.

关键词： Economic and social effects

来源：评论

学校读者我要写书评

暂无评论

Model-Based Feature Selection for Neural Networks: A Mixed-Integer Programming Approach 1

引用

17th International Conference on Learning and Intelligent Optimization, LION-17 2023

作者： Zhao, Shudian Tsay, Calvin Kronqvist, Jan Optimization and Systems Theory Department of Mathematics KTH Royal Institute of Technology Stockholm Sweden Department of Computing Imperial College London London United Kingdom

ISBN: (数字)9783031445057

ISBN: (纸本)9783031445040

In this work, we develop a novel input feature selection framework for ReLU-based deep neural networks (DNNs), which builds upon a mixed-integer optimization approach. While the method is generally applicable to various classification tasks, we focus on finding input features for image classification for clarity of presentation. The idea is to use a trained DNN, or an ensemble of trained DNNs, to identify the salient input features. The input feature selection is formulated as a sequence of mixed-integer linear programming (MILP) problems that find sets of sparse inputs that maximize the classification confidence of each category. These "inverse" problems are regularized by the number of inputs selected for each category and by distribution constraints. Numerical results on the well-known MNIST and FashionMNIST datasets show that the proposed input feature selection allows us to drastically reduce the size of the input to ∼ 15% while maintaining a good classification accuracy. This allows us to design DNNs with significantly fewer connections, reducing computational effort and producing DNNs that are more robust towards adversarial attacks. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.

关键词： Deep neural networks

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：