检索结果-内蒙古大学图书馆

Conference on Computer Vision and Pattern Recognition (CVPR)

作者： Yushun Tang Ce Zhang Heng Xu Shuoshuo Chen Jie Cheng Luziwei Leng Qinghai Guo Zhihai He Department of Electronic and Electrical Engineering Southern University of Science and Technology Shenzhen China Advanced Computing and Storage Laboratory Huawei Technologies Co. Ltd. Shenzhen China Pengcheng Laboratory Shenzhen China

Fully test-time adaptation aims to adapt the network model based on sequential analysis of input samples during the inference stage to address the cross-domain performance degradation problem of deep neural networks. We take inspiration from the biological plausibility learning where the neuron responses are tuned based on a local synapse-change procedure and activated by competitive lateral inhibition rules. Based on these feed-forward learning rules, we design a soft Hebbian learning process which provides an unsupervised and effective mechanism for online adaptation. We observe that the performance of this feed-forward Hebbian learning for fully test-time adaptation can be significantly improved by incorporating a feedback neuromodulation layer. It is able to fine-tune the neuron responses based on the external feedback generated by the error backpropagation from the top inference layers. This leads to our proposed neuro-modulated Hebbian learning (NHL) method for fully test-time adaptation. With the unsupervised feed-forward soft Hebbian learning being combined with a learned neuromodulator to capture feedback from external responses, the source model can be effectively adapted during the testing process. Experimental results on benchmark datasets demonstrate that our proposed method can significantly improve the adaptation performance of network models and outperforms existing state-of-the-art methods.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Abnormal Thermal Conductivity Increase in β-Ga2O3 by an Unconventional Bonding Mechanism Using Machine-Learning Potential

SSRN

引用

SSRN 2024年

作者： Zhou, Wu-Xing Wu, Cheng-Wei Cao, Hao-Ran Zeng, Yu-Jia Xie, Guofeng Zhang, Gang School of Materials Science and Engineering Hunan Provincial Key Laboratory of Advanced Materials for New Energy Storage and Conversion Hunan University of Science and Technology Xiangtan411201 China Institute of High Performance Computing A*STAR Singapore138632 Singapore

β-Ga2O3, with its ultrawide band gap (~4.9 eV) and high critical electric field, holds potential in power electronics but is limited by low thermal conductivity, which is critical to the performance and reliability because the high level of heat flux density induced by the extremely high levels of power density. Combining first-principles calculations, machine-learning potentials, and solving the phonon Boltzmann transport equation, we found that substituting octahedral-coordinated Ga with Al significantly enhances thermal conductivity from 100K to 800K. At room temperature, Al-substituted β-Ga2O3 achieves 38.91 W/mK, more than 2-fold that of pristine β-Ga2O3 (17.10 W/mK) and even higher than β-Al2O3 (34.77 W/mK). This enhancement, unusual due to the heavier atomic mass and mixed mass distribution, is rooted in suppressed anharmonic characteristics caused by reduced bonding strength inhomogeneity. Our results may inspire the rational design of materials with tailored thermal properties through chemical bonding mechanisms. © 2024, The Authors. All rights reserved.

关键词： Thermal conductivity

来源：评论

学校读者我要写书评

暂无评论

A Low-Power Approximate Multiplier with Sign-Focus Compressor and Error Compensation

A Low-Power Approximate Multiplier with Sign-Focus Compresso...

引用

IEEE Asia-Pacific Conference on Circuits and Systems

作者： Laimin Du Leibin Ni Xiong Liu Wei Mao Hao Yu School of Microelectronics Southern University of science and technology Advanced Computing and Storage Laboratory Huawei Technologies Co. Ltd Shenzhen China

Approximate computing is an emerging and effective means of reducing energy consumption in digital circuits, which is critical for improving the energy efficiency of edge computing devices. In this paper, we propose a novel low power signed approximate multiplier based on a sign-focus compressor and error compensation. The sign-focus compressor is a compressor customized for partial product matrix of signed operands. After probabilistic analysis, it is reduced to the simplest logic circuit. At the same time, we truncate the low N-I columns of the partial product matrix and perform error compensation. By logic synthesis evaluation, the proposed approximate multiplier saves 49.84% power, 46.41% area and 24.65% delay compared to the exact multiplier. Compared with other state-of-the-art approximate multipliers, the proposed approximate multiplier has better accuracy-performance trade-off.

关键词： Energy consumption Circuits and systems Logic circuits Asia Error compensation Probabilistic logic Energy efficiency

来源：评论

学校读者我要写书评

暂无评论

NavG: Risk-Aware Navigation in Crowded Environments Based on Reinforcement Learning with Guidance Points

arXiv

引用

arXiv 2025年

作者： Zhang, Qianyi Luo, Wentao Liu, Boyi Zhang, Ziyang Wang, Yaoyuan Liu, Jingtai Institute of Robotics and Automatic Information System Nankai University Tianjin Key Laboratory of Intelligent Robotics Tianjin China Advanced Computing and Storage Lab Huawei 2012 Lab China Department of Electronic and Computer Engineering The Kong Kong University of Science and Technology Hong Kong

Motion planning in navigation systems is highly susceptible to upstream perceptual errors, particularly in human detection and tracking. To mitigate this issue, the concept of guidance points—a novel directional cue within a reinforcement learning-based framework—is introduced. A structured method for identifying guidance points is developed, consisting of obstacle boundary extraction, potential guidance point detection, and redundancy elimination. To integrate guidance points into the navigation pipeline, a perception-to-planning mapping strategy is proposed, unifying guidance points with other perceptual inputs and enabling the RL agent to effectively leverage the complementary relationships among raw laser data, human detection and tracking, and guidance points. Qualitative and quantitative simulations demonstrate that the proposed approach achieves the highest success rate and near-optimal travel times, greatly improving both safety and efficiency. Furthermore, real-world experiments in dynamic corridors and lobbies validate the robot’s ability to confidently navigate around obstacles and robustly avoid pedestrians. © 2025, CC BY.

关键词： Travel time

来源：评论

学校读者我要写书评

暂无评论

Statistical physics of deep neural networks: Initialization toward optimal channels

引用

Physical Review Research 2023年第2期5卷 023023-023023页

作者： Kangyu Weng Aohua Cheng Ziyang Zhang Pei Sun Yang Tian Tsien Excellence in Engineering Program Tsinghua University Beijing 100084 China Laboratory of Advanced Computing and Storage Central Research Institute 2012 Laboratories Huawei Technologies Co. Ltd. Beijing 100084 China Department of Psychology & Tsinghua Laboratory of Brain and Intelligence Tsinghua University Beijing 100084 China

In deep learning, neural networks serve as noisy channels between input data and its latent representation. This perspective naturally relates deep learning with the pursuit of constructing channels with optimal performance in information transmission and representation. While considerable efforts are concentrated on realizing optimal channel properties during network optimization, we study a frequently overlooked possibility that neural networks can be initialized toward optimal channels. Our theory, consistent with experimental validation, identifies primary mechanics underlying this unknown possibility and suggests intrinsic connections between statistical physics and deep learning. Unlike the conventional theories that characterize neural networks applying the classic mean-field approximation, we offer analytic proof that this extensively applied simplification scheme is not appropriate in studying neural networks as information channels. To fill this gap, we develop a restricted mean-field framework applicable for characterizing the limiting behaviors of information propagation in neural networks without strong assumptions on inputs. Based on it, we propose an analytic theory to prove that mutual information maximization is realized between inputs and propagated signals when neural networks are initialized at dynamic isometry, a case where information transmits via norm-preserving mappings. These theoretical predictions are validated by experiments on real neural networks, suggesting the robustness of our theory against finite-size effects. Finally, we analyze our findings with information bottleneck theory to confirm the precise relations among dynamic isometry, mutual information maximization, and optimal channel properties in deep learning. Our work may lay a cornerstone for promoting deep learning in terms of network initialization and suggest general statistical physics mechanisms underlying diverse deep learning techniques.

关键词： Neuronal dynamics Artificial neural networks Deep learning Information theory Random matrix theory

来源：评论

学校读者我要写书评

暂无评论

Laplacian dynamics of convergent and divergent swarm behaviors

arXiv

引用

arXiv 2023年

作者： Tian, Yang Xu, Yunhui Sun, Pei Department of Psychology Tsinghua Laboratory of Brain and Intelligence Tsinghua University Beijing100084 China Laboratory of Advanced Computing and Storage Central Research Institute 2012 Laboratories Huawei Technologies Co. Ltd. Beijing100084 China Department of Physics Tsinghua University Beijing100084 China

Swarming phenomena are ubiquitous in various physical, biological, and social systems, where simple local interactions between individual units lead to complex global patterns. A common feature of diverse swarming phenomena is that the units exhibit either convergent or divergent evolution in their behaviors, i.e., becoming increasingly similar or distinct, respectively. The associated dynamics changes across time, leading to complex consequences on a global scale. In this study, we propose a generalized Laplacian dynamics model to describe both convergent and divergent swarm behaviors, where the trends of convergence and divergence compete with each other and jointly determine the evolution of global patterns. We empirically observe non-trivial phase-transition-like phenomena between the convergent and divergent evolution phases, which are controlled by local interaction properties. We also propose a conjecture regarding the underlying phase transition mechanisms and outline the main theoretical difficulties for testing this conjecture. Overall, our framework may serve as a minimal model of swarm behaviors and their intricate phase transition dynamics. © 2023, CC BY.

关键词： Dynamics

来源：评论

学校读者我要写书评

暂无评论

Network comparison via encoding, decoding, and causality

引用

Physical Review Research 2023年第3期5卷 033129-033129页

作者： Yang Tian Hedong Hou Guangzheng Xu Ziyang Zhang Pei Sun Department of Psychology & Tsinghua Laboratory of Brain and Intelligence Tsinghua University Beijing 100084 China and Laboratory of Advanced Computing and Storage Central Research Institute 2012 Laboratories Huawei Technologies Co. Ltd. Beijing 100084 China Institut de Mathématiques d'Orsay 14 Rue Joliot-Curie 91190 Gif-sur-Yvette France Department of Computer Science University College London London WC1E 6AE United Kingdom Laboratory of Advanced Computing and Storage Central Research Institute 2012 Laboratories Huawei Technologies Co. Ltd. Beijing 100084 China Department of Psychology & Tsinghua Laboratory of Brain and Intelligence Tsinghua University Beijing 100084 China

Quantifying the relations (e.g., similarity) between complex networks paves the way for studying the latent information shared across networks. However, fundamental relation metrics are not well-defined between networks. As a compromise, prevalent techniques measure network relations in data-driven manners, which are inapplicable to analytic derivations in physics. To resolve this issue, we present a theory for obtaining an optimal characterization of network topological properties. We show that a network can be fully represented by a Gaussian variable defined by a function of the Laplacian, which simultaneously satisfies network-topology-dependent smoothness and maximum entropy properties. Based on it, we can analytically measure diverse relations between complex networks. As illustrations, we define encoding (e.g., information divergence and mutual information), decoding (e.g., Fisher information), and causality (e.g., Granger causality and conditional mutual information) between networks. We validate our framework on representative networks (e.g., random networks, protein structures, and chemical compounds) to demonstrate that a series of science and engineering challenges (e.g., network evolution, embedding, and query) can be tackled from a new perspective. An implementation of our theory is released as a multiplatform toolbox.

关键词： Networks & random structures Random graphs Graph theory Information theory

来源：评论

学校读者我要写书评

暂无评论

Large contributions from optical phonons to thermal transport in hexagonal carbon-boron-nitrogen monolayers

引用

Physical Review B 2023年第23期108卷 235420-235420页

作者： Xue-Kun Chen (陈学坤) Yue Zhang (张越) Qing-Qing Luo (罗青清) Xue Chen (陈雪) Pin-Zhen Jia (贾聘真) Wu-Xing Zhou (周五星) School of Mathematics and Physics and Hunan Key Laboratory of Mathematical Modeling and Scientific Computing University of South China Hengyang 421001 China Department of Mathematics and Physics Hunan Institute of Technology Hengyang 421002 China School of Materials Science and Engineering and Hunan Provincial Key Laboratory of Advanced Materials for New Energy Storage and Conversion Hunan University of Science and Technology Xiangtan 411201 People's Republic of China

As representative semiconducting hexagonal carbon-boron-nitride lattices, C6BN and C2BN are experimentally realized two-dimensional (2D) plane materials and have recently become the focus of research. Herein, combining first-principles calculations with the Boltzmann transport equation, we performed a comprehensive study on the phonon interaction and thermal conductivity in C6BN and C2BN monolayers. It is found that the thermal conductivities of C6BN and C2BN monolayers at room temperature are reduced by 79% and 73%, respectively, due to four-phonon scattering, compared with the results including three-phonon scattering only. We can attribute this phenomenon to giant four-phonon scattering exclusive for the heat-carrying out-of-plane acoustic (ZA) phonons, because the reflection symmetry allows four-ZA processes much higher than three-ZA processes, and the quasiparallel behavior between the ZA and low-lying out-of-plane optical (ZO) branches contribute to a broad phase space for four-phonon scattering as well. Moreover, C6BN monolayer exhibits unusual behavior that optical phonons contribute about ∼60% to the overall thermal conductivity under the four-phonon picture, which differs from the traditional case that acoustic phonons dominate thermal conductivity. Unexpectedly, two low-lying ZO modes have as high as 38% contributions to the thermal transport at 300 K under the four-phonon picture, causing 60% contribution of optical phonon modes, apparently larger than that of the three-phonon case (15%) and many other 2D materials, also indicating the four-phonon scattering has a more significant effect on acoustic phonons than on optical phonons. This finding not only highlights insight into the nature of phonon transport, but also provides a promising strategy for manipulation of heat transport based on optical phonon modes.

关键词： Phonons Thermal conductivity Boron nitride Graphene First-principles calculations Lattice-Boltzmann methods

来源：评论

学校读者我要写书评

暂无评论

Koopman Neural Operator as a Mesh-Free Solver of Non-Linear Partial Differential Equations

SSRN

引用

SSRN 2023年

作者： Xiong, Wei Huang, Xiaomeng Zhang, Ziyang Deng, Ruixuan Sun, Pei Tian, Yang Department of Earth System Science Tsinghua University Beijing100084 China Department of Psychology Tsinghua Laboratory of Brain and Intelligence Tsinghua University Beijing100084 China Department of Biomedical Engineering Tsinghua University Beijing100084 China Laboratory of Advanced Computing and Storage Central Research Institute Laboratories Huawei Technologies Co. Ltd 2012 China

The lacking of analytic solutions of diverse partial differential equations (PDEs) gives birth to a series of computational techniques for numerical solutions. Although numerous latest advances are accomplished in developing neural operators, a kind of neural-network-based PDE solver, these solvers become less accurate and explainable while learning long-term behaviors of non-linear PDE families. In this paper, we propose Koopman neural operator (KNO), a new neural operator, to overcome these challenges. With the same objective of learning an infinite-dimensional mapping between Banach spaces that serves as the solution operator of the target PDE family, our approach differs from existing models by formulating a non-linear dynamic system of equation solution. By approximating the Koopman operator, an infinite-dimensional operator governing all possible observations of the dynamic system, to act on the flow mapping of the dynamic system, we can equivalently learn the solution of a non-linear PDE family by solving simple linear prediction problems. In zero-shot prediction and long-term prediction experiments on representative PDEs (e.g., the Navier-Stokes equation), KNO exhibits notable advantages in breaking the tradeoff between accuracy and efficiency (e.g., model size) while previous state-of-the-art models are limited. These results suggest the potential of KNO in supporting diverse science and engineering applications. © 2023, The Authors. All rights reserved.

关键词： Partial differential equations

来源：评论

学校读者我要写书评

暂无评论

Token-Mol 1.0：tokenized drug design with large language model

arXiv

引用

arXiv 2024年

作者： Wang, Jike Qin, Rui Wang, Mingyang Fang, Meijing Zhang, Yangyang Zhu, Yuchen Su, Qun Wan, Xiaozhe Gou, Qiaolin Shen, Chao Zhang, Odin Wu, Zhenxing Jiang, Dejun Zhang, Xujun Zhao, Huifeng Ge, Jingxuan Wu, Zhourui Liu, Liwei Kang, Yu Hsieh, Chang-Yu Hou, Tingjun College of Pharmaceutical Sciences Zhejiang University Zhejiang Hangzhou310058 China Advanced Computing and Storage Laboratory Central Research Institute 2012 Laboratories Huawei Technologies Co. Ltd. Jiangsu Nanjing210000 China Paul Allen School University of Washington Seattle United States Key Laboratory of Spine and Spinal cord Injury Repair and Regeneration Ministry of Education Tongji University Shanghai China

Significant interests have recently risen in leveraging sequence-based large language models (LLMs) for drug design. However, most current applications of LLMs in drug discovery lack the ability to comprehend 3D structures, thereby limiting their effectiveness in tasks that specifically involve molecular conformations. In this study, we introduce Token-Mol, a token-only 3D drug design model. This model encodes all pertinent molecular information, including both 2D and 3D structures, as well as molecular property data, into tokens. It transforms classification and regression tasks in drug discovery into probabilistic prediction problems, ultimately enabling a comprehensive and cohesive learning approach through a unified paradigm. Token-Mol is constructed on the foundation of the transformer decoder architecture, and trained using random causal masking techniques. Moreover, we propose the Gaussian cross-entropy loss function to overcome the challenges encountered in regression tasks, which significantly enhances the capacity of LLMs to learn continuous numerical values. Experimental validations demonstrate that Token-Mol surpasses state-of-the-art methods in the molecular conformation generation task, achieving improvements across various metrics of over 10% and 20% on two datasets, respectively. In a series of molecular property prediction tasks, Token-Mol exhibits an average improvement of 30% in regression tasks compared to existing token-only models. In the pocket-based molecular generation task, Token-Mol can generate molecules with a notable improvement in the drug-likeness (QED) and synthetic accessibility by ~11% and 14%, respectively, meanwhile maintaining a comparable level of affinity, as estimated by the Vina score, compared to existing methods. When compared to expert models based on diffusion architecture that exhibit similar performance, Token-Mol demonstrates an impressive generation speed that is 35 times faster. In real-world drug design scenarios,

关键词： Structural dynamics

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：