检索结果-内蒙古大学图书馆

39th Annual AAAI Conference on Artificial Intelligence, AAAI 2025

作者： Wang, Lu Xu, Liuchi Yang, Xiong Huang, Zhenhua Cheng, Jun School of Computer Science and Engineering Northeastern University Shenyang China Institute of Applied Artificial Intelligence of the Guangdong-Hong Kong-Macao Greater Bay Area Shenzhen Polytechnic University Shenzhen China School of Computer Science South China Normal University Guangzhou China Guangdong Provincial Key Laboratory of Robotics and Intelligent System Shenzhen Institute of Advanced Technology CAS China The Chinese University of Hong Kong Hong Kong

ISBN: (纸本)157735897X

Knowledge distillation transfers "dark knowledge" from a large teacher model to a smaller student model, yielding a highly efficient network. To improve the network's generalization ability, existing works use a larger temperature coefficient for knowledge distillation. Nevertheless, these methods may reduce the confidence of the target category and lead to ambiguous recognition of similar samples. To mitigate this issue, some studies introduce intra-batch distillation to reduce prediction discrepancy. However, these methods overlook the inconsistency between background information and the target category, which may increase prediction bias due to noise disturbance. Additionally, label imbalance from random sampling and batch size can undermine network generalization reliability. To tackle these challenges, we propose a simple yet effective Intra-class Knowledge Distillation (IKD) method that facilitates knowledge sharing within the same class to ensure consistent predictions. First, we initialize the matrix and the vector to store logits and class counts provided by the teacher, respectively. Then, in the first epoch, we calculate the sum of logits and sample counts per class and perform KD to prevent knowledge omission. Finally, in subsequent training, we update the matrix to obtain the average logits and compute the KL divergence between the student's output and the updated matrix according to the label index. This process ensures intra-class consistency and improves the student's performance. Furthermore, this method theoretically reduces prediction bias by ensuring intra-class consistency. Extensive experiments on the CIFAR-100, ImageNet-1K, and Tiny-ImageNet datasets validate the superiority of IKD. Copyright © 2025, Association for the Advancement of Artificial Intelligence (***). All rights reserved.

关键词： Students

来源：评论

学校读者我要写书评

暂无评论

Bioinspired Sensing of Undulatory Flow Fields Generated by Leg Kicks in Swimming

arXiv

引用

arXiv 2025年

作者： Wang, Jun Shen, Tongsheng Zhao, Dexin Zhang, Feitian Department of Advanced Manufacturing and Robotics The State Key Laboratory of Turbulence and Complex Systems College of Engineering Peking University Beijing100871 China The National Innovation Institute of Defense Technology Beijing China National Innovation Institute of Defense Technology Beijing100071 China

The artificial lateral line (ALL) is a bioinspired flow sensing system for underwater robots, comprising of distributed flow sensors. The ALL has been successfully applied to detect the undulatory flow fields generated by body undulation and tail-flapping of bioinspired robotic fish. However, its feasibility and performance in sensing the undulatory flow fields produced by human leg kicks during swimming has not been systematically tested and studied. This paper presents a novel sensing framework to investigate the undulatory flow field generated by swimmer’s leg kicks, leveraging bioinspired ALL sensing. To evaluate the feasibility of using the ALL system for sensing the undulatory flow fields generated by swimmer leg kicks, this paper designs an experimental platform integrating an ALL system and a lab-fabricated human leg model. To enhance the accuracy of flow sensing, this paper proposes a feature extraction method that dynamically fuses time-domain and time-frequency characteristics. Specifically, time-domain features are extracted using one-dimensional convolutional neural networks and bidirectional long short-term memory networks (1DCNN-BiLSTM), while time-frequency features are extracted using short-term Fourier transform and two-dimensional convolutional neural networks (STFT-2DCNN). These features are then dynamically fused based on attention mechanisms to achieve accurate sensing of the undulatory flow field. Furthermore, extensive experiments are conducted to test various scenarios inspired by human swimming, such as leg kick pattern recognition and kicking leg localization, achieving satisfactory results. Copyright © 2025, The Authors. All rights reserved.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Exfoliated multi-layered graphene anode with the broadened delithiation voltage plateau below 0.5 V

引用

Journal of Energy Chemistry 2020年第10期29卷 233-242页

作者： Xinlong Ma Xinyu Song Yushu Tang Enzuo Liu Chenggen Xu Chuanlei Qi Yun Li Jinsen Gao Yongfeng Li State Key Laboratory of Heavy Oil Processing China University of PetroleumBeijing 102249China Key Laboratory for Advanced Technology in Environmental Protection of Jiangsu Province Yancheng Institute of TechnologyYancheng 224051JiangsuChina School of Materials Science and Engineering and Tianjin Key Laboratory of Composites and Functional Materials Tianjin UniversityTianjin 300072China

The commercial graphite(CG)is the conventional anode material for lithium ion batteries(LIBs)due to its low delithiation voltage plateau(below 0.5 V)and extraordinary ***,the further promotion of energy density of LIBs is restricted by the limited capacity below 0.5 V of ***,based on the supercritical CO2 exfoliation technique,the production of multi-layered graphene(MLG)is achieved from the pilot scale production *** great merit of the exfoliated MLG anode is that the voltage plateau below 0.5 V is broadened obviously as compared to those of natural graphite and ***,no obvious lithium dendrites are observed for MLG during the lithiation *** large delithiation capacity under the low voltage plateau of MLG is mainly benefited from the combination of Li intercalation and boundary storage mechanism,which is further confirmed by the density functional theory *** LiFePO4/MLG full cell can afford the satisfactory electrochemical property with respect to the capacity,energy density and ultralong cycling stability(90%capacity retention after 500 cycles at 2 C),significantly better than that of LiFePO4/***,this developed technique not only dedicates to producing the high-performance anode for LIBs but also opens a door for the mass production of MLG in the industrial scale.

关键词： Multi-layered graphene Supercritical CO2 exfoliation Anode Broadened delithiation voltage plateau Intercalation and boundary storage

来源：评论

学校读者我要写书评

暂无评论

Nucleation and recombination dependent desorption performance of Mg-based hydrides 5

Nucleation and recombination dependent desorption performanc...

引用

5th International Conference on Energy Engineering and Environmental Protection, EEEP 2020

作者： Xie, Lishuai Wu, Tong School of Material Science and Engineering Nanjing Institute of Technology Nanjing211167 China Jiangsu Key Laboratory of Advanced Structural Materials and Application Technology Nanjing China

High desorption temperature and sluggish desorption kinetics hinder the large-scale application of Mg-based hydrogen storage materials. Fundamental understanding of the dehydrogenation process and underlying mechanism is crucial for further modifying the desorption performance. Aiming at elucidating the dehydrogenation process of Mg-based hydrides, partially desorbed pure MgH2 and MgH2-Mg2NiH4-CeH2.73 composite have been prepared to investigate the effects of metal phase nucleation on overall desorption performance. It is found that metal Mg is formed on the surface of partially desorbed MgH2 and MgH2-Mg2NiH4-CeH2.73. Correspondingly, the desorption temperature of partially desorbed samples is significantly decreased, indicating an important role nucleation plays in desorption process. With minor Ni addition, the desorption temperature and kinetics of MgH2-Mg2NiH4-CeH2.73 are significantly improved. Promoting hydrogen recombination is another essential aspect to modify the desorption properties of Mg hydrides. © Published under licence by IOP Publishing Ltd.

关键词： Magnesium compounds

来源：评论

学校读者我要写书评

暂无评论

Regulating ethane and ethylene synthesis by proton corridor microenvironment for CO_(2) electrolysis

引用

Journal of Energy Chemistry 2023年第12期87卷 368-377,I0010页

作者： Xiaowen Zhang Bohua Ren Hao Li Shuxuan Liu Haoyang Xiong Silong Dong Yifan Li Dan Luo Yi Cui Guobin Wen Xin Wang South China Academy of Advanced Optoelectronics International Academy of Optoelectronics at ZhaoqingSouth China Normal UniversityGuangzhou 510006GuangdongChina Department of Chemical Engineering Waterloo Institute for NanotechnologyUniversity of Waterloo200 University Avenue WestWaterlooOntario N2L 3G1Canada School of Nano Technology and Nano Bionics University of Science and Technology of ChinaHefei 230026AnhuiChina i-lab Nano-X Vacuum Interconnected WorkstationSuzhou Institute of Nano-Tech and Nano-BionicsChinese Academy of SciencesSuzhou 215123JiangsuChina State Key Laboratory of Chem/Bio-Sensing and Chemometrics Provincial Hunan Key Laboratory for Graphene Materials and DevicesCollege of Chemistry and Chemical EngineeringNational Supercomputer Centers in ChangshaHunan UniversityChangsha 410082HunanChina

Electrocatalytic reduction of carbon dioxide is one of the most effective strategies to achieve carbon neutrality and energy *** high-value multi-carbon products have been widely studied,limited electrocatalysts have been reported for the selective conversion of *** importantly,the factors tuning the selectivity between ethane and ethylene have not been ***,Zn@Cu nanowire arrays(Zn@Cu-NWAs) catalyst is proposed to stimulate the maintenance of efficient CO_(2)-to-C_(2)H_(6) conversion at high current ***,in order to investigate the factors affecting the interconversion between ethane and ethylene,the counterpart catalyst that facilitates C–C coupling to ethylene was also ***-of-flight secondary-ion mass spectroscopy(TOF-SIMS),in-situ Raman spectroscopy,and simulation results show that Zn@Cu-NWAs can provide a localized proton corridor environment for the formation of ethane,accelerating the further proton-coupled CO_(2) reduction reaction(CO_(2)RR)***,this catalyst delivered an ethane Faraday efficiency of over 65% at-1.14 V *** with a total current density of 142.3 mA/cm^(2).This work provides a new perspective on regulating the local microenvironment to modify the selectivity of multi-carbon products.

关键词： Proton microenvironment CO_(2)reduction C2+product Selective conversion Cu-Zn materials

来源：评论

学校读者我要写书评

暂无评论

GTCRN: A Speech Enhancement Model Requiring Ultralow Computational Resources

GTCRN: A Speech Enhancement Model Requiring Ultralow Computa...

引用

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Xiaobin Rong Tianchi Sun Xu Zhang Yuxiang Hu Changbao Zhu Jing Lu Key Laboratory of Modern Acoustics Nanjing University Nanjing China NJU-Horizon Intelligent Audio Lab Horizon Robotics Beijing China Jiangsu Thingstar Information Technology Co. Ltd. Nanjing China

While modern deep learning-based models have significantly outperformed traditional methods in the area of speech enhancement, they often necessitate a lot of parameters and extensive computational power, making them impractical to be deployed on edge devices in real-world applications. In this paper, we introduce Grouped Temporal Convolutional Recurrent Network (GTCRN), which incorporates grouped strategies to efficiently simplify a competitive model, DPCRN. Additionally, it leverages subband feature extraction modules and temporal recurrent attention modules to enhance its performance. Remarkably, the resulting model demands ultralow computational resources, featuring only 23.7 K parameters and 39.6 MMACs per second. Experimental results show that our proposed model not only surpasses RNNoise, a typical lightweight model with similar computational burden, but also achieves competitive performance when compared to recent baseline models with significantly higher computational resources requirements.

关键词：

来源：评论

学校读者我要写书评

暂无评论

A Lightweight Hybrid Multi-Channel Speech Extraction System with Directional Voice Activity Detection

A Lightweight Hybrid Multi-Channel Speech Extraction System ...

引用

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Tianchi Sun Tong Lei Xu Zhang Yuxiang Hu Changbao Zhu Jing Lu Key Laboratory of Modern Acoustics Nanjing University Nanjing China NJU-Horizon Intelligent Audio Lab Horizon Robotics Beijing China Jiangsu Thingstar Information Technology Co. Ltd. Nanjing China

Although deep learning (DL) based end-to-end models have shown outstanding performance in multi-channel speech extraction, their practical applications on edge devices are restricted due to their high computational complexity. In this paper, we propose a hybrid system that can more effectively integrate the generalized sidelobe canceller (GSC) and a lightweight post-filtering model under the assistance of spatial speaker activity information provided by a directional voice activity detection (DVAD) module. In addition to guiding the update of the adaptive blocking matrix (ABM) and the adaptive interference canceller (AIC) used in GSC to alleviate the distortion of the desired speech, DVAD is also utilized as an auxiliary input to the post-filtering model to enhance its capability of interference suppression. The experimental results demonstrate that, with much lower computational costs, our method can achieve comparable performance with a current state-of-the-art end-to-end model on simulated data and generalize even better on real-world data.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Memristor-Based Input Delay Reservoir Computing System for Temporal Signal Prediction

SSRN

引用

SSRN 2024年

作者： Wang, Sui-Dong Lu, Zhen-Ni Ye, Jing-Ting Zhang, Zhong-Da Cai, Jia-Wei Pan, Xiang-Yu Xu, Jian-Long Gao, Xu Zhong, Ya-Nan Jiangsu Key Laboratory for Carbon-Based Functional Materials & Devices Soochow University Jiangsu Suzhou215123 China MUST-SUDA Joint Research Center for Advanced Functional Materials Macau University of Science and Technology Taipa999078 China

Reservoir computing (RC) system, featured by its recursive structure, has been utilized for temporal signal processing, offering both low power consumption and high computational speed. This work reports on a novel input delay reservoir computing (ID-RC) system based on the oxide memristors, which can be applied to temporal signal prediction. The particle swarm optimization (PSO) algorithm is employed in the ID-RC system to obtain optimal hyperparameters for multi-step prediction in the Mackey-Glass task, with a normalized root-mean-square error (NRMSE) of only 0.09 at the 20th step. Significantly, by employing the ID-RC system in temporal signal prediction of the Hénon map and the nonlinear autoregressive moving average (NARMA10), small NRMSEs of 0.047 and 0.017 were achieved, respectively. The memristor-based ID-RC system turns out to be highly promising in forecasting of chaotic time series. © 2024, The Authors. All rights reserved.

关键词： Mean square error

来源：评论

学校读者我要写书评

暂无评论

A New Dynamic Routing Network for Monocular Depth Estimation 12

A New Dynamic Routing Network for Monocular Depth Estimation

引用

12th International Conference on CYBER technology in Automation, Control, and Intelligent Systems, CYBER 2022

作者： Luo, Zhehao Luo, Sijin Liang, Guoyuan Wu, Xinyu Guangdong Provincial Key Lab of Robotics and Intelligent System Shenzhen Institute of Advanced Technology Chinese Academy of Sciences Guangdong Province Shenzhen China University of Chinese Academy of Sciences Beijing China Guangdong-Hong Kong-Macao Joint Laboratory of Human-Machine Intelligence-Synergy Systems Chinese Academy of Sciences Guangdong Province Shenzhen China Guangdong Provincial Key Laboratory of Computer Vision and Virtual Reality Technology Shenzhen Institute of Advanced Technology Chinese Academy of Sciences Guangdong Province Shenzhen China

ISBN: (数字)9781665472678

ISBN: (纸本)9781665472678

Depth estimation is an essential task for understanding the geometry of 3D scenes. Compared with multi-view-based methods, monocular depth estimation is more challenging for the requirement of integrating not only global information but also local cues from various parts. Recently, numerous Convolutional Neural Networks (CNNs) based approaches relying on pre-defined static architectures have been reported. Although many of them have achieved remarkable improvements in accuracy, there remains problem with the generality of the static model when it comes to diverse scenes. In this paper, a novel deep network with dynamic routing architecture is proposed with a dynamically adjustable structure with respect to different inputs, therefore leading to not only improvements on both accuracy and generality but also decrease of the model parameters. Extensive experiments of monocular depth estimation have been conducted on two well-known challenging datasets (i.e., NYUD-v2 and Make3D) as well as the mixed dataset, and have verified the effectiveness of the proposed dynamic framework. © 2022 IEEE.

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

Design and Evaluation of a Novel Self-Adaptive Ankle Rehabilitation Exoskeleton with Elastic Modules

Design and Evaluation of a Novel Self-Adaptive Ankle Rehabil...

引用

International Conference on advanced robotics and Mechatronics (ICARM)

作者： Meng Gao Junpeng Chen Mi Li Jian S. Dai Shenzhen Key Laboratory of Intelligent Robotics and Flexible Manufacturing Systems Southern University of Science and Technology Shenzhen China Centre for Robotics Research King's College London London United Kingdom Key Laboratory of Mechanism Theory and Equipment Design of the Ministry of Education Centre for Advanced Mechanisms and Robotics Tianjin University Tianjin China Chair Professor of Mechanisms and Robotics Centre for Robotics Research King's College London UK

This paper proposes a novel ankle rehabilitation exoskeleton based on the research of human gait cycle, aiming to assist patients with neurological injury to recover quickly. The mechanism consists of two frames equipped with ankle joints and toe joint. The lower limbs and musculature system are correlated to propose the model. Incorporating spring modules are provided on the instep, the bottom of toes and either sides of the ankles which cushions lower limbs and reduces energy consumption. Kinematics and stress analysis are conducted to verify the feasibility of the model. Besides, the simulation, which incorporates the elastic elements and springs, is presented. Finally, the modular facilitates could easy integration with other rehabilitation modules for a comprehensive system. Overall, this research provides theoretical and engineering technical support for developing lower limb rehabilitation robots with adaptive stability, diversified movement functions, and human-computer interaction safety and health under different conditions.

关键词：

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：