检索结果-内蒙古大学图书馆

Evaluating and improving the interpretability of item embeddings using item-tag relevance information

Frontiers of computer Science 2020年第3期14卷 143-158页

作者： Tao LIAN Lin DU Mingfu ZHAO Chaoran CUI Zhumin CHEN Jun MA College of Data Science Taiyuan University of TechnologyJinzhong 030600China Software College Shandong UniversityJinan 250101China School of Computer and Control Engineering University of Chinese Academy of SciencesBeijing 101408China School of Computer Science and Technology Shandong University of Finance and EconomicsJinan 250014China School of Computer Science and Technology Shandong UniversityQingdao 266237China

Matrix factorization(MF)methods have superior recommendation performance and are flexible to incorporate other side information,but it is hard for humans to interpret the derived latent ***,the item-item cooccurrence information is exploited to learn item embeddings and enhance the recommendation ***,the item-item co-occurrence information,constructed from the sparse and long-tail distributed user-item interaction matrix,is over-estimated for rare items,which could lead to bias in learned item *** this paper,we seek to evaluate and improve the interpretability of item embeddings by leveraging a dense item-tag relevance ***,we design two metrics to quantitatively evaluate the interpretability of item embeddings from different viewpoints:interpretability of individual dimensions of item embeddings and semantic coherence of local neighborhoods in the latent *** also propose a tag-informed item embedding(TIE)model that jointly factorizes the user-item interaction matrix,the item-item co-occurrence matrix and the item-tag relevance matrix with shared item embeddings so that different forms of information can co-operate with each other to learn better item *** on the MovieLens20M dataset demonstrate that compared with other state-of-the-art MF methods,TIE achieves better top-N recommendations,and the relative improvement is larger when the user-item interaction matrix becomes *** leveraging the itemtag relevance information,individual dimensions of item embeddings are more interpretable and local neighborhoods in the latent space are more semantically coherent;the bias in learned item embeddings are also mitigated to some extent.

关键词： recommender system matrix factorization item embedding item-tag relevance interpretability

来源：评论

学校读者我要写书评

暂无评论

UGNA-VPR: A Novel Training Paradigm for Visual Place Recognition Based on Uncertainty-Guided NeRF Augmentation

arXiv

引用

arXiv 2025年

作者： Shen, Yehui Zhang, Lei Li, Qingqiu Zhao, Xiongwei Wang, Yue Lu, Huimin Chen, Xieyuanli College of Intelligence Science and Technology National University of Defense Technology China Faculty of Robot Science and Engineering Northeastern University China School of Electronic and Information Engineering Harbin Institute of Technology Shenzhen China State Key Laboratory of Industrial Control Technology Institute of Cyber-Systems and Control Zhejiang University China SIASUN Robot& Automation Co. Ltd China School of Computer Science Fudan University China

Visual place recognition (VPR) is crucial for robots to identify previously visited locations, playing an important role in autonomous navigation in both indoor and outdoor environments. However, most existing VPR datasets are limited to single-viewpoint scenarios, leading to reduced recognition accuracy, particularly in multi-directional driving or feature-sparse scenes. Moreover, obtaining additional data to mitigate these limitations is often expensive. This paper introduces a novel training paradigm to improve the performance of existing VPR networks by enhancing multi-view diversity within current datasets through uncertainty estimation and NeRF-based data augmentation. Specifically, we initially train NeRF using the existing VPR dataset. Then, our devised self-supervised uncertainty estimation network identifies places with high uncertainty. The poses of these uncertain places are input into NeRF to generate new synthetic observations for further training of VPR networks. Additionally, we propose an improved storage method for efficient organization of augmented and original training data. We conducted extensive experiments on three datasets and tested three different VPR backbone networks. The results demonstrate that our proposed training paradigm significantly improves VPR performance by fully utilizing existing data, outperforming other training approaches. We further validated the effectiveness of our approach on self-recorded indoor and outdoor datasets, consistently demonstrating superior results. Our dataset and code have been released at https://***/nubot-nudt/UGNA-VPR. Copyright © 2025, The Authors. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Synergistically Learning Class-specific Tokens for Multi-class Whole Slide Image Classification

Synergistically Learning Class-specific Tokens for Multi-cla...

引用

IEEE International Conference on Bioinformatics and Biomedicine (BIBM)

作者： Pengzhong Sun Wei Wang Xiangyu Li Suyu Dong Shuo Li Kuanquan Wang Gongning Luo Faculty of Computing Harbin Institute of Technology Harbin China Faculty of Computing Harbin Institute of Technology Shenzhen China College of Computer and Control Engineering Northeast Forestry University Harbin China Case School of Engineering Case Western Reserve University Cleveland USA

The application of transformer architecture in analyzing whole slide images (WSIs) has become increasingly popular due to its remarkable ability to learn complex associations. Nevertheless, a significant drawback emerges in the multiclass analysis of WSIs. The majority of the transformer-based methods available currently rely primarily on a single, class-agnostic token. This approach might not ideally capture the subtleties of class-discriminative information. To address this challenge, we present an innovative approach tailored for multi-class WSI analysis that harnesses the power of class-specific tokens. Central to our method is a novel attention mechanism designed to foster a synergistic learning relationship between patch and class tokens, enhancing the granularity of information captured and ensuring a more comprehensive representation of the WSI. Complementing this, we introduce a dynamic class-centric training strategy designed to optimize token representation learning, ensuring each token is informatively aligned with its corresponding class. Through extensive experimentation on three challenging multi-class WSI analysis datasets, our method consistently demonstrates superior performance, underscoring its potential as a robust solution for multi-class WSI analysis tasks.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Adaptive Positional Encoding for Bundle-Adjusting Neural Radiance Fields

Adaptive Positional Encoding for Bundle-Adjusting Neural Rad...

引用

International Conference on computer Vision (ICCV)

作者： Zelin Gao Weichen Dai Yu Zhang State Key Laboratory of Industrial Control Technology College of Control Science and Engineering Zhejiang University Key Laboratory of Brain Machine Collaborative Intelligence of Zhejiang Province and School of Computer Science Hangzhou Dianzi University Key Laboratory of Collaborative Sensing and Autonomous Unmanned Systems of Zhejiang Province

Neural Radiance Fields have shown great potential to synthesize novel views with only a few discrete image observations of the world. However, the requirement of accurate camera parameters to learn scene representations limits its further application. In this paper, we present adaptive positional encoding (APE) for bundle-adjusting neural radiance fields to reconstruct the neural radiance fields from unknown camera poses (or even intrinsics). Inspired by Fourier series regression, we investigate its relationship with the positional encoding method and therefore propose APE where all frequency bands are trainable. Furthermore, we introduce period-activated multilayer perceptrons (PMLPs) to construct the implicit network for the high-order scene representations and fine-grained gradients during backpropagation. Experimental results on public datasets demonstrate that the proposed method with APE and PMLPs can outperform the state-of-the-art methods in accurate camera poses and high-fidelity view synthesis.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Design of control Method for Improving Current Quality of Wind Power Grid-Connected Inverter 5

Design of Control Method for Improving Current Quality of Wi...

引用

5th International Conference on New Energy and Future Energy System, NEFES 2020

作者： Li, Y. Zhang, Z.Y. Wang, P.G. Zhuang, Y. Liu, S. Beijing Guodian Star Technology Co. Ltd Beijing100000 China College of Electronic Information Engineering Hebei University Baoding071002 China College of Electronic Information Engineering Hebei University Baoding071002 China School of Control and Computer Engineering North China Electric Power University Baoding071000 China

The problem of suppression of nonlinear random impulse disturbance in grid-connected wind power generation is studied in this paper. The mathematical model of grid side converter of wind power system is obtained by using Kirchhoff law. A feedback synchronization controller is designed to suppress the disturbance of current. With the help of Lyapunov stability theory, the effectiveness of the synchronization controller is verified. Further effectively restrain the nonlinear random disturbance caused by the current connected to the grid, so that the current generated by the wind turbine is synchronized with the current generated by the power grid. Finally, the simulation verification is carried out. The simulation results show that the effectiveness of the proposed control method. © Published under licence by IOP Publishing Ltd.

关键词： Synchronization

来源：评论

学校读者我要写书评

暂无评论

A Resilient Decentralized Learning Approach against Byzantine Attack via Reputation Evaluation

A Resilient Decentralized Learning Approach against Byzantin...

引用

Youth Academic Annual Conference of Chinese Association of Automation (YAC)

作者： Heng Zhang Wenting Wang Xuguo Jiao Ming Yang Xiaoming Wu Key Laboratory of Computing Power Network and Information Security Ministry of Education Shandong Computer Science Center Qilu University of Technology (Shandong Academy of Sciences) Jinan P. R. China Shandong Provincial Key Laboratory of Computer Networks Shandong Fundamental Research Center for Computer Science Jinan P. R. China College of Control Science and Engineering Zhejiang University Hangzhou China State Grid Shandong Electric Power Research Institute Jinan China School of Information and Control Engineering Qingdao University of Technology Qingdao China

ISBN: (数字)9798350379228

ISBN: (纸本)9798350390780

Machine learning is a key component of many applications, among which decentralized learning has attracted wide attention because of its cost-effectiveness and high efficiency. However, decentralized learning is vulnerable to various Byzantine attacks. Existing defense models face challenges in defending against complex attacks and are often constrained by strict network topologies. To address this issue, a parameter aggregation rule based on reputation evaluation (RPV) is proposed in this paper. This rule establishes a reputation model for neighboring nodes and continuously updates it based on their distance performance during each iteration. By applying the reputation model of neighboring nodes to the Beta distribution, their reliability is obtained. Based on the reliability of neighboring nodes, the set of reliable nodes for this iteration and select parameter values from them for aggregation are determined. Finally, we gradually eliminate the influence of Byzantine nodes. Our experimental results on the MNIST dataset demonstrate that the proposed algorithm is resilient to attacks from any number of Byzantine nodes and outperforms previous defense models in terms of network topology constraints, training accuracy, and computational costs.

关键词： Training Learning systems Accuracy Network topology Computational modeling Machine learning Classification algorithms

来源：评论

学校读者我要写书评

暂无评论

SSD-MonoDETR: Supervised Scale-aware Deformable Transformer for Monocular 3D Object Detection

arXiv

引用

arXiv 2023年

作者： He, Xuan Yang, Fan Yang, Kailun Lin, Jiacheng Fu, Haolong Wang, Meng Yuan, Jin Li, Zhiyong The College of Computer Science and Electronic Engineering Hunan University Changsha410082 China The School of Robotics Hunan University China410012 China The National Engineering Research Center of Robot Visual Perception and Control Technology Hunan University Changsha410082 China The School of Computer Science Hefei University of Technology Hefei230009 China

Transformer-based methods have demonstrated superior performance for monocular 3D object detection recently, which aims at predicting 3D attributes from a single 2D image. Most existing transformer-based methods leverage both visual and depth representations to explore valuable query points on objects, and the quality of the learned query points has a great impact on detection accuracy. Unfortunately, existing unsupervised attention mechanisms in transformers are prone to generate low-quality query features due to inaccurate receptive fields, especially on hard objects. To tackle this problem, this paper proposes a novel "Supervised Scale-aware Deformable Attention" (SSDA) for monocular 3D object detection. Specifically, SSDA presets several masks with different scales and utilizes depth and visual features to adaptively learn a scale-aware filter for object query augmentation. Imposing the scale awareness, SSDA could well predict the accurate receptive field of an object query to support robust query feature generation. Aside from this, SSDA is assigned with a Weighted Scale Matching (WSM) loss to supervise scale prediction, which presents more confident results as compared to the unsupervised attention mechanisms. Extensive experiments on the KITTI and Waymo Open datasets demonstrate that SSDA significantly improves the detection accuracy, especially on moderate and hard objects, yielding state-of-the-art performance as compared to the existing approaches. Our code will be made publicly available at https://***/mikasa3lili/SSD-MonoDETR. Copyright © 2023, The Authors. All rights reserved.

关键词： Object detection

来源：评论

学校读者我要写书评

暂无评论

Sinusoidal-Assisted Synchrosqueezing Transform: Algorithms and Biomedical Applications

SSRN

引用

SSRN 2023年

作者： Chen, Tao Zheng, Qian Xie, Lei Su, Hongye State Key Laboratory of Industrial Control Technology Zhejiang University Hangzhou310027 China College of Computer Science and Control Engineering Shenzhen Institutes of Advanced Technology Chinese Academy of Sciences Shenzhen518055 China Hangzhou Flexolink Technology Co. Ltd. Hangzhou China

Intermittent oscillation signals (IOSs) exist widely in biomedical systems. They are always contaminated by various randomness like noise and artifacts. Compared with traditional time-frequency analysis (TFA) methods, adaptive TFA methods show the significant advantages in separating IOSs. However, these methods are implemented by assuming that the signal is in a weakly noisy environment and have narrow-band limitation. Besides, it is difficult for these methods to alleviate the noise in the gap among time where IOSs happen. Therefore, in this paper, a new method called the sinusoidal-assisted synchrosqueezing transform (SASST) is proposed for separating IOSs in strongly noisy environment. The SASST method locates the IOSs with the aid of a sinusoidal signal. Such a sinusoidal signal fills the energy blank in the gap between time where IOSs happen. This enables that the ridge detection approach, which is the necessary tool for mode retrieval of SST-based methods, focuses on the assisted sinusoidal signal, rather than noise. Thus, the background noise in the gap is suppressed. Moreover, this sinusoidal-assisted framework can be introduced into any SST-based TFA methods. Depending on the characteristics of different biomedical signals, different SST-based methods are chosen to achieve better separation. As a result, SASST is capable of extracting both narrow- and wide-band IOSs in a strongly noisy environment. The effectiveness and advantages of the proposed algorithm are verified by both numerical simulations and real-world cases. Even in a strong noise environment of -10dB, SASST achieve output SNR of 12.66dB and 10.53dB in narrow- and wide-band simulation, respectively. © 2023, The Authors. All rights reserved.

关键词： Medical applications

来源：评论

学校读者我要写书评

暂无评论

A Novel Relocating Recognition Algorithm in Overlapping Region for Multi-UAVs Cluster

A Novel Relocating Recognition Algorithm in Overlapping Regi...

引用

2022 IEEE International Conference on Unmanned Systems, ICUS 2022

作者： Li, Chen Qi, Xuelei Zhao, Juntao Liao, Xuanhong Lin, Zitian Ma, Hongjun Liang, Xinkai College of Information Science and Engineering Northeastern University Shenyang China Hong Kong University of Science and Technology Department of Electronic and Computer Engineering Hong Kong Hong Kong High School Attached to Northeast Normal University International Curriculum Center Changchun China School of Automation Science and Engineering South China University of Technology Guangzhou China Beijing Electro-Mechanical Engineering Institute Science and Technology on Complex System Control and Intelligent Agent Cooperation Laboratory Beijing China

ISBN: (数字)9781665484565

ISBN: (纸本)9781665484565

How improve the mapping efficiency and location accuracy of the multi-UAV cluster based on the distributed SLAM technology is a significant problem in overlapping regions. Therefore, this paper mainly proposes a novel relocation recognition algorithm for the multi-UAV cluster. First, the autonomous localization algorithm based on visual-inertial fusion is proposed. Then, the algorithm for vocabulary updating and relocating is designed. Finally, practical testing is implemented to verify and improve the efficiency of exploration ability and the accuracy of the autonomous navigation and positioning in the complex overlapping regions. © 2022 IEEE.

关键词： Unmanned aerial vehicles (UAV)

来源：评论

学校读者我要写书评

暂无评论

A Method for Measuring and Calibrating the Thickness of Thin Films Based on Infrared Interference Technology

SSRN

引用

SSRN 2023年

作者： Sun, Jiaxing Li, Xinyu Zhang, Haojie Song, Jinlong Li, Zhisong College of Machine Shanghai Dianji University Shanghai201306 China School of Control and Computer Engineering North China Electric Power University Beijing102206 China Semiconductor Equipment Co. Shanghai201114 China

Transparent thin films are critical industrial components widely used in advanced optics, microelectronics and materials science, among other relevant fields. The rapid and stable measurement of sub-micron industrial film thickness is of particular importance. In this study, a novel method based on the principles of infrared interference and laser profilometry is proposed to calculate the periodogram of the original spectrum and accurately extract the maximum frequency. In addition, a high-precision, compact and low-cost thin-film thickness measurement and calibration system is developed. This system achieves accurate measurement of single-layer film thickness and adapts to correct errors caused by slight tilt due to mechanical vibration or misalignment during film placement. Experimental data show that, compared to conventional infrared interference thickness measurement systems, the use of the proposed maximum frequency extraction algorithm improves measurement accuracy by approximately 90% and stability by approximately 70%. Furthermore, even when measuring the thickness of films with inclined components, the use of the constructed thin film thickness measurement and calibration system provides equivalent results. © 2023, The Authors. All rights reserved.

关键词： Film thickness

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：