检索结果-内蒙古大学图书馆

2022 IEEE International Conference on Systems, Man, and Cybernetics, SMC 2022

作者： Dai, Zhigang Cai, Bolun Chen, Junying South China University of Technology Key Lab. of Big Data & Intell. Robot MoE School of Software Engineering Guangzhou China Tencent Wechat Ai Guangzhou China

ISBN: (数字)9781665452588

ISBN: (纸本)9781665452588

Momentum Contrast (MoCo) achieves great success for unsupervised visual representation learning. However, there are a lot of supervised and semi-supervised datasets, which are already lab.led. To fully utilize the lab.l annotations, we propose Unified Momentum Contrast (UniMoCo), which extends MoCo to support arbitrary ratios of lab.led data and unlab.led data training. Compared with MoCo, UniMoCo has two modifications as follows: (1) Different from a single positive pair in MoCo, we maintain multiple positive pairs on-the-fly by comparing the query lab.l to a lab.l queue. (2) We propose a Unified Contrastive (UniCon) loss to support an arbitrary number of positives and negatives in a unified pair-wise optimization perspective. Our UniCon is more reasonable and powerful than the supervised contrastive loss in theory and practice. In our experiments, we pre-train multiple UniMoCo models with different ratios of ImageNet lab.ls and evaluate the performance on various downstream tasks. Experiment results show that UniMoCo generalizes well for unsupervised, semi-supervised and fully-supervised visual representation learning. Besides, we surprisingly find that UniMoCo performs best with 60% ImageNet lab.ls for COCO and VOC transfer learning. The code is availab.e: https://***/dddzg/unimoco. © 2022 IEEE.

关键词： Momentum

来源：评论

学校读者我要写书评

暂无评论

Semi-Supervised Learning via Weight-aware Distillation under Class Distribution Mismatch

Semi-Supervised Learning via Weight-aware Distillation under...

引用

International Conference on Computer Vision (ICCV)

作者： Pan Du Suyun Zhao Zisen Sheng Cuiping Li Hong Chen Key Lab of Data Engineering and Knowledge Engineering of MOE Renmin University of China Renmin University of China Beijing China

Semi-Supervised Learning (SSL) under class distribution mismatch aims to tackle a challenging problem wherein unlab.led data contain lots of unknown categories unseen in the lab.led ones. In such mismatch scenarios, traditional SSL suffers severe performance damage due to the harmful invasion of the instances with unknown categories into the target classifier. In this study, by strict mathematical reasoning, we reveal that the SSL error under class distribution mismatch is composed of pseudo-lab.ling error and invasion error, both of which jointly bound the SSL population risk. To alleviate the SSL error, we propose a robust SSL framework called Weight-Aware Distillation (WAD) that, by weights, selectively transfers knowledge beneficial to the target task from unsupervised contrastive representation to the target classifier. Specifically, WAD captures adaptive weights and high-quality pseudo-lab.ls to target instances by exploring point mutual information (PMI) in representation space to maximize the role of unlab.led data and filter unknown categories. Theoretically, we prove that WAD has a tight upper bound of population risk under class distribution mismatch. Experimentally, extensive results demonstrate that WAD outperforms five state-of-the-art SSL approaches and one standard baseline on two benchmark datasets, CIFAR10 and CIFAR100, and an artificial cross-dataset. The code is availab.e at https://***/RUC-DWBI-ML/research/tree/main/WAD-master.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Source separation improves music emotion recognition 14

Source separation improves music emotion recognition

引用

2014 4th ACM International Conference on Multimedia Retrieval, ICMR 2014

作者： Xu, Jieping Li, Xirong Hao, Yun Yang, Gang Multimedia Computing Lab. School of Information Renmin University of China Beijing 100872 China Key Lab of Data Engineering and Knowledge Engineering Renmin University of China Beijing 100872 China Graduate School of Library and Information Science University of Illinois at Urbana-Champaign United States

ISBN: (纸本)1595930361

Despite the impressive progress in music emotion recognition, it remains unclear what aspect of a song, i.e., singing voice and accompanied music, carries more emotional information. As an initial attempt to answer the question, we introduce source separation into a standard music emotion recognition system. This allows us to compare systems with and without source separation, and consequently reveal the influence of singing voice and accompanied music on emotion recognition. Classification experiments on a set of 267 songs with *** annotations verify the new finding that source separation improves song music emotion recognition. Copyright is held by the owner/author(s).

关键词： Source separation

来源：评论

学校读者我要写书评

暂无评论

Semi-Supervised Learning via Weight-aware Distillation under Class Distribution Mismatch

arXiv

引用

arXiv 2023年

作者： Du, Pan Zhao, Suyun Sheng, Zisen Li, Cuiping Chen, Hong Key Lab of Data Engineering and Knowledge Engineering of MOE Renmin University of China China Renmin University of China Beijing China

关键词： Distillation

来源：评论

学校读者我要写书评

暂无评论

Superclass Learning with Representation Enhancement

Superclass Learning with Representation Enhancement

引用

Conference on Computer Vision and Pattern Recognition (CVPR)

作者： Zeyu Gan Suyun Zhao Jinlong Kang Liyuan Shang Hong Chen Cuiping Li Key Lab of Data Engineering and Knowledge Engineering of MOE Renmin University of China Beijing China Renmin University of China Beijing China

In many real scenarios, data are often divided into a handful of artificial super categories in terms of expert knowledge rather than the representations of images. Concretely, a superclass may contain massive and various raw categories, such as refuse sorting. Due to the lack of common semantic features, the existing classification techniques are intractable to recognize superclass without raw class lab.ls, thus they suffer severe performance damage or require huge annotation costs. To narrow this gap, this paper proposes a superclass learning framework, called SuperClass Learning with Representation Enhancement(SCLRE), to recognize super categories by leveraging enhanced representation. Specifically, by exploiting the self-attention technique across the batch, SCLRE collapses the boundaries of those raw categories and enhances the representation of each superclass. On the enhanced representation space, a superclassaware decision boundary is then reconstructed. Theoretically, we prove that by leveraging attention techniques the generalization error of SCLRE can be bounded under superclass scenarios. Experimentally, extensive results demonstrate that SCLRE outperforms the baseline and other contrastive-based methods on CIFAR-100 datasets and four high-resolution datasets.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Context-aware recommendation using GPU based parallel tensor decomposition

Lecture Notes in Computer Science (including subseries Lectu...

引用

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 2014年 8933卷 213-226页

作者： Zou, Benyou Lan, Mengwei Li, Cuiping Tan, Liwen Chen, Hong Key Lab of Data Engineering and Knowledge Engineering of MOE Renmin University of China Beijing100872 China School of Information Renmin University of China Beijing100872 China

Recommender system plays an important role in many practical applications that help users to deal with information overload and provide personalized recommendations to them. The context in which a choice is made has been recognized as an important factor for recommendation systems. Recently, researchers extend the classical matrix factorization and allows for a generic integration of contextual information by modeling the data as a tensor. However, current tensor factorization methods suffer from the limitation that the computing cost can be very high in practice. In this paper, we propose GALS, a GPU based parallel tensor factorization algorithm, to accelerate the tensor factorization on large data sets to support the efficient context-aware recommendation. Experiments show that the proposed method can achieve 10 times faster than the current tensor factorization methods. © Springer International Publishing Switzerland 2014.

关键词： Collab.rative filtering

来源：评论

学校读者我要写书评

暂无评论

Location-based recommendation using incremental tensor factorization model

Lecture Notes in Computer Science (including subseries Lectu...

引用

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 2014年 8933卷 227-238页

作者： Zou, Benyou Zou, Benyou Li, Cuiping Tan, Liwen Tan, Liwen Chen, Hong Key Lab of Data Engineering and Knowledge Engineering of MOE Renmin University of China Beijing100872 China School of Information Renmin University of China Beijing100872 China

Newly emerging location-based online social services, such as Meetup and Douban, have experienced increased popularity and rapid growth. The classical Matrix Factorization methods usually only consider the user-item matrix. Recently, Researchers have extended the matrix adding location context as a tensor and used the Tensor Factorization methods for this scenario. However, in real scenario, the users and events are changing over time, the classical Tensor Factorization methods suffers the limitation that it can only be applied for static settings. In this paper, we propose a general Incremental Tensor Factorization model, which models the appearance changes of a tensor by adaptively updating its previous factorized components rather than recomputing them on the whole data every time the data changed. Experiments show that the proposed methods can offer more effective recommendations than baselines, and significantly improve the efficiency of providing location recommendations. © Springer International Publishing Switzerland 2014.

关键词： Location based services

来源：评论

学校读者我要写书评

暂无评论

Contrastive Coding for Active Learning under Class Distribution Mismatch

Contrastive Coding for Active Learning under Class Distribut...

引用

International Conference on Computer Vision (ICCV)

作者： Pan Du Suyun Zhao Hui Chen Shuwen Chai Hong Chen Cuiping Li Key Lab of Data Engineering and Knowledge Engineering of MOE Renmin University of China Beijing China Renmin University of China Beijing China

ISBN: (纸本)9781665428132

Active learning (AL) is successful based on the assumption that lab.led and unlab.led data are obtained from the same class distribution. However, its performance deteriorates under class distribution mismatch, wherein the un-lab.led data contain many samples out of the class distribution of lab.led data. To effectively handle the problems under class distribution mismatch, we propose a contrastive coding based AL framework named CCAL. Unlike the existing AL methods that focus on selecting the most informative samples for annotating, CCAL extracts both semantic and distinctive features by contrastive learning and combines them in a query strategy to choose the most informative un-lab.led samples with matched categories. Theoretically, we prove that the AL error of CCAL has a tight upper bound. Experimentally, we evaluate its performance on CIFAR10, CIFAR100, and an artificial cross-dataset that consists of five datasets; consequently, CCAL achieves state-of-the-art performance by a large margin with remarkably lower annotation cost. To the best of our knowledge, CCAL is the first work related to AL for class distribution mismatch.

关键词： Computer vision Costs Upper bound Annotations Semantics Text categorization Computer architecture

来源：评论

学校读者我要写书评

暂无评论

What matters for ad-hoc video search? A large-scale evaluation on TRECVID

arXiv

引用

arXiv 2021年

作者： Chen, Aozhu Hu, Fan Wang, Zihan Zhou, Fangming Li, Xirong AIMC Lab. School of Information Renmin University of China Key Lab of Data Engineering and Knowledge Engineering Renmin University of China

For quantifying progress in Ad-hoc Video Search (AVS), the annual TRECVID AVS task is an important international evaluation. Solutions submitted by the task participants vary in terms of their choices of cross-modal matching models, visual features and training data. As such, what one may conclude from the evaluation is at a high level that is insufficient to reveal the influence of the individual components. In order to bridge the gap between the current solution-level comparison and the desired component-wise comparison, we propose in this paper a large-scale and systematic evaluation on TRECVID. By selected combinations of state-of-the-art matching models, visual features and (pre-)training data, we construct a set of 25 different solutions and evaluate them on the TRECVID AVS tasks 2016–2020. The presented evaluation helps answer the key question of what matters for AVS. The resultant observations and learned lessons are also instructive for developing novel AVS solutions. Copyright © 2021, 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

Scientific Value Weights more than Being Open or Toll Access:An analysis of the OA advantage in Nature and Science

引用

Journal of data and Information Science 2021年第4期6卷 62-75页

作者： Howell Y.Wang Shelia X.Wei Cong Cao Xianwen Wang Fred Y.Ye International Joint Informatics Laboratory&Jiangsu Key Laboratory of Data Engineering and Knowledge Service School of Information ManagementNanjing UniversityNanjing 210023China Nottingham University Business School China University of Nottingham Ningbo ChinaNingbo315100China WISE Lab. Dalian University of TechnologyDalian 116024China

Purpose:We attempt to find out whether OA or TA really affects the dissemination of scientific ***/methodology/approach:We design the indicators,hot-degree,and R-index to indicate a topic OA or TA ***,according to the OA classification of the Web of Science(WoS),we collect data from the WoS by downloading OA and TA articles,letters,and reviews published in Nature and Science during 2010–*** papers are divided into three broad disciplines,namely biomedicine,physics,and ***,taking a discipline in a journal and using the classical Latent Dirichlet Allocation(LDA)to cluster 100 topics of OA and TA papers respectively,we apply the Pearson correlation coefficient to match the topics of OA and TA,and calculate the hot-degree and R-index of every OA-TA topic ***,characteristics of the discipline can be *** qualitative comparison,we choose some high-quality papers which belong to Nature remarkable papers or Science breakthroughs,and analyze the relations between OA/TA and citation ***:The result shows that OA hot-degree in biomedicine is significantly greater than that of TA,but significantly less than that of TA in *** on the R-index,it is found that OA advantages exist in biomedicine and TA advantages do in ***,the dissemination of average scientific discoveries in all fields is not necessarily affected by OA or ***,OA promotes the spread of important scientific discoveries in high-quality *** limitations:We lost some citations by ignoring other open sources such as arXiv and *** limitation came from that Nature employs some strong measures for access-promoting subscription-based articles,on which the boundary between OA and TA became *** implications:It is useful to select hot topics in a set of publications by the hotdegree *** finding comprehensively reflects the differences of OA and TA in different disciplines,which is a u

关键词： Open access Toll access Scientific discovery Academic dissemination data analysis

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：