检索结果-内蒙古大学图书馆

38th Conference on Neural Information Processing systems, NeurIPS 2024

作者： Zhang, Di Lv, Bowen Zhang, Hai Yang, Feifan Zhao, Junqiao Yu, Hang Huang, Chang Zhou, Hongtu Ye, Chen Jiang, Changjun Department of Computer Science Tongji University Shanghai China MOE Key Lab of Embedded System and Service Computing Tongji University Shanghai China

A primary challenge for visual-based Reinforcement Learning (RL) is to generalize effectively across unseen environments. Although previous studies have explored different auxiliary tasks to enhance generalization, few adopt image reconstruction due to concerns about exacerbating overfitting to task-irrelevant features during training. Perceiving the pre-eminence of image reconstruction in representation learning, we propose SMG (Separated Models for Generalization), a novel approach that exploits image reconstruction for generalization. SMG introduces two model branches to extract task-relevant and task-irrelevant representations separately from visual observations via cooperatively reconstruction. Built upon this architecture, we further emphasize the importance of task-relevant features for generalization. Specifically, SMG incorporates two additional consistency losses to guide the agent's focus toward task-relevant areas across different scenarios, thereby achieving free from overfitting. Extensive experiments in DMC demonstrate the SOTA performance of SMG in generalization, particularly excelling in video-background settings. Evaluations on robotic manipulation tasks further confirm the robustness of SMG in real-world applications. Source code is available at https://***/r/SMG/. © 2024 Neural information processing systems foundation. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

An efficient gene selection algorithm based on tolerance rough set theory 1

引用

12th International Conference on Rough Sets, Fuzzy Sets, Data Mining and Granular computing, RSFDGrC 2009

作者： Jiao, Na Miao, Duoqian Department of Computer Science and Technology Tongji University Shanghai 201804 China Key Laboratory of Embedded System and Service Computing Ministry of Education of China Tongji University Shanghai 201804 China

ISBN: (数字)9783642106460

ISBN: (纸本)3642106455

Gene selection, a key procedure of the discriminant analysis of microarray data, is to select the most informative genes from the whole gene set. Rough set theory is a mathematical tool for further reducing redundancy. One limitation of rough set theory is the lack of effective methods for processing real-valued data. However, most of gene expression data sets are continuous. Discretization methods can result in information loss. This paper investigates an approach combining feature ranking together with feature selection based on tolerance rough set theory. Compared with gene selection algorithm based on rough set theory, the proposed method is more effective for selecting high discriminative genes in cancer classification task. © 2009 Springer-Verlag Berlin Heidelberg.

关键词： Rough set theory

来源：评论

学校读者我要写书评

暂无评论

Quantized robust H-two filtering for Markovian jump linear systems over networks with nonaccessible mode information

引用

控制理论与应用（英文版） 2011年第4期9卷 505-512页

作者： Tao LIU Qijun CHEN Hao ZHANG Iko MIYAZAWA Department of Electronics and Information Engineering Tongji University Shanghai 201804 China Key Laboratory of Embedded System and Service Computing Tongji University Ministry of Education Shanghai 201804 China Department of Machine Control Technology Kanagawa Industrial Technology Center Kanagawa 243-0435 Japan

This paper is concerned with the problems of H-two filtering for discrete-time Markovian jump linear systems subject to logarithmic quantization. We assume that only the output of the system is available, and therefore the mode information is nonaccessible. In this paper, a mode-independent quantized H-two filter is designed such that filter error system is stochastically stable. To this end, sufficient conditions for the existence of an upper bound of H-two norm are presented in terms of linear matrix inequalities. Considering uncertainty of system matrices, a robust H-two filter is designed. The proposed method is also applicable to cover the case where the transition probability matrix is not exactly known but belongs to a given polytope. Finally, numerical examples are provided to demonstrate the effectiveness of the proposed approach.

关键词： Markovian jump linear systems Quantized robust H-two filter Mode independent Sector bound approach Linear matrix inequalities

来源：评论

学校读者我要写书评

暂无评论

CDMC2011 data mining competition: LEVIS group results overview

CDMC2011 data mining competition: LEVIS group results overvi...

引用

2012 International Conference on system Simulation, ICUSS 2012

作者： Zhao, Q.C. Liu, J.M. Wang, X. You, M.Y. Li, G.Z. MOE Key Laboratory of Embedded Systems and Service Computing Department of Control Science and Engineering Tongji University Shanghai 201804 China

ISBN: (纸本)9781849195386

The task for the data mining contest organized in conjunction with the ICONIP2011 conference was to learn three predictive models (i.e. a classifier) capable of distinguishing between different classes for three separate tasks. Different combinations of data preprocess, feature selection and classifier learning methods were tried to obtain the best results.

关键词： Intrusion detection

来源：评论

学校读者我要写书评

暂无评论

Task allocation for maximizing reliability of distributed computing systems using honeybee mating optimization

引用

Journal of systems and Software 2010年第11期83卷 2165-2174页

作者： Kang, Qin-Ma He, Hong Song, Hui-Min Deng, Rong Key Laboratory of Embedded System and Service Computing Ministry of Education Tongji University Shanghai 201804 China School of Information Engineering Shandong University at Weihai 180 Wenhua Xilu Weihai 264209 China School of Mathematics and Statistics Shandong University at Weihai Weihai 264209 China

This paper deals with the problem of task allocation (i.e,;to which processor should each task of an application be assigned) in heterogeneous distributed computing systems with the goal of maximizing the system reliability. The problem of finding an optimal task allocation is known to be NP-hard in the strong sense. We propose a new swarm intelligence technique based on the honeybee mating optimization (HBMO) algorithm for this problem. The HBMO based approach combines the power of simulated annealing, genetic algorithms with a fast problem specific local search heuristic to find the best possible solution within a reasonable computation time. We study the performance of the algorithm over a wide range of parameters such as the number of tasks, the number of processors, the ratio of average communication time to average computation time, and task interaction density of applications. The effectiveness and efficiency of our algorithm are demonstrated by comparing it with recently proposed task allocation algorithms for maximizing system reliability available in the literature. © 2010 Elsevier Inc. All rights reserved.

关键词： Reliability

来源：评论

学校读者我要写书评

暂无评论

Symptom selection for multi-label data of inquiry diagnosis in traditional Chinese medicine

引用

Science China(Information Sciences) 2013年第5期56卷 238-250页

作者： SHAO Huan LI GuoZheng LIU GuoPing WANG YiQin School of Computer Engineering and Science Shanghai University Department of Control Science and Engineering Key Laboratory of Ministry of Education for Service Computing and Embedded SystemsTongji University Laboratory of Information Access and Synthesis of TCM Four Diagnosis Shanghai University of Traditional Chinese Medicine

In traditional Chinese medicine(TCM) diagnosis,a patient may be associated with more than one syndrome tags,and its computer-aided diagnosis is a typical application in the domain of multi-label learning of high-dimensional *** is common that a great deal of symptoms can occur in traditional Chinese medical diagnosis,which affects the modeling of diagnostic *** selection entails choosing the smallest feature subset of relevant symptoms,and maximizing the generalization performance of the *** present there are rare researches on feature selection on multi-label data.A hybrid optimization technique is introduced to symptom selection for multi-label data in TCM diagnosis in this paper,and modeling is made by means of four multi-label learning algorithms like k nearest neighbors,*** compare the performance of the algorithm with the current popular dimension reduction algorithms like MEFS(embedded feature selection for multi-Label learning),MDDM(multi-label dimensionality reduction via dependence maximization) on the UCI Yeast gene functional data set and an inquiry diagnosis dataset of coronary heart disease(CHD).Experimental results show that the algorithm we present has significantly improved the *** particular,the improvement on the average precision for the classifier is up to 10.62% and 14.54%.Syndrome inquiry modeling of CHD in TCM is realized in this paper,providing effective reference for the diagnosis of CHD and analysis of other multi-label data.

关键词： multi-label learning feature selection high-dimensionality inquiry of traditional Chinese medicine coronary heart disease

来源：评论

学校读者我要写书评

暂无评论

Behavior-aware Trustworthiness Study of Networked Software

引用

International Journal of Computational Intelligence systems 2010年第5期3卷 542-552页

作者： Fang, Xianwen Jiang, Changjun Fan, Xiaoqin Department of Computer Science and Technology Tongji University Shanghai 201804 China Key Lab of Embedded System and Service Computing Ministry of Education Tongji University Shanghai 201804 China

The essence characteristics of software trustworthiness are software execution effect and behavior can be anticipated, which is an important index of software quality. Under the open and dynamic environments, some uncertainty factors cause the behavior of software to be uncontrolled, uncertainty and unpredictable. Behavior-aware networked software trustworthiness research methods are proposed in the paper. Firstly, we propose the analysis methods of the consistency between the inferred specification models with component specification model. Then, for analyzing the component interaction behavior, the behavior relativity analyzing method is presented based on Petri net. Finally, aimed to the outer factors, we analyze the behavioral congruence between theoretical composite models with dynamic behavior model based on running logs. Theoretical analysis and the example analysis indicated that this method is benefit to analyze the trustworthiness of networked software. © 2010, the authors.

关键词： consistency networked software Petri net software behavior trustworthiness

来源：评论

学校读者我要写书评

暂无评论

Synchronization of Multi-Weighted and Directed Network Under Pinning Impulsive Control

Synchronization of Multi-Weighted and Directed Network Under...

引用

2021 China Automation Congress, CAC 2021

作者： Hu, Xiaohan Liu, Xiwei Department of Computer Science and Technology Tongji University Shanghai201804 China Department of Computer Science and Technology Key Laboratory of Embedded System Service Computing Tongji University Shanghai201804 China

ISBN: (纸本)9781665426473

In this paper, multi-weighted and directed complex dynamic networks (MWDCDNs) under the control of hybrid impulses are analyzed. Two cases are considered. In the first case, impulses can be only applied on one fixed node. In the other case, impulses are applied on different nodes according to whether they are synchronized or de-synchronized in MWDCDNs with unbounded time delay. By limiting impulsive interval and Chebyshev distance between the normalized left eigenvectors (NLEVec) of multiple coupling matrices corresponding to the zero eigenvalue, we obtain some criteria for the synchronization of MWDCNs. Some numerical simulation are also given. © 2021 IEEE

关键词： Eigenvalues and eigenfunctions

来源：评论

学校读者我要写书评

暂无评论

Use the core clusters for the initialization of the clustering based on one-class support vector machine

Use the core clusters for the initialization of the clusteri...

引用

4th International Conference on Advances in Swarm Intelligence, ICSI 2013

作者： Gu, Lei Key Laboratory of Embedded System and Service Computing Ministry of Education Tongji University Shanghai 200092 China School of Computer Science and Technology Nanjing University of Posts and Telecommunications Nanjing 210023 China

ISBN: (纸本)9783642387142

The clustering method based on one-class support vector machine has been presented recently. Although this approach can improve the clustering accuracies, it often gains the unstable clustering results because some random datasets are employed for its initialization. In this paper, a novel initialization method based on the core clusters is used for the clustering algorithm based one-class support vector machine. The core clusters are gained by constructing the neighborhood graph and they are regarded as the initial datasets of the clustering algorithm based one-class support vector machine. To investigate the effectiveness of the proposed approach, several experiments are done on four datasets. Experimental results show that the new presented method can improve the clustering performance compared to the previous clustering algorithm based on one-class support vector machine and k-means approach. © 2013 Springer-Verlag Berlin Heidelberg.

关键词： Support vector machines

来源：评论

学校读者我要写书评

暂无评论

Correct after Answer: Enhancing Multi-Span Question Answering with Post-Processing Method

Correct after Answer: Enhancing Multi-Span Question Answerin...

引用

2024 Conference on Empirical Methods in Natural Language Processing, EMNLP 2024

作者： Lin, Jiayi Zhang, Chenyang Tong, Haibo Zhang, Dongyu Hong, Qingqing Hou, Bingxuan Wang, Junli Key Laboratory of Embedded System and Service Computing Tongji University Ministry of Education Shanghai201804 China Collaborative Innovation Center for Financial Network Security Tongji University Shanghai201804 China

ISBN: (纸本)9798891761681

Multi-Span Question Answering (MSQA) requires models to extract one or multiple answer spans from a given context to answer a question. Prior work mainly focuses on designing specific methods or applying heuristic strategies to encourage models to predict more correct predictions. However, these models are trained on gold answers and fail to consider the incorrect predictions. Through a statistical analysis, we observe that models with stronger abilities do not predict less incorrect predictions compared with other models. In this work, we propose Answering-Classifying-Correcting (ACC) framework, which employs a post-processing strategy to handle incorrect predictions. Specifically, the ACC framework first introduces a classifier to classify the predictions into three types and exclude "wrong predictions", then introduces a corrector to modify "partially correct predictions". Experiments on several MSQA datasets show that ACC framework significantly improves the Exact Match (EM) scores, and further analysis demostrates that ACC framework efficiently reduces the number of incorrect predictions, improving the quality of predictions. © 2024 Association for Computational Linguistics.

关键词： Prediction models

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：