检索结果-内蒙古大学图书馆

2010 International Conference on Artificial Intelligence and Computational Intelligence, AICI 2010

作者： Xie, Jiongyun Lu, Hongtao Nan, Deng Nengbin, Cai MOE-MS Key Laboratory for Intelligent Computing and Intelligent Systems Dept. of Computer Science Shanghai Jiaotong University Shanghai China Shanghai Forensic Center China

ISBN: (纸本)364216529X

It has been shown that the Deep Belief Network is good at modeling input distribution, and can be trained efficiently by the greedy layer-wise unsupervised learning. Hoglak Lee et al. (2008) introduced a sparse variant of the Deep Belief Network, which applied the Gaussian linear units to model the input data with a sparsity constraint. However, it takes much more weight updates to train the RBM (Restricted Boltzmann Machine) with Gaussian visible units, and the reconstruction error is much larger than training an RBM with binary visible units. Here, we propose another version of Sparse Deep Belief Net which applies the differentiable sparse coding method to train the first level of the deep network, and then train the higher layers with RBM .This hybrid model, combining the advantage of the Deep architecture and the sparse coding model, leads to state-of-the-art performance on the classification of handwritten digits. © 2010 Springer-Verlag.

关键词： Character recognition

来源：评论

学校读者我要写书评

暂无评论

A parallel skeleton library for embedded multicores

A parallel skeleton library for embedded multicores

引用

International Workshop on Scheduling and Resource Management for Parallel and Distributed systems

作者： Liu, Xin Zhou, Jingyu Zhang, Daqiang Shen, Yao Guo, Minyi Department of Computer Science MOE-MS Key Laboratory for Intelligent Computing and Intelligent Systems Shanghai Jiao Tong University 800 Dongchuan Road Shanghai 200240 China

ISBN: (纸本)9780769541570

Many SoCs adopt multicore architectures. As a result, embedded programmers are also facing the challenge of parallel programming. We propose a parallel skeleton library that can be used on embedded multicores. Our library is implemented in standard C++ using template features. We propose two parallel skeletons to support common program patterns on multicores. In our skeleton library, programmers can easily choose underlying parallel implementations with no code changes. Experimental results show that many applications can take advantage of these two skeletons for performance improvement, sometimes better than hand-parallelized code. © 2010 IEEE.

关键词： C++ (programming language)

来源：评论

学校读者我要写书评

暂无评论

Flatten hierarchies for large-scale hierarchical text categorization

Flatten hierarchies for large-scale hierarchical text catego...

引用

International Conference on Digital Information Management

作者： Wang, Xiao-Lin Lu, Bao-Liang Center for Brain-Like Computing and Machine Intelligence Department of Computer Science and Engineering China MOE-Microsoft Key Laboratory for Intelligent Computing and Intelligent Systems Shanghai Jiao Tong University China

ISBN: (纸本)9781424475728

Hierarchies are very popular in organizing documents and web pages, hence automated hierarchical classification techniques are desired. However, the current dominant hierarchical approach of top-down method suffers accuracy decrease compared with flat classification approaches, because of error propagation and bottom nodes' data sparsity. In this paper we flatten hierarchies to relieve such accuracy decrease in top-down method, which aims to make hierarchies both effective enough to make large-scale classification tasks feasible, and simple enough to ensure high classification accuracy. We propose two flattening strategies based on these two causes of the accuracy decrease, and experimental results show that the flattening strategy designed for error propagation is more effective, which suggests that hierarchies with lots of branches at top layers can provide high classification accuracy. Besides, we analyze the computational complexity before and after flattening, which approximately agree with the experimental results. ©2010 IEEE.

关键词： Websites

来源：评论

学校读者我要写书评

暂无评论

A unified character-based tagging framework for chinese word segmentation

引用

ACM Transactions on Asian Language Information Processing 2010年第2期9卷 1–32页

作者： Zhao, Hai Huang, Chang-Ning Li, Mu. Lu, Bao-Liang Department of Computer Science and Engineering MOE-Microsoft Key Laboratory for Intelligent Computing and Intelligent Systems Shanghai Jiao Tong University 800 Dongchuan Road Minhang District Shanghai 200240 China Language Computing Group Microsoft Research Asia 49 Zhichun Road Haidian District Beijing 100080 China

Chinese word segmentation is an active area in Chinese language processing though it is suffering from the argument about what precisely is a word in Chinese. Based on corpus-based segmentation standard, we launched this study. In detail, we regard Chinese word segmentation as a character-based tagging problem. We show that there has been a potent trend of using a character-based tagging approach in this field. In particular, learning from segmented corpus with or without additional linguistic resources is treated in a unified way in which the only difference depends on how the feature template set is selected. It differs from existing work in that both feature template selection and tag set selection are considered in our approach, instead of the previous feature template focus only technique. We show that there is a significant performance difference as different tag sets are selected. This is especially applied to a six-tag set, which is good enough for most current segmented corpora. The linguistic meaning of a tag set is also discussed. Our results show that a simple learning system with six n-gram feature templates and a six-tag set can obtain competitive performance in the cases of learning only from a training corpus. In cases when additional linguistic resources are available, an ensemble learning technique, assistant segmenter, is proposed and its effectiveness is verified. Assistant segmenter is also proven to be an effective method as segmentation standard adaptation that outperforms existing ones. Based on the proposed approach, our system provides state-of-the-art performance in all 12 corpora of three international Chinese word segmentation bakeoffs. © 2010 ACM 1530-0226/2010/06-ART5 $10.00.

关键词： Learning systems

来源：评论

学校读者我要写书评

暂无评论

Comparative analysis of motor imagery on different scales based on brain computer interface

Comparative analysis of motor imagery on different scales ba...

引用

作者： Li, Junhua Zhang, Liqing Department of Computer Science and Engineering MOE-Microsoft Key Laboratory for Intelligent Computing and Intelligent Systems Shanghai Jiao Tong University Shanghai 200240 China

The event-related brain oscillatory responses have a substantial role in motor imagery experiments. There are two patterns, i.e. event-related desynchronization (ERD) and eventrelated synchronization (ERS). In clinical applications and scientific experiments, an obvious event-related potential (ERP) is pivotal to achieve required tasks. Therefore, we explore how subjects generate the obvious ERP more easily. Eight healthy subjects participated in the experiments and completed motor imagery on two different scales (index finger versus arm). The results indicate that the motor imagery of arm reaches the better performance compared with the motor imagery of index finger. Furthermore, we make a comparative analysis in averaged power spectrum in order to reveal inherent phenomenon. The analysis of power spectrum further confirms that arm motor imagery has an advantage over index finger motor imagery.

关键词： Brain computer interface

来源：评论

学校读者我要写书评

暂无评论

A Study on Recommendation Features for an RSS Reader

A Study on Recommendation Features for an RSS Reader

引用

International Conference on Cyber-Enabled Distributed computing and Knowledge Discovery, CyberC

作者： Cansheng Ji Jingyu Zhou School of Software MOE-MS Key Laboratory for Intelligent Computing and Intelligent Systems Shanghai Jiaotong University Shanghai China

In the era of web 2.0, everyone can create and update content, and everyone can host a personal web site with little effort, making it hard to gather valuable information from different web sites. With RSS, people can read information from different resources in a uniform way, and in a single tool, such as an RSS reader. However, most of the RSS readers only display items in chronological order, which doesn't work well when users are inundated with too many items in the feeds. We propose using recommendation to help people find items in an RSS reader. Specifically, we consider profiled based features (i.e., text similarity and favorite fraction), update frequency, as well as Post Rank values for RSS recommendation. Experimental results indicate that favorite fraction and update frequency perform better than text similarity. Additionally, we also study the effect of feature combination and find that the combination of similarity and favorite fraction performs the best.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Vigilance Analysis Based on Continuous Wavelet Transform of EEG Signals

Vigilance Analysis Based on Continuous Wavelet Transform of ...

引用

International Conference on Biomedical Engineering and Computer Science

作者： Tian Ouyang Hong-Tao Lu MOE-MS Key Laboratory for Intelligent Computing and Intelligent Systems Department of Computer Science and Engineering Shanghai Jiao Tong University

ISBN: (纸本)9781424453153

Electroencephalography (EEG) is considered a reliable indicator of a person's vigilance level. In this paper, we use EEG recordings to discriminate three vigilance states of a person, namely alert, drowsy, and sleep, while driving a car in a simulation environment. The proposed framework explores the use of continuous wavelet transform in EEG signal processing. A large set of features is extracted from the wavelet coefficients, which are computed from EEG signals with multiple wavelet functions. We use random forest to rank the plenty of features and select the most important ones for later classification. Samples of EEG data are then trained and classified by SVM (Support Vector Machine). On datasets acquired from 5 subjects, our method reveals high classification accuracy (over 96%).

关键词： Electroencephalography (EEG) Vigilance Alert Drowsy Sleep Continuous Wavelet Transform Random Forest SVM

来源：评论

学校读者我要写书评

暂无评论

A Parallel Skeleton Library for Embedded Multicores

A Parallel Skeleton Library for Embedded Multicores

引用

International Conference on Parallel Processing Workshops (ICPPW)

作者： Xin Liu Jingyu Zhou Daqiang Zhang Yao Shen Minyi Guo EPCC Department of Computer Science MOE-MS Key Laboratory of Intelligent Computing and Intelligent Systems Shanghai Jiaotong University Shanghai China

关键词： Skeleton Libraries Multicore processing Kernel System-on-a-chip Runtime Computational modeling

来源：评论

学校读者我要写书评

暂无评论

Hedge Detection and Scope Finding by Sequence Labeling with Normalized Feature Selection 14

Hedge Detection and Scope Finding by Sequence Labeling with ...

引用

14th Conference on Computational Natural Language Learning, CoNLL 2010

作者： Zhang, Shaodian Zhao, Hai Zhou, Guodong Lu, Bao-Liang Center for Brain-Like Computing and Machine Intelligence Dept of Computer Science and Engineering Shanghai Jiao Tong University China MOE-Microsoft Key Laboratory for Intelligent Computing and Intelligent Systems Shanghai Jiao Tong University China School of Computer Science and Technology Soochow University China

ISBN: (纸本)1932432841

This paper presents a system which adopts a standard sequence labeling technique for hedge detection and scope finding. For the first task, hedge detection, we formulate it as a hedge labeling problem, while for the second task, we use a two-step labeling strategy, one for hedge cue labeling and the other for scope finding. In particular, various kinds of syntactic features are systemically exploited and effectively integrated using a large-scale normalized feature selection method. Evaluation on the CoNLL-2010 shared task shows that our system achieves stable and competitive results for all the closed tasks. Furthermore, post-deadline experiments show that the performance can be much further improved using a sufficient feature selection. © 2010 Association for Computational Linguistics

关键词： Wooden fences

来源：评论

学校读者我要写书评

暂无评论

A comparative study on two large-scale hierarchical text classification tasks' solutions

A comparative study on two large-scale hierarchical text cla...

引用

International Conference on Machine Learning and Cybernetics

作者： Zhang, Jian Zhao, Hai Lu, Bao-Liang Department of Computer Science and Engineering Center for Brain-Like Computing and Machine Intelligence Shanghai Jiao Tong University 800 Dong Chuan Road Shanghai 200240 China MOE-Microsoft Key Laboratory for Intelligent Computing and Intelligent Systems Shanghai Jiao Tong University 800 Dong Chuan Road Shanghai 200240 China

ISBN: (纸本)9781424465262

Patent classification is a large scale hierarchical text classification (LSHTC) task. Though comprehensive comparisons, either learning algorithms or feature selection strategies, have been fully made in the text categorization field, few work was done for a LSHTC task due to high computational cost and complicated structural label characteristics. For the first time, this paper compares two popular learning frameworks, namely hierarchical support vector machine (SVM) and k nearest neighbor (k-NN) that are applied to a LSHTC task. Experiment results show that the latter outperforms the former in this LSHTC task, which is quite different from the usual results for normal text categorization tasks. Then this paper does a comparative study on different similarity measures and ranking approaches in k-NN framework for LSHTC task. Conclusions can be drawn that k-NN is more appropriate for the LSHTC task than hierarchical SVM and for a specific LSHTC task. BM25 outperforms other similarity measures and List Weak gains a better performance than other ranking approaches. We also find an interesting phenomenon that using all the labels of the retrieved neighbors can remarkably improve classification performance over only using the first label of the retrieved neighbors. © 2010 IEEE.

关键词： Learning algorithms

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：