检索结果-内蒙古大学图书馆

International Joint Conference on Neural Networks

作者： Yiqin Leng Li Zhang Jiwen Yang School of Computer Science and Technology & Provincial Key Laboratory for Computer Information Processing Technology Soochow University

ISBN: (纸本)9781479914821

Locally Linear Embedding (LLE) is a sort of powerful nonlinear dimensionality reduction algorithms. The basic idea behind the LLE method is that each data point and its neighbors lie on or close to a locally linear patch of the manifold if there is sufficient data. Then the local geometry of these patches is described by using linear coefficients which can reconstruct each data point from its neighbors. However, LLE operates in a batch way and its dimension reduction cannot be generalized to unseen samples. If a test sample arrives, LLE must run repeatedly and the former computational results are discarded. Thus, some incremental methods have been proposed for LLE to solve this problem. In these incremental methods, the neighbor number is globally fixed, which may result in selecting points from another linear space as neighbors. This paper presents LLE based on orthogonal matching pursuit (OMP) and applies it to classification tasks. In the classification tasks, dimensionality reduction on test samples is implemented by applying dimension reduction on training samples. The new LLE method could select a more appropriate neighbors from the selected neighbors. OMP is applied to not only LLE for training samples, but also the incremental learning of LLE for test samples. Compared with other linear incremental methods, experimental results show that the proposed method is promising.

关键词： Joints Conferences Neural networks

来源：评论

学校读者我要写书评

暂无评论

Feature Ensemble Learning based on Sparse Autoencoders for Image Classification

Feature Ensemble Learning based on Sparse Autoencoders for I...

引用

International Joint Conference on Neural Networks

作者： Yaping Lu Li Zhang Bangjun Wang Jiwen Yang School of Computer Science and Technology & Provincial Key Laboratory for Computer Information Processing Technology Soochow University

ISBN: (纸本)9781479914821

Deep networks are well known for their powerful function approximations. To train a deep network efficiently, greedy layer-wise pre-training and fine tuning are required. Typically, pre-training, aiming to initialize a deep network, is implemented via unsupervised feature learning, with multiple feature representations generated. However, in general only the last layer representation is to be employed because of its abstraction and compactness being the best with comparisons to the ones of lower layers. To make full use of the representations of all layers, this paper proposes a feature ensemble learning method based on sparse autoencoders for image classification. Specifically, we train three softmax classifiers by using the representations of different layers, instead of one classifier trained by applying the last layer representation. Of the three softmax classifiers, two are obtained by training stacked auto-encoders with fine tuning, and the other one is obtained by directly using a concatenation of two representations. To improve accuracy and stability of a single softmax classifier, the ensemble of multiple classifiers is considered, and some Naive Bayes combination rules are introduced to integrate the three classifiers. Experimental results on the MNIST and COIL datasets are presented, with comparisons to other classification methods.

关键词： Deep network Feature representation Feature ensemble Autoencoder Softmax Naive Bayes

来源：评论

学校读者我要写书评

暂无评论

Research on improved coding and decoding algorithm for fountain codes

Research on improved coding and decoding algorithm for fount...

引用

International Conference on Wireless Communications and Signal processing (WCSP)

作者： Zhifeng Zhu Liulei Zhou Jia Hou School of Electronics and Information Engineering Soochow University Jiangsu China Provincial Key Laboratory for Computer Information Processing Technology Soochow University Jiangsu China

ISBN: (纸本)9781479973408

Fountain code is a class of graph-based linear erasure codes, which can effectively solve the problems such as network congestion and feedback cracking for its characteristics of rateless and can resume when interrupted, and has a lower complexity of encoding and decoding. However, there are still some problems in the process of encoding and decoding, including the degree distribution structure may be destroyed, parameters of the generating matrix are not fixed, and it cannot recover source datas from the remaining encoded packets with no degree one. So, the basic theory of fountain codes from three aspects are introduced in this paper, i.e., degree distribution, encoding and decoding principles. Therefore the improved algorithms according to the above three aspects are presented. Simulation results show that the proposed algorithm is more efficient than the previous one.

关键词： Decoding Encoding Algorithm design and analysis Solitons Robustness Vectors Signal processing algorithms

来源：评论

学校读者我要写书评

暂无评论

Solving Unbalanced Problems in Similarity Learning using SVM Ensemble

Solving Unbalanced Problems in Similarity Learning using SVM...

引用

International Joint Conference on Neural Networks

作者： Peipei Xia Li Zhang School of Computer Science and Technology & Provincial Key Laboratory for Computer Information Processing Technology Soochow University

ISBN: (纸本)9781479914821

Similarity learning is one of the most fundamental notions in machine learning and pattern recognition. In real-world problems, the number of the paired-samples in similarity set is far less than the ones in dissimilarity set. In other word, there is an unbalanced problem in the paired-samples of similarity learning. This paper presents a scheme of SVM ensemble to solve it. In our scheme, we randomly select some of samples to construct paired-samples, not producing all the paired-samples, and introduces multiple classifiers to obtain higher stability and reliability. As a result, the SVM ensemble can effectively decrease the number of paired-samples in similarity learning and solve the unbalanced data learning to some degree. In the experiments, the SVM ensemble is compared with some classic unbalanced learning algorithms. The results on classification tasks show that the SVM ensemble gains better performance.

关键词： Training Support vector machines Sampling methods Databases Accuracy Iris Educational institutions

来源：评论

学校读者我要写书评

暂无评论

Hybridizing infeasibility driven and constrained-domination principle with MOEA/D for constrained multiobjective evolutionary optimization

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 8916卷 249-261页

作者： Lin, Huibiao Fan, Zhun Cai, Xinye Li, Wenji Wang, Sheng Li, Jian Zhang, Chengdian School of Engineering Shantou University Guangdong515063 China College of Computer Science and Technology Nanjing University of Aeronautics and Astronautics NanjingJiangsu210016 China College of Science Shantou University Guangdong515063 China Guangdong Provincial Key Laboratory of Digital Signal and Image Processing Techniques Shantou University Guangdong515063 China

This paper presents a novel multiobjective constraint handling approach, named as MOEA/D-CDP-ID, to tackle constrained optimization problems. In the proposed method, two mechanisms, namely infeasibility driven (ID) and constrained-domination principle (CDP) are embedded into a prominent multiobjective evolutionary algorithm called MOEA/D. Constraineddomination principle defined a domination relation of two solutions in constraint handling problem. Infeasibility driven preserves a proportion of marginally infeasible solutions to join the searching process to evolve offspring. Such a strategy allows the algorithm to approach the constraint boundary from both the feasible and infeasible side of the search space, thus resulting in gaining a Pareto solution set with better distribution and convergence. The efficiency and effectiveness of the proposed approach are tested on several well-known benchmark test functions. In addition, the proposed MOEA/D-CDP-ID is applied to a real world application, namely design optimization of the two-stage planetary gear transmission system. Experimental results suggest that MOEA/D-CDP-ID can outperform other state-of-the-art algorithms for constrained multiobjective evolutionary optimization. © Springer International Publishing Switzerland 2014.

关键词： Constraint handling

来源：评论

学校读者我要写书评

暂无评论

Towards the Independent Spanning Trees in Conditional BC Networks

Towards the Independent Spanning Trees in Conditional BC Net...

引用

2013 International Conference on Advanced computer Science and Electronics information(ICACSEI 2013)

作者： Baolei Cheng Jianxi Fan Shukui Zhang Yan Wang Xi Wang Wenjun Liu School of Computer Science and Technology Soochow University Provincial Key Laboratory for Computer Information Processing Technology Soochow University

Independent spanning trees (ISTs for short) in networks have applications such as reliable communication protocols, the multi-node broadcasting, one-to-all broadcasting, reliable broadcasting, and secure message distribution. However, it is an open problem whether there are n ISTs rooted at any node in any n connected network with n 5. In this paper, we consider the construction of ISTs in a family of hypercube variants, called conditional BC networks. A recursive algorithm based on two common rules is proposed to construct n ISTs rooted at any node in any n -dimensional conditional BC network n X . We also show that our constructive method is adaptive to not only the existing hypercube variants, but also some other ones.

关键词： Conditional BC network, independent spanning trees, node-disjoint paths, recursive algorithm

来源：评论

学校读者我要写书评

暂无评论

An Efficient Image Enhancement Method Based on Image Saliency Detection for Portrait Photo

An Efficient Image Enhancement Method Based on Image Salienc...

引用

2014 International Conference on Advanced computer Science and Engineering(ACSE 2014)

作者： Si-qi WU Yao HU Wei-yu YU School of Electronic and Information Engineering South China University of Technology Provincial Key Laboratory for Computer Information Processing Technology Soochow University

The portrait region of portrait photo may be too dark due to backlight during the process of photo taking. Aiming at this problem, we propose a new image enhancement method based on histogram-based contrast method(HC) image saliency detection. Portrait area in portrait photo is the most salient region, thus we first use HC to measure saliency and then extract the portrait area from the portrait photo with the help of saliency map. We simply enhance the portrait region, keeping the background unchanged. Experiment results prove that proposed algorithm is able to compress the dynamic range of the original image and yield pleasing enhancement result.

关键词： Image enhancement Image saliency detection Portrait photo

来源：评论

学校读者我要写书评

暂无评论

Efficient personalized probabilistic retrieval of Chinese calligraphic manuscript images in mobile cloud environment

引用

ACM Transactions on Asian Language information processing 2014年第4期13卷 1-38页

作者： Zhuang, Yi Li, Qing Chiu, Dickson K. W. Wu, Zhiang Hu, Haiyang College of Computer and Information Engineering Zhejiang Gongshang University Hangzhou310018 China Department of Computer Science City University of Hong Kong Hong Kong Hong Kong Faculty of Education University of Hong Kong Hong Kong Hong Kong Jiangsu Provincial Key Laboratory of E-Business Nanjing University of Finance and Economics Nanjing China College of Computer Science Hangzhou Dianzi University Hangzhou China State Key Laboratory of Networking and Switching Technology Beijing University of Posts and Telecommunications China

Ancient language manuscripts constitute a key part of the cultural heritage of mankind. As one of the most important languages, Chinese historical calligraphy work has contributed to not only the Chinese cultural heritage but also the world civilization at large, especially for Asia. To support deeper and more convenient appreciation of Chinese calligraphy works, based on our previous work on the probabilistic retrieval of historical Chinese calligraphic character manuscripts repositories, we propose a system framework of the multi-feature-based Chinese calligraphic character images probabilistic retrieval in the mobile cloud network environment, which is called the DPRC. To ensure retrieval efficiency, we further propose four enabling techniques: (1) DRL-based probability propagation, (2) optimal data placement scheme, (3) adaptive data robust transmission algorithm, and (4) index support filtering scheme. Comprehensive experiments are conducted to testify the effectiveness and efficiency of our proposed DPRC method. © 2014 ACM.

关键词： Efficiency

来源：评论

学校读者我要写书评

暂无评论

A model based on crossed cubes for VoD services

A model based on crossed cubes for VoD services

引用

International Conference on Systems and Informatics (ICSAI)

作者： Haifei Shen Jianxi Fan Baolei Cheng Cheng-Kuan Lin School of Computer Science and Technology Soochow University Suzhou China Provincial Key Laboratory for Computer Information Processing Technology Soochow University China

ISBN: (纸本)9781479954599

Content distribution schemes for Video on Demand (VoD) systems, based on the peer-to-peer (P2P) technology, have attracted more and more attention. Recently, people mainly focus on the latency performance, security, interaction, scalability, and so on. We propose a new network topology model - Similarity Crossed Cube(S-CQ for short), which combines with the feature of crossed cube and establishes multiple independent spanning trees for data distribution in the layer. This network model processes the properties such as good self-organization, delay, etc. The simulation results also show that it can effectively reduce the playback delay and maintain a low delay. The S-CQ network model can provide good quality of VoD service and effectively improve the user experience.

关键词： Routing Delays Bandwidth Streaming media Network topology Peer-to-peer computing Servers

来源：评论

学校读者我要写书评

暂无评论

Model Checking Parallel Interval Logic on Parallel Run Structures

Model Checking Parallel Interval Logic on Parallel Run Struc...

引用

2012 Third International Conference on Theoretical and Mathematical Foundations of computer Science(ICTMF 2012)

作者： Zining Cao Department of Computer Science and Technology Nanjing University of Aero.& Astro. Provincial Key Laboratory for Computer Information Processing Technology Soochow University

In this paper,we firstly present a kind of models for truly concurrent systems named parallel run *** we present a parallel interval logic,called PIL,which is an extension of interval ***,we give an algorithm of model... 详细信息

关键词： Model Checking,Truly Concurrent Systems,Temporal Logic

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：