TAILIEUCHUNG - Mixture of hyperspheres for novelty detection

In this paper, we present a mixture of support vector data descriptions (mSVDD) for one-class classification or novelty detection. A mixture of optimal hyperspheres is automatically discovered to characterize data. | Vietnam J Comput Sci 2016 3 223-233 DOI s40595-016-0069-x CrossMark REGULAR PAPER Mixture of hyperspheres for novelty detection Duy Nguyen 1 Vinh Lai1 Khanh Nguyen1 Trung Le1 Received 30 November 2015 Accepted 6 May 2016 Published online 4 June 2016 The Author s 2016. This article is published with open access at Abstract In this paper we present a mixture of support vector data descriptions mSVDD for one-class classification or novelty detection. A mixture of optimal hyperspheres is automatically discovered to characterize data. The model includes two parts log likelihood to control the fit of data to model . empirical risk and regularization quantizer to control the generalization ability of model . general risk . Expectation maximization EM principle is employed to train our proposed mSVDD. We demonstrate the advantage of the proposed model if learning mSVDD in the input space it simulates learning a single hypersphere in the feature space and the accuracy is thus comparable but the training time is significantly shorter. Keywords Mixture of experts Mixture model Kernel method One-class classification 1 Introduction Novelty detection is an interesting research topic in many data analytics and machine learning tasks ranging from video security surveillance network abnormality detection and detection of abnormal gene expression sequence to name a few. Different from the binary classification which focuses mainly on balance dataset novelty detection aims to learn from imbalance dataset where a majority in the dataset is normal data and abnormal data or outliers constitute a minor portion of dataset. The purpose of novelty detection is to find patterns in data that do not conform to expected behaviors B Trung Le trunglm@ 1 Faculty of Information Technology HCMc University of Pedagogy Ho Chi Minh City Vietnam 5 . To obtain this aim a data description is constructed to capture all characteristics of normal data and .

• Quản trị mạng
• Mixture of experts
• Mixture model
• Kernel method
• One class classification
• Mixture of support vector data descriptions
• Modeling Wisdom
• Crowds Using Latent Mixture
• Discriminative Experts
• báo cáo khoa học
• báo cáo ngôn ngữ
• ngôn ngữ tự nhiên