基于RJMCMC的多维尺度分析维数选择研究.pdfVIP

Proceedings of the 26th Chinese Control Conference July 26-31, 2007, Zhangji aj ie, Hunan, China 基于RJMCMC 的多维尺度分析维数选择* 卿湘运，王行愚 东理工大学信息科学与工程学院, 上海 200237 E-mail: xywang@ecust.edu.cn 要：多维尺度分析是模式识别与数据挖掘领域一个有力的降维工具。在贝叶斯多维尺度分析的基础上，我们将多 维尺度的本质低维维数决定问题视作模型选择问题。在贝叶斯框架下提出一种可逆转跳马尔可夫链蒙特卡罗(RJMCMC) 的算法，能在形成多维尺度分析的低维主坐标的同时决定本质维数。在仿真数据和真实数据上的试验结 验证了本算 法的有效性。 关键词：多维尺度分析,可逆转跳马尔可夫链蒙特卡罗,本质维,多变量贝叶斯统计 Choice of Dimension Using Reversible Jump Markov Chain * Monte Carlo in the Multidimensional Scaling Qing Xiangyun, Wang Xingyu College of Information Science and Technology, East China University of Science and Technology, Shanghai 200237, P. R. China E-mail: xywang@ecust.edu.cn Abstract: Multidimensional scaling is a powerful tool for dimensionality reduction in the field of pattern recognition and data mining. Based on the bayesian multidimensional scaling (MDS), we consider the problem of determining the number of in- trinsic low dimensions of MDS as a model selection problem. A Reversible Jump Markov chain Monte Carlo (RJMCMC) al- gorithm is proposed for performing low-dimensional coordinate and choice of dimension simultaneously within the bayesian framework. Experiments results on simulated data and real data are presented to demonstrate the effectiveness of our RJMCMC method. Key Words: Multidimensional Scaling, Reversible Jump Markov Chain Monte Carlo, Intrinsic Dimension, Multivariate Bayesian Statistics 1 究者很难凭经验或先验知识决定其本质维数。 1 引言(Introduction) 在文献[6] 中Man-Suk Oh 等提出一种贝叶斯多维 多维尺度