Data Dimensionality Reduction And Classification Based On Gaussian Process Latent Variable Model

A lot of information can be expressed by high dimensional data in the nature.The structure of these data are complex,and the size of these data is large.so,it is difficult to deal with them directly.How to reduce the dimension of these data is an important content in the field of machine learning,pattern recognition.The improvement and optimization of its technology are of great significance to the analysis of high dimensional data.The Gaussian process latent variable model is an unsupervised probability model for dimensionality reduction.It is a method based on the probability principal component analysis.It do not have the ability to make use of label information,so,it don't suit for classification.Meanwhile,it assumes that each sample is independent of each other,so it don't model the dynamic connection between samples for time series data,so,it don't suit to the modeling of time series data.In view of the above problems,this paper extends the model.we suppose the distribution of the latent variable,and make the model use the label information,and extend it to a supervised dimensionality reduction model with classification learning ability.The experimental results show that the model proposed in this paper has better classification ability than the other models.At the same time,we model the dynamic connection between latent variables with the Markov process,and we also introduce the strategy of multi-kernel learning to improve the generalization ability of the model.The experimental results show that the model proposed in this paper not only has good retention for dynamic connection of human motion,but also has good generalization ability for different types of human motion.In this paper,we model the distribution of latent variables,and we also introduce the strategy of multi-kernel learning,then,we proposed a extend model of Gaussian process latant variable model,and The extended model improves the ability of data dimensionality reduction and classification...