A low-cost variational-Bayes technique for merging mixtures of probabilistic principal component analyzers

Abstract

International audienceMixtures of probabilistic principal component analyzers (MPPCA) have shown effective for modeling high-dimensional data sets living on nonlinear manifolds. Briefly stated, they conduct mixture model estimation and dimensionality reduction through a single process. This paper makes two contributions: first, we disclose a Bayesian technique for estimating such mixture models. Then, assuming several MPPCA models are available, we address the problem of aggregating them into a single MPPCA model, which should be as parsimonious as possible. We disclose in detail how this can be achieved in a cost-effective way, without sampling nor access to data, but solely requiring mixture parameters. The proposed approach is based on a novel variational-Bayes scheme operating over model parameters. Numerous experimental results and discussion are provided

Face recognition using supervised probabilistic principal component analysis mixture model in dimensionality reduction without loss framework

“…(vii) Proj‐SPPCAMM, which applies SPPCAMM and then the projection penalty idea with a non‐linear SVM with polynomial kernel. (viii) 2DPCA, which uses exactly the method proposed in [9]. (ix) 2DPCA‐fusion, which applies feature fusion approach for 2DPCA, exactly as proposed in [10].…”

“…In [9], two new techniques for estimation of mixtures of probabilistic principal component analysers, and aggregation of these models is proposed. A fully probabilistic and Bayesian framework, along with the possibility to deal with high‐dimensional data motivated this approach.…”

Method of speed data fusion based on Bayesian combination algorithm and high‐order multi‐variable Markov model

“…In the next section, we will represent how to use Bayesian methods for data fusion. In many studies [13, 18], Bayesian methods are classified as effective methods for the fusion of multiple data sources; such methods are based on the principle of Bayes' theorem: posterior is proportional to prior times likelihood.…”