Mixture of Bilateral-Projection Two-dimensional Probabilistic Principal Component Analysis

TL;DR

This paper introduces a mixture of bilateral-projection 2D probabilistic PCA model that effectively captures complex data structures, improving reconstruction and recognition performance over traditional PCA methods.

Contribution

It proposes a novel mixture model with Bayesian inference for 2D data, enhancing modeling flexibility and accuracy compared to existing PCA-based algorithms.

Findings

Improved reconstruction errors

Higher recognition rates

Effective modeling of complex data structures

Abstract

The probabilistic principal component analysis (PPCA) is built upon a global linear mapping, with which it is insufficient to model complex data variation. This paper proposes a mixture of bilateral-projection probabilistic principal component analysis model (mixB2DPPCA) on 2D data. With multi-components in the mixture, this model can be seen as a soft cluster algorithm and has capability of modeling data with complex structures. A Bayesian inference scheme has been proposed based on the variational EM (Expectation-Maximization) approach for learning model parameters. Experiments on some publicly available databases show that the performance of mixB2DPPCA has been largely improved, resulting in more accurate reconstruction errors and recognition rates than the existing PCA-based algorithms.