Abstract
The paper introduces a new framework for feature learning in classification motivated by information theory. We first systematically study the information structure and present a novel perspective revealing the two key factors in information utilization: class-relevance and redun- dancy. We derive a new information decomposition model where a novel concept called class-relevant redundancy is introduced. Subsequently a new algorithm called Conditional Informative Feature Extraction is for- mulated, which maximizes the joint class-relevant information by explic- itly reducing the class-relevant redundancies among features. To address the computational difficulties in information-based optimization, we in- corporate Parzen window estimation into the discrete approximation of the ob jective function and propose a Local Active Region method which substantially increases the optimization efficiency. To effiectively utilize the extracted feature set, we propose a Bayesian MAP formulation for feature fusion, which unifies Laplacian Sparse Prior and Multivariate Logistic Regression to learn a fusion rule with good generalization ca- pability. Realizing the ineficiency caused by separate treatment of the extraction stage and the fusion stage, we further develop an improved design of the framework to coordinate the two stages by introducing a feedback from the fusion stage to the extraction stage, which significantly enhances the learning efficiency. The results of the comparative experiments show remarkable improvements achieved by our framework.