Energy-Based Feature Selection and Its Ensemble Version

摘要

Variable and feature selection has been a research topic with practical significance in many areas such as statistics, pattern recognition, machine learning and data mining. The task of feature selection is to choose an effective feature subset out of a given feature set to reduce the feature space dimensionality. In this paper, along with the guidelines of Energy-based model, a unified energy-based framework for feature selection and a feature ranking algorithm under this framework is presented. On the other hand, in order to increase the stability of our algorithm, an ensemble feature selection is introduced. Some experiments are conducted on the real world and synthesis data sets to demonstrate the ability of our feature selection algorithm and the stability improvement of the ensemble feature selection.