Knowledge Transfer-Based Multiple Data Sets Collaborative Analysis for Hyperspectral Band Selection

Abstract

The paper presents a knowledge transfer based collaborative analysis method for hyperspectral band selection. This collaborative analysis method establishes different band selection tasks for multiple data sets, cooperatively analyzes the shared spectral spatial structure between hyperspectral data sets, so as to improve the performance of band selection tasks for each data set. The transfer probability is adjusted dynamically to realize spectral knowledge transfer effectively and improve the cooperation ability of collaborative analysis. Experiments indicate that the proposed collaborative analysis method works more efficiently than the comparison methods.