判决性字典学习方法综述

为了解决在基于稀疏表示的分类算法中,传统字典学习框架下学习得到的字典仅可用于信号重构而并不针对分类的问题,分析和总结了具有代表性的面向分类的字典学习算法,也称判决性字典学习。判决性字典学习算法总体上分为两类:直接使得字典具有判决性和使得表示系数具有判决性。对两类方法进行分析和总结可为本领域算法的发展提供参考,并引起更多研究。

Review on Discriminative Dictionary Learning Algorithms

Dictionaries learned from the traditional dictionary learning framework can only be used for signal reconstruction, but not for classification problems. In order to solve this problem in the classification algorithms based on sparse representation, we summarize and analyze typical dictionary learning algorithms for classification, also called discriminative dictionary learning. To be specific, we can roughly divide the discriminative dictionary learning methods into two categories: directly making the dictionary be discriminative and making the representation coefficients be discriminative. The analysis and summary in this paper will provide references for the development of algorithms in this field, and lead to further researches.