Image classification by iterative semi-supervised sparse coding

Abstract

In this paper, we propose a novel algorithm, Iterative Semisupervised Sparse Coding(ISSC), which combines sparse coding and graphbased semi-supervised learning into a unified framework, in order to learn the discriminative sparse codes as well as an effective classification function. The ISSC algorithm fully exploits initial labels and the subsequently predicted labels for sparse codes learning. At the same time, during the graph-based semi-supervised learning stage, the similarity matrix is firstly adjusted through the latest learned sparse codes, and then is utilized to obtain a better classification function. In particular, by solving quadratic optimization, the ISSC approach can give rise to closed-form solution for learned sparse codes. We have extensively evaluated the proposed ISSC method over the widely used datasets for image classification task. The experimental results in terms of classification accuracy demonstrate that the proposed ISSC method is robust and can also achieve significant performance improvements with respect to the state-of-the-arts. © Springer International Publishing Switzerland 2013.