Boosted similarity learning based on discriminative graphs

Similarity measurement is crucial for unsupervised learning and semi-supervised learning. Unsupervised methods need a similarity to do clustering. Semi-supervised algorithms need a similarity to take advantage of unlabeled data. In this paper, we develop a boosted similarity learning algorithm. Based on the manifold assumption, our similarity is learned iteratively by a few discriminative graphs. So our similarity adopts the local structure information underlying the data. We propose “within graph-cluster scatter Sw” and “between graph-cluster scatter Sb”. Sw and Sb are used to analyze the discrimination of a given graph. Experimental results on both synthetic and public available data sets show that the proposed method outperforms the state-of-the-art approaches.