关键词: 稀疏低秩逼近, 双边判别框架, 主成分分析

Abstract: “High dimensionality and small size samples” is widely encountered in many real world machine learning applications. Low rank approximation to parametric matrix has recently been proven to be an effective method to control the complexity of models. Most of the previous low rank methods are required to specify the target rank by hand beforehand (e.g., principal component analysis), however, imposing the sparsity constraints on the parametric matrix can avoid this. In particular, under a bilinear discriminative framework, decomposing the parametric matrix and simultaneously constraining their ranks with the sparsity-inducing regularization will perform well. The result-ing problem can be efficiently solved with coordinate descent. This method is able to take the spatial information of structured data into account, leading to improved generalization capability. The feasibility and effectiveness of the proposed method is demonstrated on a security biometric application.

Key words: sparse low rank approximation, bilinear discriminative framework, principal component analysis