Smooth Sparse Coding via Marginal Regression for Learning Sparse Representations

摘要: We propose and analyze a novel framework for learning sparse representations, based on two statistical techniques: kernel smoothing marginal regression. The proposed approach provides flexible incorporating feature similarity or temporal information present in data sets, via non-parametric smoothing. provide generalization bounds dictionary using smooth coding show how the sample complexity depends L1 norm of function used. Furthermore, we regression obtaining codes, which significantly improves speed allows one to scale large sizes easily. demonstrate advantages approach, both terms accuracy by extensive experimentation several real sets. In addition, can be used improving semi-supervised coding.