Structural Dictionary Learning based on Non-convex Surrogate of l2,1 Norm for Classification

Recently, group sparse representation which is based on a hypothesis about correlation of coefficient variables has attracted much attention due to its effectiveness and robustness in dictionary learning. Traditional group sparse representation methods use l(2,1) norm to enforce the estimation of models with joint sparsity patterns, which often leads to over-punishment phenomenon. To solve this issue, we replace l(2,1) with non-convex surrogate of l(2,0), and give a general solver for the corresponding optimization algorithm. Experimental results confirm the effectiveness of our proposed method.