Sparse model selection and parameter identification

Model selection generally considers the tradeoff between the model fidelity and its complexity, which is realized by structural risk minimization. By the regularization-based sparse component analysis method, a novel regularization function is constructed, then a new principle of model selection is presented with corresponding algorithm of coefficient solution. Finally, a noisy polynomial signal with missing term is presented, which is processed to get the coefficients by our method and the least square method respectively. Numeric results demonstrate that our method can identify the missing term and solve the coefficient accurately at the same time. However, the least square method must be used combining with different model order selection criteria and the estimated coefficients are not accurate.