Reviews: Group Sparse Additive Machine

This paper presents a new additive classification model (GroupSAM) with grouped variables based on group based additive models in kernel Hilbert spaces, and an upper bound for the generalization error is provided in Theorem 1 to show that GroupSAM converges to the optimal error with a polynomial decay under certain conditions. Experiments are performed on synthetic data and seven UCI datasets that have been modified in order to get some group structure in the feature space. The results provided in the paper show that the proposed method is able to take into account grouped features while yielding good performance in terms of classification accuracy. As far as I know, the paper presents a novel method for group sparse nonlinear classification with additive models. Is not clear to me as the framework presented in the above paper seems to be easy to extend to classification settings.