Abstract: In order to use low-level visual features of Chinese paintings to predict the author, we present a supervised heterogeneous sparse feature selection method for Chinese paintings classification.Firstly, a variety of low-level heterogeneous visual features are used to describe the painting style of Chinese paintings and mapped to high-level semantic information.Then a subset of features are selected to best represent the author's unique style from these heterogeneous features, so as to achieve the correspondence and transformation between the different painting styles of Chinese paintings and the underlying sparse features.Finally, the subset of features are used to forecast Chinese painting author and the classification task.Experimental results show that our method has good classification performance on traditional Chinese paintings.