Entropy-based Common Process Extraction for Multi-datasets

In this paper, we focus on surveying recent research results about the common process extraction methods in the field of machine learning and data mining. Not only the similarity measure between latent processes but also some classical higher-order statistic information based common process extraction algorithms are summarized. In addition, we propose a novel common process extraction method from multi-datasets. The new method exploits the entropy approximation as the similarity measure. Furthermore, an alternate fixed point based algorithm is proposed to search the optimal solution. Compared with the conventional approaches, the proposed method can handle more various distributions. The simulation results demonstrate that the proposed method outperforms the state-of the-art methods in the problems of common process extraction and joint blind source separation.