Use of graph kernels in Estimation of Distribution Algorithms

The graph-related problems, which solutions are represented by graphs, have attracted much attention since there are a large number of application areas in bioinformatics and social science. In this paper, we propose a novel Estimation of Distribution Algorithm which can effectively cope with graphs. The proposed method employs graph kernels in estimation and sampling phases in the EDAs. The preliminary experiments on edge-max problems and edge-min problems elucidate the effectiveness of the proposed method.