摘要 |
<P>PROBLEM TO BE SOLVED: To perform high-speed clustering having the same clustering result as that of an original Affinity Propagation. <P>SOLUTION: A clustering device according to the present invention outputs an upper limit and a lower limit of a message of a data pair with similarity of the data pair defined as an input, outputs a nondense graph structure with the upper limit and the lower limit of the message defined as the input, determines a convergence value of an edge corresponding to the edge in the graph by repeat operation, with the graph structure defined as the input, determines the convergence value of the data pair that is not the edge, with the convergence value of the edge defined as the input, and determines an exemplar of each data point from the convergence values of all the data pair. <P>COPYRIGHT: (C)2013,JPO&INPIT |