Being the clustering distance study device which studies the parameter of the Mahalanobis distance which the occasion where

摘要

PPROBLEM TO BE SOLVED: To provide a clustering device estimating a parameter of a Mahalanobis distance used in semi-supervised clustering, from a learning data, without performing repeated computation. PSOLUTION: The clustering device 1 includes a storage means 21 for storing a value of the cluster center in a feature amount space of a data, for each cluster; a learning data input means 22 for inputting a learning data that is known to which cluster it belongs, and a parameter computing means 24 for computing a matrix as a parameter, while providing a predetermined constraint condition in the sum of orthogonal components of a product of a matrix for calculating the Mahalanobis distance and a transposed matrix of the matrix, based on the learning data input by the learning data input means 22, and based on the cluster center stored in the cluster center storage means corresponding to the cluster of the learning data. PCOPYRIGHT: (C)2010,JPO&INPIT