LEARNING VECTOR QUANTIZATION LEARNING SYSTEM, PARALLEL LEARNING METHOD, AND PROGRAM FOR THE SAME

摘要

PROBLEM TO BE SOLVED: To provide a learning vector quantization learning system, a parallel learning method, and a program for the same, reducing the ratio of a synchronous processing time to the overall processing time for learning without requiring continuous correction to a reference vector.;SOLUTION: The learning vector quantization learning system includes: a learning vector storage part 41 receiving learning vectors in parallel; a reference vector storage part 42 storing a plurality of reference vectors; a parallel data processing part 47 calculating distances between each of the learning vectors and all the reference vectors in parallel; a reference vector selection part 45 sequentially selecting close reference vectors from the reference vectors for each learning vector based on the distance calculation; a reference vector update part 46 updating the reference vector selected by the reference vector selection part 45 of the reference vectors stored in the reference vector storage part 42; and a distance data correction part 48 correcting a distance calculation result by the parallel data processing part 47 based on the updated reference vector.;COPYRIGHT: (C)2010,JPO&INPIT