CLUSTERING PROCESSOR, CLUSTERING PROCESSING METHOD AND PROGRAM

PROBLEM TO BE SOLVED: To reduce memory consumption in a system for reducing a distance calculation amount of k-Means on the basis of a triangle inequality, and to quicken inter-vector distance calculation.SOLUTION: A clustering processor includes: data compression means for compressing data relating...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: ONIZUKA MAKOTO, FUKUMOTO KEISHI
Format: Patent
Sprache:eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:PROBLEM TO BE SOLVED: To reduce memory consumption in a system for reducing a distance calculation amount of k-Means on the basis of a triangle inequality, and to quicken inter-vector distance calculation.SOLUTION: A clustering processor includes: data compression means for compressing data relating to a distance, and for storing the compressed distance data in a memory; data-for-distance calculation cache means for calculating a non-zero index list and a Bloom filter about an input vector and a centroid vector as the object of distance calculation, and for storing the non-zero index list and Bloom filter in a memory; and vector distance calculation means for performing the omission determination of distance calculation by using the distance data compressed by the data compression means, for performing distance calculation when the distance calculation is not omitted by using the non-zero index list and the Bloom filter, and for calculating a most neighboring centroid vector with respect to the input vector.