Learning an optimal distance metric in a linguistic vector space

Authors:
Daichi Mochihashi;Genichiro Kikui;Kenji Kita
Affiliations:
ATR Spoken Language Communication Research Laboratories, Kyoto, 619-0288 Japan and Graduate School of Information Science, NAIST, Nara, 630-0192 Japan;ATR Spoken Language Communication Research Laboratories, Kyoto, 619-0288 Japan;ATR Spoken Language Communication Research Laboratories, Kyoto, 619-0288 Japan and Center for Advanced Information Technology, Tokushima University, Tokushima, 770-8506 Japan
Venue:
Systems and Computers in Japan
Year:
2006

Citing 0
Cited 3

Matrix Learning for Topographic Neural Maps

ICANN '08 Proceedings of the 18th international conference on Artificial Neural Networks, Part I
2010 Special Issue: Local matrix learning in clustering and applications for manifold visualization

Neural Networks
Local matrix adaptation in topographic neural maps

Neurocomputing

Quantified Score

Hi-index	0.00

Visualization

Abstract

Much natural language processing still depends on the Euclidean distance function between the two feature vectors, but the Euclidean distance suffers from severe defects as to feature weightings and feature correlations. In this paper we propose an optimal metric distance function that can be used as an alternative to the Euclidean distance, accommodating the two problems at the same time. This metric is optimal in the sense of global quadratic minimization, and can be obtained from the clusters in the training data in a supervised fashion.We have confirmed the effect of the proposed metric by the sentence retrieval, document retrieval, and K-means clustering of general vectorial data. © 2006 Wiley Periodicals, Inc. Syst Comp Jpn, 37(9): 12–21, 2006; Published online in Wiley InterScience (). DOI 10.1002/scj.20533