A Variable Kernel Classifier using ALOPEX Optimization and its Application to Bearing Fault Diagnosis

K. Kith, B.J. van Wyk, and M.A. van Wyk (South Africa)


Metric learning, variable kernel similarity metric, general ized variable kernel similarity metric, ALOPEX, bearing data classification.


A gradient free stochastic optimization ALOPEX-based Generalized Variable kernel Similarity Metric (GVSM) learning algorithm is presented. In this paper we will out line the application of this data pre-processing technique to boost the classification of a nearest neighbours classifier on a bearing data classification task using only simple statis tical features such as the variance and central moments. It is shown that by preprocessing the bearing data using the GVSM technique the performance of a nearest neighbour classifier can be boosted to be comparable to that of a more complex neural network.

Important Links:

Go Back