Experimental study on the multiple use of training patterns in confidence-weighted on-line learning for fuzzy classifiers

On-line learning for fuzzy classifiers are investigated in this paper. In the on-learning problems, it is assumed that there are only a single training patterns available at a time. This paper extends the assumption by allowing multiple training patterns available for incrementally update the fuzzy classifiers. Confidence-weighted learning is used for the on-line learning of the fuzzy classifiers. Computational experiments are conducted to show the high-performance of the fuzzy classifiers with the confidence-weighted learning.