Handwriting Recognition with Transductive Confidence Machines

Abstract

The recently introduced transductive confidence machines (TCMs) framework allows to extend classifiers such that their performance can be set by the user prior to classification. In this paper we apply the TCM framework, with a plugged in k-nearest neighbor classifer, to the domain of (on-line) handwriting recognition. First, we modify the original TCM algorithm to make it much more efficient. Then, we use this modified algorithm to classify the NicIcon database of iconic gestures. Results show that the modified TCM algorithm is a promising way to classify handwriting.