Evaluating Threshold for Retraining Rule in Semi-supervised Learning Using Multi-expert System

The creation of training set, for pattern recognition, is a difficult, expensive and time consuming task because it requires the efforts of experienced human annotators. On the other hand, unlabeled data can be obtained cheaply, but there are few ways to use them. Semi-Supervised learning uses both labeled and unlabeled data for classification task. In this paper we propose three methods to apply semi-supervised learning to re-train individual classifiers in a multi-expert scenario. More specifically, these methods are focused on a feedback-based process and on the acceptance threshold that defines what data must be selected for feedback. For the experimental results, carried out on the CEDAR (handwritten digits) database, a SVM classifier and two different combination techniques at measurement level have been used. The results show the effectiveness of the proposed approach with respect to Self-Training and Co-Training algorithms.