Tri-training: exploiting unlabeled data using three classifiers

TLDR

Unlabeled data are abundant in many data‑mining tasks while labeled examples are costly, making semi‑supervised methods such as co‑training attractive. This work introduces tri‑training, a new co‑training‑style semi‑supervised learning algorithm. Tri‑training creates three classifiers from the labeled set and iteratively refines them by labeling unlabeled instances only when the other two classifiers agree, without requiring multiple views or algorithmic constraints, thus extending applicability beyond prior co‑training methods. Experiments on UCI datasets and web‑page classification demonstrate that tri‑training effectively leverages unlabeled data to improve learning performance.

Abstract

In many practical data mining applications, such as Web page classification, unlabeled training examples are readily available, but labeled ones are fairly expensive to obtain. Therefore, semi-supervised learning algorithms such as co-training have attracted much attention. In this paper, a new co-training style semi-supervised learning algorithm, named tri-training, is proposed. This algorithm generates three classifiers from the original labeled example set. These classifiers are then refined using unlabeled examples in the tri-training process. In detail, in each round of tri-training, an unlabeled example is labeled for a classifier if the other two classifiers agree on the labeling, under certain conditions. Since tri-training neither requires the instance space to be described with sufficient and redundant views nor does it put any constraints on the supervised learning algorithm, its applicability is broader than that of previous co-training style algorithms. Experiments on UCI data sets and application to the Web page classification task indicate that tri-training can effectively exploit unlabeled data to enhance the learning performance.