Weighted Multi-view Clustering for Handwritten Numerals

Abstract

Many problems in educational data mining involve datasets that come from multiple different views or sources, which make the data mining task more challenging. However, most existing methods rely equally on every view, something lead to performance degradation in the case of incompatible views. In this work, we focus on a typical multi-view problem, the handwritten numerals clustering. In the proposed algorithm, each view is assigned a weight to express its importance and a simple yet efficient dynamical weight updating strategy is given.