Studies on the Constrained Group-Wise K-Means Algorithm

Abstract

In this paper, the constrained group-wise k-means algorithm is proposed to select groups of features during clustering. The idea of the algorithm is it not only constrains the weight of individual features via the lasso penalty, but also restrains group-wise weight of features by the group lasso penalty. Under the effect of these two penalties, this algorithm not only performs better in clustering, but is also comparable to other approaches in its ability to obtain the correct features with costing less computational time than the sparse k-means algorithm.