A semi-supervised sparse K-Means algorithm

We consider the problem of data clustering with unidentified feature quality and when a small amount labelled is provided. An unsupervised sparse method can be employed in order to detect subgroup features necessary for semi-supervised use create constraints enhance solution. In this paper we propose K-Means variant that employs these techniques. show algorithm maintains high performance other algorithms addition preserves ability identify informative from uninformative features. examine on synthetic real world sets. scenarios different types as well initialisation methods.