Improving k-Means Clustering Performance with Disentangled Internal Representations