Fig. 1From: Improving clustering performance using independent component analysis and unsupervised feature learningPipeline for processing. Each of the components contains the options available for implementation. The simplest processing pipeline to obtain clustering results consists of a L2-normalization on the data, followed by K-means clustering. The processing stream with the most components would consist: (1) L2-normalization followed by UFL using either RICA or SFT; (2) similarity graph construction; (3) GNMF or spectral decomposition followed by ICA blind source separation; and (4) K-means clusteringBack to article page