What Is Meant by Clustering of ICA (Independent Component Analysis) Components?

Question

I'm following the EEGLAB tutorial about STUDY creation and manipulation of the datasets, and came across the concept of clustering of ICA components. I'm not too familiar with processing of ICA components other than its physical meaning which is separating a multivariate signal into additive subcomponents. If someone could explain clustering in as simple way to understand intuitively, that would be great!

Answer 1

Since ICA is a generalization of PCA (Principal Component Analysis) one could use it for intuition about the process.

In PCA we create a new coordinate system where we can represent each data sample.
So if we have 10 Dimensional data for each sample we have 10 components on the new coordinate system.
Classic clustering approach is to use less than 10 components (Process called Dimensionality Reduction). Then one could apply clustering algorithm of his choice with K-Means Clustering being the classic choice.

The same idea goes for ICA. Once you have a representation of the data in the new system of representation you're applying on it (Or part of it) a clustering algorithm.

Some hands on examples are given by:

1. Kaggle - Unsupervised Learning: PCA and Clustering.
2. SciKit Learn - Unsupervised Learning: Seeking Representations of the Data.