Contrastive independent component analysis

TL;DR

This paper introduces contrastive ICA (cICA), a novel extension of independent component analysis designed to identify salient features in experimental data relative to control data, using a new tensor decomposition algorithm.

Contribution

It proposes a new contrastive ICA method with a linear algebra-based tensor decomposition, establishing its identifiability and demonstrating its effectiveness on various datasets.

Findings

cICA effectively identifies salient features in contrastive datasets

The tensor decomposition algorithm is efficient and scalable

cICA outperforms existing methods in pattern discovery and data visualization

Abstract

In recent years, there has been growing interest in jointly analyzing a foreground dataset, representing an experimental group, and a background dataset, representing a control group. The goal of such contrastive investigations is to identify salient features in the experimental group relative to the control. Independent component analysis (ICA) is a powerful tool for learning independent patterns in a dataset. We generalize it to contrastive ICA (cICA). For this purpose, we devise a new linear algebra based tensor decomposition algorithm, which is more expressive but just as efficient and identifiable as other linear algebra based algorithms. We establish the identifiability of cICA and demonstrate its performance in finding patterns and visualizing data, using synthetic, semi-synthetic, and real-world datasets, comparing the approach to existing methods.