Deep Deterministic Independent Component Analysis for Hyperspectral Unmixing

TL;DR

This paper introduces a neural network-based ICA method that minimizes dependence among components using R{é}nyi's entropy, optimized via SGD, and demonstrates its effectiveness in hyperspectral unmixing.

Contribution

The paper presents a novel neural network ICA approach utilizing R{é}nyi's entropy, enabling direct optimization without variational or adversarial methods, applied successfully to hyperspectral unmixing.

Findings

ICA plays a significant role in hyperspectral unmixing.

The proposed DDICA outperforms traditional methods in unmixing tasks.

Code and implementation details are publicly available.

Abstract

We develop a new neural network based independent component analysis (ICA) method by directly minimizing the dependence amongst all extracted components. Using the matrix-based R{\'e}nyi's $\alpha$-order entropy functional, our network can be directly optimized by stochastic gradient descent (SGD), without any variational approximation or adversarial training. As a solid application, we evaluate our ICA in the problem of hyperspectral unmixing (HU) and refute a statement that "\emph{ICA does not play a role in unmixing hyperspectral data}", which was initially suggested by \cite{nascimento2005does}. Code and additional remarks of our DDICA is available at https://github.com/hongmingli1995/DDICA.