Noise signal as input data in self-organized neural networks

TL;DR

This paper explores the use of self-organizing neural networks to analyze various types of noise signals, including experimental noise from a 2D electron system, without preprocessing, highlighting its potential in solid-state physics research.

Contribution

It introduces a novel application of self-organizing neural networks to analyze uncorrelated noise signals directly, including experimental noise from a sliding Wigner-crystal-like structure.

Findings

Successfully analyzed different noise distributions (normal, triangular, uniform)

First-time analysis of noise from a sliding Wigner-crystal-like structure

Discussed potential for analyzing experimental and simulated data in solid-state physics

Abstract

Self-organizing neural networks are used to analyze uncorrelated white noises of different distribution types (normal, triangular, and uniform). The artificially generated noises are analyzed by clustering the measured time signal sequence samples without its preprocessing. Using this approach, we analyze, for the first time, the current noise produced by a sliding "Wigner-crystal"-like structure in the insulating phase of a 2D electron system in silicon. The possibilities of using the method for analyzing and comparing experimental data obtained by observing various effects in solid-state physics and simulated numerical data using theoretical models are discussed.