Machine learning the 2D percolation model

TL;DR

This paper applies deep learning techniques, specifically convolutional neural networks, to analyze 2D percolation models, aiming to characterize percolation states and recognize the importance of percolating clusters.

Contribution

It demonstrates the effectiveness of deep learning in identifying key features of 2D percolation models, including densities, correlation lengths, and the role of percolating clusters.

Findings

Deep learning accurately characterizes percolation densities.

CNNs recognize the significance of percolating clusters.

The approach offers a new tool for studying percolation phenomena.

Abstract

We use deep-learning strategies to study the 2D percolation model on a square lattice. We employ standard image recognition tools with a multi-layered convolutional neural network. We test how well these strategies can characterise densities and correlation lengths of percolation states and whether the essential role of the percolating cluster is recognised.