Unsupervised learning of topological phase transitions using Calinski-Harabaz index

TL;DR

This paper introduces a new unsupervised learning approach using the Calinski-Harabaz index to identify topological phase transitions in statistical models, demonstrating accuracy comparable to supervised methods and applicability to experimental data.

Contribution

The authors propose using the Calinski-Harabaz index for detecting phase transitions, providing a robust, unsupervised method applicable to both topological and non-topological phases.

Findings

The $ch$ index peaks align with known critical points in the Ising model.

In the XY model, the $ch$ index peaks converge over parameter ranges.

The method performs comparably to supervised learning in phase detection.

Abstract

Machine learning methods have been recently applied to learning phases of matter and transitions between them. Of particular interest is the topological phase transition, such as in the XY model, which can be difficult for unsupervised learning such as the principal component analysis. Recently, authors of [Nature Physics \textbf{15},790 (2019)] employed the diffusion-map method for identifying topological order and were able to determine the BKT phase transition of the XY model, specifically via the intersection of the average cluster distance $\bar{D}$ and the within cluster dispersion $\bar\sigma$ (when the different clusters vary from separation to mixing together). However, sometimes it is not easy to find the intersection if $\bar{D}$ or $\bar{\sigma}$ does not change too much due to topological constraint. In this paper, we propose to use the Calinski-Harabaz ($ch$) index, defined roughly as the ratio $\bar D/\bar \sigma$, to determine the critical points, at which the $ch$ index reaches a maximum or minimum value, or jump sharply. We examine the $ch$ index in several statistical models, including ones that contain a BKT phase transition. For the Ising model, the peaks of the quantity $ch$ or its components are consistent with the position of the specific heat maximum. For the XY model both on the square lattices and honeycomb lattices, our results of the $ch$ index show the convergence of the peaks over a range of the parameters $\varepsilon/\varepsilon_0$ in the Gaussian kernel. We also examine the generalized XY model with $q=2$ and $q=8$ and at the value away from the pure XY limit. Our method is thus useful to both topological and non-topological phase transitions and can achieve accuracy as good as supervised learning methods previously used in these models, and may be used for searching phases from experimental data.