Source-linked AI summary
Discovering Phase Transitions with Unsupervised Learning
Lei Wang
TL;DR
Identifying phase transitions becomes difficult when useful order parameters are unknown or nonlocal, and supervised approaches require previously solved training data. This paper applies PCA and clustering directly to raw many-body configurations, finding phase indicators such as magnetization and structure factors, while noting that linear PCA remains limited for subtler topological transitions.
Problem
Useful indicators of phase transitions can be difficult to identify for new states of matter with elusive nonlocal order parameters, while supervised methods require previously solved training data.
Method
The paper uses PCA to extract low-dimensional features from raw configurations and clustering analysis to divide samples into phases without assuming a transition or critical-point location.
Results
The approach identifies the Ising model’s order parameter and, for the constrained model, a structure factor that indicates the phase transition.
Takeaways & Limitations
Unsupervised learning can extract physically meaningful phase indicators directly from many-body configurations without prior lattice geometry or Hamiltonian knowledge.
Takeaways & Limitations
Because PCA is limited to linear transformations, identifying subtler transitions associated with topological order remains challenging.
Abstract
from arXiv · showhide
Unsupervised learning is a discipline of machine learning which aims at discovering patterns in big data sets or classifying the data into several categories without being trained explicitly. We show that unsupervised learning techniques can be readily used to identify phases and phases transitions of many body systems. Starting with raw spin configurations of a prototypical Ising model, we use principal component analysis to extract relevant low dimensional representations the original data and use clustering analysis to identify distinct phases in the feature space. This approach successfully finds out physical concepts such as order parameter and structure factor to be indicators of the phase transition. We discuss future prospects of discovering more complex phases and phase transitions using unsupervised learning techniques.