Abstract:As powerful as machine learning (ML) techniques are in solving problems involving data with large dimensionality, explaining the results from the fitted parameters remains a challenging task of utmost importance, especially in physics applications. This work shows how this can be accomplished for the ferromagnetic Ising model, the main target of several ML studies in statistical physics. Here it is demonstrated that the successful unsupervised identification of the phases and order parameter by principal component analysis, a common method in those studies, detects that the magnetization per spin has its greatest variation with the temperature, the actual control parameter of the phase transition. Then, by using a neural network (NN) without hidden layers (the simplest possible) and informed by the symmetry of the Hamiltonian, an explanation is provided for the strategy used in finding the supervised learning solution for the critical temperature of the model's continuous phase transition. This allows the prediction of the minimal extension of the NN to solve the problem when the symmetry is not known, which becomes also explainable. These results pave the way to a physics-informed explainable generalized framework, enabling the extraction of physical laws and principles from the parameters of the models.

What problem does this paper attempt to address?

This paper discusses the interpretability issues of machine learning in solving physical problems, especially in the context of the Ising model in statistical physics. The Ising model is a major focus of many machine learning studies, particularly in identifying phase transitions and order parameters. The authors point out that although machine learning techniques perform well on high-dimensional data problems, interpreting the model parameter fitting results remains a significant challenge, especially in physics applications. The paper demonstrates how principal component analysis (PCA) can be used to interpret the unsupervised learning solution for the ferromagnetic Ising model. This approach successfully identified the phases and order parameters - the variations of magnetization with temperature, which are crucial parameters controlling phase transitions. Subsequently, by using a single-layer neural network (SLNN) without hidden layers and exploiting the symmetry of the Hamiltonian, the paper provides an explanation of the strategy to find the critical temperature in supervised learning. This enables the possibility of predicting the minimal extended neural network (NN) to solve the problem without knowledge of the symmetry, and is also interpretable. The paper further proposes a physics-based interpretable general framework that can extract physical laws and principles from model parameters. The authors emphasize the insight of the solution as crucial for overcoming adoption barriers of machine learning in healthcare applications and other fields, while also having profound impacts on technological advancements and scientific research. In the unsupervised learning part, PCA is shown to effectively detect the direction of the maximum variation of magnetization with temperature. In the supervised learning part, the determination of the critical temperature is achieved through a simple single-layer neural network (SLNN), revealing how the model learns the key features of phase transitions by employing symmetry. This simplified model can provide approximate solutions for different two-dimensional and three-dimensional lattices, although with lower accuracy for the cubic lattice as it belongs to a different universality class. Through these analyses, the paper not only provides a deeper understanding but also paves the way for the development of interpretable and efficient machine learning models that may unveil new physical laws and facilitate new discoveries.

Explaining the Machine Learning Solution of the Ising Model

Machine Learning of Nonequilibrium Phase Transition in an Ising Model on Square Lattice

Machine Learning-based estimation and explainable artificial intelligence-supported interpretation of the critical temperature from magnetic ab initio Heusler alloys data

Network science: Ising states of matter

Interpretable machine learning for inferring the phase boundaries in a nonequilibrium system

Machine learning force-field model for kinetic Monte Carlo simulations of itinerant Ising magnets

Influence of anisotropy on the study of critical behavior of spin models by machine learning methods

Network science Ising states of matter

Accurate Prediction of Magnetocaloric Effect in NiMn-based Heusler Alloys by Prioritizing Phase Transitions Through Explainable Machine Learning

Explainable Machine Learning Identification of Superconductivity from Single-Particle Spectral Functions

Learning phase transitions from regression uncertainty: A new regression-based machine learning approach for automated detection of phases of matter

Heuristic machinery for thermodynamic studies of SU(N) fermions with neural networks

Machine Learning Phases of Strongly Correlated Fermions

Self-organizing maps as a method for detecting phase transitions and phase identification

Predicting the Curie temperature of ferromagnets using machine learning

Few-shot Machine Learning in the Three-Dimensional Ising Model

Unsupervised machine learning of quenched gauge symmetries: A proof-of-concept demonstration

Persistent self-organized states in non-equilibrium magnetic models

Adding machine learning within Hamiltonians: Renormalization group transformations, symmetry breaking and restoration

Machine Learning of Phase Transitions in the Percolation and XY Models

Machine Learning Magnetic Parameters from Spin Configurations