Abstract:This paper presents the investigation of convolutional neural network (CNN) prediction successfully recognizing the temperature of the non-equilibrium phases and phase transitions in two-dimensional (2D) Ising spins on square-lattice. The model uses image snapshots of ferromagnetic 2D spin configurations as an input shape to provide the average out put predictions. By considering supervised machine learning techniques, we perform the (modified) Metropolis Monte Carlo (MC) simulations to generate the equilibrium (and non-equilibrium) configurations. In equilibrium Ising model, the Metropolis algorithm respects detailed balance condition (DBC), while its modified non-equilibrium version violates the DBC. Violating the DBC of the algorithm is characterized by a parameter $-8 < \varepsilon < 8$. We find the exact result of the transition temperature in terms of $\varepsilon$. This solution is used to encode the two (high-and low-temperature) phases through an order parameter of the model. If we set $\varepsilon = 0$, the usual single spin flip algorithm can be restored and the equilibrium configurations (training dataset) generated with such set up are used to train our model. For $\varepsilon \neq 0$, the system attains the non-equilibrium steady states (NESS), and the modified algorithm generates NESS configurations (test dataset), not defined by Boltzmann distribution. Finally, the trained model has been validated and successfully tested on the test dataset. Our result shows that CNN can correctly determine the nonequilibrium phase transition temperature $T_c$ for various $\varepsilon$ values, consistent with the exact result (our study) and also in agreement with MC result (literature).

Deep neural networks for direct, featureless learning through observation: the case of 2d spin models

Learning Spin Liquids on a Honeycomb Lattice with Artificial Neural Networks.

Deep Learning the Ising Model Near Criticality

Neural Networks as Spin Models: From Glass to Hidden Order Through Training

Nearest-Neighbours Neural Network architecture for efficient sampling of statistical physics models

Short-sighted deep learning

An Efficient Learning Algorithm for Direct Training Deep Spiking Neural Networks

Deep CovDenseSNN: A Hierarchical Event-Driven Dynamic Framework with Spiking Neurons in Noisy Environment

Sample generation for the spin-fermion model using neural networks

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

Equivariant Neural Networks for Spin Dynamics Simulations of Itinerant Magnets

On the generalizability of artificial neural networks in spin models

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

Complex Spin Hamiltonian Represented by an Artificial Neural Network

Simulating the Ising Model with a Deep Convolutional Generative Adversarial Network

Message passing variational autoregressive network for solving intractable Ising models

Deep learning nonlocal and scalable energy functionals for quantum Ising models

Sparse Autoregressive Neural Networks for Classical Spin Systems

Deep Learning Illuminates Spin and Lattice Interaction in Magnetic Materials

Forecasting the Outcome of Spintronic Experiments with Neural Ordinary Differential Equations