1D Ising Model

Code for simulating 1D Ising Model

1. Run ising_1D to define true model, parameters, and observables

2. Run diffusion_ising for the Simple MLP // training --> Defined parameters in this section

   • Uses standard noise prediction, MLE Loss
   • Uses reverse time Euler sampling

3. Run plot_correlations to visualize xi v. T, domain wall density v. T

• Outputs two visuals: correlation length xi and domain wall density based on xi and rho data from diffusion_ising script

4. plot_diffusion_v_true_1D

• loads true ising data and diffusion results to plot the domain wall density vs. temperature for both the true and generated 1D ising

2D Ising Model

1. 2D_MC_Initialization

• Initialize 2D monte carlo

2. 2D_diffusion_utils

• Define utilities needed for diffusion model

3. 2D_Model

• UNet, attention, and diffusion classes

4. 2D_Training

• Train model --> 100 Epochs --> 300 timesteps

5. 2D_sampling

• Draw samples

6. 2D_JS_Div

• JS Divergence

7. 2D_Loop

• Iterate through temps