Authors: Yanting Teng

ml_toric_code is the code base for our paper [Classifying topological neural network quantum states via diffusion maps][arXiv link to be added] The code base is implemented in JAX.

• To run the variational Monte Carlo component, see the demo notebook Optimization_demo.ipynb.
• For visualizing the dataset linked here , see DM_data.ipynb.

The experiments are run on FASRC Cannon cluster at Harvard University. It is set up with config files in exp_cluster/configs/ folder. To generate and characterize the datasets, follow these steps:

1. Optimize for intial ``seeds'' ${\Lambda^0}$ at various field valuess $h$ (specified in config_opt.py)

sbatch opt_v1.slurm.sh

2. Estimate properties of the optimized states
   e.g. for the optimized states with the tracker id 10331255 using config_opt_est_v1_10331255:

sbatch opt_est_v1.slurm.sh

3. Generate ensembles for each state in ``seeds'' e.g. to generate the ensembles the tracker id 10910143 using config_ens_v2_10910143.py:

sbatch ens_v2.slurm.sh 

4. Estimate properties of the states in ensembles e.g. use config_est_ens.py

sbatch est_ens.slurm.sh

5. Perform dimensional reduction of the dataset using diffusion_maps module. This is done in a colab notebook. Some examples of diffusion map spectra are in DM_data.ipynb.