The preprint of the paper is avilable here.
The files in this repository should be run on Google Colab. All dependencies will be automatically taken care of when running the code on Colab.
The instructions to run the simulations are given at the begining of each notebook. In summary, all you need to do is to provide the path for pretrained models and output path to save the resulting simulations. Once that is done, you are all set. Press Run button.
In case you want to train a model from scratch, please use the train notebook. Don't forget to save the trained model. You can later use the saved model to run simulations.
For two levels of noise and two threshold levels, simulation results for each model will be ready in less than 5 minutes for the whole MNIST dataset. Expected example output is provided here.
To download the pretrained models, you can go to our OSF page.