PAC-Bayes with Backprop (PBB)

This repository includes the code associated to the paper:

"Tighter risk certificates for neural networks", M. Perez-Ortiz, O. Rivasplata, J. Shawe-Taylor and C. Szepesvari, 2020.

There are a few examples of how to run the code in the running_example.py script.

Additional functionalities will be added in subsequent versions.

To create a conda environment to run these experiments in, simply run:

$ conda create --name <envname> --file requirements.txt

If you have any questions don't hesitate to contact me!