PAC-Bayes-In-Medical-Imaging

Introduction

The repo for the work "PAC Bayesian Performance Guarantees for Deep(Stochastic) Networks in Medical Imaging." Accepted to MICCAI 2021. Available at: https://arxiv.org/abs/2104.05600

Preparation

Prerequisites

• Python 3.6
• Pytorch 1.4
• numpy
• tqdm
• pandas
• PIL

Dataset Preparation

• Run get_data.sh to retrieve the ISIC2018 challenge data.
• Run make_split.py to generate a train test split.
• Run python3 -m src.main **kwargs to train models and compute bounds.

Training

To reproduce all results run the following scripts:

• sh scripts/run-lw.sh
• sh scripts/run-rn.sh
• sh scripts/run-un.sh

To run individual experiments, please check the comments which identify each subcall with the correct experiment.