Learning Global Pairwise Interactions with Bayesian Neural Networks

This repository contains the appendix for the paper 'Learning Global Pairwise Interactions with Bayesian Neural Networks'.

---

Toy example on the energy efficiency dataset:

Please find a demostration of GEH on the energy efficiency dataset in "/demo/GEH-Demo.ipynb". We provide a trained concrete dropout BNN on that dataset for demonstration:

• "/demo/main_effect_CD_softReLU_Energy.pth" for the main effect model.
• "/demo/mlp_CD_softReLU_Energy.pth" for the MLP.

We also provide the training code of concrete dropout in "/demo/GEH-Demo.ipynb" (adopted from the original code) for users to apply on their own datasets.