Introduction

This repo contains code for reproducing the results in the paper Query Training: Learning a Worse Model to Infer Better Marginals in Undirected Graphical Models with Hidden Variables at the 35th AAAI Conference on Artificial Intelligence (AAAI 2021).

Setting up the environment

1. Install miniconda (if not already installed)
2. Set up the virtual environment

conda env create -f environment.yml
conda activate qt
python setup.py develop

3. Download the data and point BASE in query_training/__init__.py to the data directory

The code was tested on Ubuntu 18.04 with CUDA 10.1.

Reproducing the results

Use the scripts in the scripts folder to reproduce results in the paper.