Code for our paper:

Region-based Energy Neural Network for Approximate Inference

Environment Setup

Create a virtual environment with a python2 interpreter at 'path/to/your/evn/'

$ virtualenv -p python3.6 pyenv

Then activate your environment:

$ source pyenv/bin/activate

and install the requirement file:

$ pip install -r requirements.txt

$ python setup.py develop

experiment examples

To compare marginals for on grid graphs:

make ising_infer

To compare marginals for on complete graphs:

make -f completeMakefile ising_infer

To compare train a MRF of a grid graph:

make ising_train

To compare train MRF of a complete graph:

make -f completeMakefile ising_train