maximevandegar / neb

Code for reproducing the experiments in the paper Neural Empirical Bayes: Source Distribution Estimation and its Applications to Simulation-Based Inference.

Getting Started

$ pip install -r requirements.txt

The implementation provides a simple interface that allows inference in a single line of code.

from inference.neb import McBiasedEstimator, McUnbiasedEstimator, ElboEstimator, IwEstimator
# Define your data, likelihood function, source model, ...
estimator = McBiasedEstimator() 
estimator.infer(observations, source_model, optimizer, log_likelihood_fct)

The source code for the different estimators was written to be self-contained in a single file for a quick and easy understanding.

Getting Started

Cite

If you make use of this code in your work, please cite our paper:

@misc{vandegar2020neural,
      title={Neural Empirical Bayes: Source Distribution Estimation and its Applications to Simulation-Based Inference}, 
      author={Maxime Vandegar and Michael Kagan and Antoine Wehenkel and Gilles Louppe},
      year={2020},
}