Code to show the simulation results in ARM: Augment-REINFORCE-Merge Gradient for Stochastic Binary Networks
The MNIST data is self-contained and the Omniglot data is in the repository.
Below are the paper to cite if you find the algorithms in this repository useful in your own research:
@inproceedings{yin2018arm,
title={{ARM}: Augment-{REINFORCE}-Merge Gradient for Stochastic Binary Networks},
author={Mingzhang Yin and Mingyuan Zhou},
booktitle={International Conference on Learning Representations},
year={2019}
}
This code is offered under the MIT License.