https://github.com/gletarte/dichotomize-and-generalize
PAC-Bayesian Binary Activated Deep Neural Networks
https://github.com/gletarte/dichotomize-and-generalize
deep-learning generalization learning-theory neural-network pac-bayes
Last synced: 9 months ago
JSON representation
PAC-Bayesian Binary Activated Deep Neural Networks
- Host: GitHub
- URL: https://github.com/gletarte/dichotomize-and-generalize
- Owner: gletarte
- License: gpl-3.0
- Created: 2019-10-21T15:15:49.000Z (over 6 years ago)
- Default Branch: master
- Last Pushed: 2022-06-22T01:23:35.000Z (almost 4 years ago)
- Last Synced: 2023-08-01T17:56:03.700Z (almost 3 years ago)
- Topics: deep-learning, generalization, learning-theory, neural-network, pac-bayes
- Language: Python
- Size: 19.6 MB
- Stars: 5
- Watchers: 6
- Forks: 3
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Dichotomize and Generalize: PAC-Bayesian Binary Activated Deep Neural Networks
This repository contains an implementation of PBGNet (**P**AC-Bayesian **B**inary **G**radient **Net**work) and all related experiments presented in "[Dichotomize and Generalize: PAC-Bayesian Binary Activated Deep Neural Networks](https://papers.nips.cc/paper/8911-dichotomize-and-generalize-pac-bayesian-binary-activated-deep-neural-networks)" by Letarte, Germain, Guedj and Laviolette, accepted at *NeurIPS 2019*.
## Requirements
- Python 3.6
- Numpy 1.14.3
- Pytorch 1.2.0
- Poutyne 1.2
- Scikit-learn 0.20.3
- Pandas 0.23.0
- Click 6.7
## Launching
To reproduce the experiment presented in Section 6 of the paper, run:
```zsh
python launch.py
```
To launch a single learning experiment with custom options, use ``experiment.py``.
Here is an example:
```zsh
python experiment.py -d mnist17 -n pbgnet --experiment-name my_exp --sample-size 50 --hidden-size 25
```
For all possible options and their description, see ``python experiment.py --help``.
## BiBTeX
```
@inproceedings{letarte2019dichotomize,
title={Dichotomize and generalize: Pac-bayesian binary activated deep neural networks},
author={Letarte, Ga{\"e}l and Germain, Pascal and Guedj, Benjamin and Laviolette, Fran{\c{c}}ois},
booktitle={Advances in Neural Information Processing Systems},
pages={6869--6879},
year={2019}
}
```