https://github.com/xmodar/network_moments
A toolkit for computing some probabilistic moments of deep neural networks.
https://github.com/xmodar/network_moments
deep-learning neural-networks probability
Last synced: about 1 year ago
JSON representation
A toolkit for computing some probabilistic moments of deep neural networks.
- Host: GitHub
- URL: https://github.com/xmodar/network_moments
- Owner: xmodar
- License: mit
- Created: 2018-06-16T06:21:16.000Z (almost 8 years ago)
- Default Branch: master
- Last Pushed: 2020-04-20T06:06:36.000Z (about 6 years ago)
- Last Synced: 2025-03-22T20:26:16.290Z (about 1 year ago)
- Topics: deep-learning, neural-networks, probability
- Language: Python
- Size: 487 KB
- Stars: 6
- Watchers: 2
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Network Moments
Network Moments is a toolkit that enables computing some probabilistic moments of deep neural networks given a specific input distribution. The current implementation allows you to compute the first and second Gaussian network moments (GNM) of affine-ReLU-affine networks i.e., the output mean and variance subject to Gaussian input.
The main backend framework is [
](./network_moments/torch/) but also [
](./network_moments/tensorflow/) and [
](./matlab/) are supported.
### Requirements
Network Moments was developed and tested with the following:
- [Python](https://www.python.org/) v3.6.3+
- **Option 1**: [`PyTorch`](./network_moments/torch/)
- [torch](https://pytorch.org/) v0.4.1+
- [torchvision](https://github.com/pytorch/vision) v0.2.1+
- **Option 2**: [`TensorFlow`](./network_moments/tensorflow/)
- [numpy](http://www.numpy.org) v1.14.2+
- [tensorflow](https://www.tensorflow.org/) v1.8.0+
- **Option 3**: [`MatLab`](./matlab/)
- [matlab](https://www.mathworks.com/products/matlab.html) vR2017b+
You need [Jupyter](https://jupyter.org/) to run [tightness](./static/tightness.ipynb). It is recommended that you have [Jupyter Lab](https://github.com/jupyterlab/jupyterlab).
### Installation
After installing the requirements, to install or update this package run the following in the terminal:
```sh
pip install -U git+https://github.com/ModarTensai/network_moments.git
```
Now go to the [tightness notebook](./static/tightness.ipynb) to see how to use this tool with the default backend framework.
To uninstall the package:
```sh
pip uninstall network_moments
```
### Usage
To import the [`PyTorch`](./network_moments/torch/) sub-package:
```python
import network_moments.torch as nm
```
The basic usage is demonstrated in the [tightness notebook](./static/tightness.ipynb).
To import the [`TensorFlow`](./network_moments/tensorflow/) sub-package:
```python
import network_moments.tensorflow as nm
```
Please, refer to [tensorflow tests notebook](./static/tensorflow_tests.ipynb) for examples to compare [`PyTorch`](./network_moments/torch/) and [`TensorFlow`](./network_moments/tensorflow/) implementations.
### Cite
This is the official implementation of the method described in [this paper](http://openaccess.thecvf.com/content_cvpr_2018/html/Bibi_Analytic_Expressions_for_CVPR_2018_paper.html) (checkout the [poster](https://drive.google.com/file/d/1S_HurI9vwhhqyzAabDiaq0ZJBPBrV_xh/view)):
```bibtex
@InProceedings{Bibi_2018_CVPR,
author = {Bibi, Adel and Alfadly, Modar and Ghanem, Bernard},
title = {Analytic Expressions for Probabilistic Moments of PL-DNN With Gaussian Input},
booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2018}
}
```
### License
MIT
### Author
[Modar M. Alfadly](https://github.com/ModarTensai/network_moments/)
### Contributors
I would gladly accept any pull request that improves any aspect of this repository.