fw-rde

Official implementation of the paper "Interpretable Neural Networks with Frank-Wolfe: Sparse Relevance Maps and Relevance Orderings" by J. Macdonald, M. Besançon, and S. Pokutta (2021).

https://github.com/zib-iol/fw-rde

Science Score: 67.0%

This score indicates how likely this project is to be science-related based on various indicators:

✓
CITATION.cff file
Found CITATION.cff file
✓
codemeta.json file
Found codemeta.json file
✓
.zenodo.json file
Found .zenodo.json file
✓
DOI references
Found 3 DOI reference(s) in README
✓
Academic publication links
Links to: arxiv.org, zenodo.org
○
Academic email domains
○
Institutional organization owner
○
JOSS paper metadata
○
Scientific vocabulary similarity
Low similarity (9.2%) to scientific vocabulary

Last synced: 10 months ago · JSON representation ·

Repository

Official implementation of the paper "Interpretable Neural Networks with Frank-Wolfe: Sparse Relevance Maps and Relevance Orderings" by J. Macdonald, M. Besançon, and S. Pokutta (2021).

Basic Info

Host: GitHub
Owner: ZIB-IOL
License: mit
Language: Python
Default Branch: main
Size: 183 KB

Statistics

Stars: 4
Watchers: 2
Forks: 0
Open Issues: 0
Releases: 0

Created over 4 years ago · Last pushed over 2 years ago

Metadata Files

Readme License Citation

Interpretable Neural Networks with Frank-Wolfe: Sparse Relevance Maps and Relevance Orderings

This repository provides the official implementation of the paper Interpretable Neural Networks with Frank-Wolfe: Sparse Relevance Maps and Relevance Orderings by J. Macdonald, M. Besançon and S. Pokutta (2021).

TL;DR: We use a constrained optimization formulation of the Rate-Distortion Explanations (RDE) (Macdonald et al., 2019) method for relevance attribution and Frank-Wolfe algorithms for obtaining interpretable neural network predictions.

Content

This repository contains subfolders with code for two independent experimental scenarios.

mnist : Sparse relevance maps (relevance attribution) and relevance orderings for a relatively small LeNet-inspired neural network classifier on the MNIST dataset of greyscale images of handwritten digits.
stl10 : Sparse relevance maps (relevance attribution) for a larger VGG-16 based neural network classifier on the STL-10 dataset of color images.

Requirements & Setup

The package versions we used are specified in Project.toml, Manifest.toml, and setup.jl.
To reproduce our computational environment run:

console julia setup.jl

To test the installation run: console test_installation.jl

This should print all the installed Julia and Python packages.

Usage

The script rde.jl can be used to obtain sparse relevance mappings.

The script rde_birkhoff.jl can be used to obtain relevance orderings with deterministic Frank-Wolfe algorithms.

The script rde_birkhoff_stochastic.jl can be used to obtain relevance orderings with stochastic Frank-Wolfe algorithms.

License

This repository is MIT licensed, as found in the LICENSE file.

Owner

Name: IOL Lab
Login: ZIB-IOL
Kind: organization
Location: Germany

Website: https://iol.zib.de
Repositories: 27
Profile: https://github.com/ZIB-IOL

Working on optimization and learning at the intersection of mathematics and computer science

Citation (CITATION.bib)

@InProceedings{pmlr-v162-macdonald22a,
  title = 	 {Interpretable Neural Networks with Frank-Wolfe: Sparse Relevance Maps and Relevance Orderings},
  author =       {Macdonald, Jan and Besan{\c{c}}on, Mathieu E. and Pokutta, Sebastian},
  booktitle = 	 {Proceedings of the 39th International Conference on Machine Learning},
  pages = 	 {14699--14716},
  year = 	 {2022},
  editor = 	 {Chaudhuri, Kamalika and Jegelka, Stefanie and Song, Le and Szepesvari, Csaba and Niu, Gang and Sabato, Sivan},
  volume = 	 {162},
  series = 	 {Proceedings of Machine Learning Research},
  month = 	 {17--23 Jul},
  publisher =    {PMLR},
  pdf = 	 {https://proceedings.mlr.press/v162/macdonald22a/macdonald22a.pdf},
  url = 	 {https://proceedings.mlr.press/v162/macdonald22a.html},
  abstract = 	 {We study the effects of constrained optimization formulations and Frank-Wolfe algorithms for obtaining interpretable neural network predictions. Reformulating the Rate-Distortion Explanations (RDE) method for relevance attribution as a constrained optimization problem provides precise control over the sparsity of relevance maps. This enables a novel multi-rate as well as a relevance-ordering variant of RDE that both empirically outperform standard RDE and other baseline methods in a well-established comparison test. We showcase several deterministic and stochastic variants of the Frank-Wolfe algorithm and their effectiveness for RDE.}
}

GitHub Events

Total

Last Year

Issues and Pull Requests

Last synced: about 1 year ago

All Time

Total issues: 2
Total pull requests: 0
Average time to close issues: 30 days
Average time to close pull requests: N/A
Total issue authors: 1
Total pull request authors: 0
Average comments per issue: 1.0
Average comments per pull request: 0
Merged pull requests: 0
Bot issues: 0
Bot pull requests: 0

Past Year

Issues: 0
Pull requests: 0
Average time to close issues: N/A
Average time to close pull requests: N/A
Issue authors: 0
Pull request authors: 0
Average comments per issue: 0
Average comments per pull request: 0
Merged pull requests: 0
Bot issues: 0
Bot pull requests: 0

ecosyste.ms

Data

Tools

Indexes

Applications

Experiments

Open Source Science