https://github.com/borealisai/monotonicity-mixup
Code of "Not Too Close and Not Too Far: Enforcing Monotonicity Requires Penalizing The Right Points"
https://github.com/borealisai/monotonicity-mixup
monotonicity neurips-2021 pytorch xai4debugging
Last synced: 30 days ago
JSON representation
Code of "Not Too Close and Not Too Far: Enforcing Monotonicity Requires Penalizing The Right Points"
- Host: GitHub
- URL: https://github.com/borealisai/monotonicity-mixup
- Owner: BorealisAI
- License: other
- Created: 2021-10-29T01:46:57.000Z (almost 4 years ago)
- Default Branch: main
- Last Pushed: 2021-12-16T23:09:30.000Z (almost 4 years ago)
- Last Synced: 2025-02-16T03:35:58.488Z (8 months ago)
- Topics: monotonicity, neurips-2021, pytorch, xai4debugging
- Language: Python
- Homepage:
- Size: 42 KB
- Stars: 1
- Watchers: 4
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Enforcing monotonicity in neural networks
This repo contains the code for the paper:
#### [Not Too Close and Not Too Far: Enforcing Monotonicity Requires Penalizing The Right Points](https://openreview.net/forum?id=xdFqKVlDHnY)
by Joao Monteiro1, Mohamed Osama Ahmed2, Hossein Hajimirsadeghi2, and Greg Mori2
1. Institut National de la Recherche Scientifique
2. Borealis AI
## Running experiments
We provide scripts to easily launch experiments once requirememnts are installed.
Examples:
```
./submit_all_reg.sh blogData cmn_MLP
```
Or, for experiments with synthetic data:
```
./synth_train_all_reg.sh cmn_MLP
```
### Data preparation
Data needs to be prepared in advance and placed under ./exp/data/
We provide scripts to prepare data and to generate the data required for synthetic experiments under ./data_utils/
Raw data for a subset of the datasets we consider can be found at:
- COMPAS: https://github.com/gnobitab/CertifiedMonotonicNetwork/blob/main/compas/compas_scores_two_years.csv
- BlogFeedback: https://archive.ics.uci.edu/ml/datasets/BlogFeedback#.
## Citation:
```
@inproceedings{
monteiro2021not,
title={Not Too Close and Not Too Far: Enforcing Monotonicity Requires Penalizing The Right Points},
author={Joao Monteiro and Mohamed Osama Ahmed and Hossein Hajimirsadeghi and Greg Mori},
booktitle={eXplainable AI approaches for debugging and diagnosis.},
year={2021},
url={https://openreview.net/forum?id=xdFqKVlDHnY}
}
```