BorealisAI / monotonicity-mixup

Code of "Not Too Close and Not Too Far: Enforcing Monotonicity Requires Penalizing The Right Points"

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

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}
}