Skip to content

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

License

Notifications You must be signed in to change notification settings

BorealisAI/monotonicity-mixup

Repository files navigation

Enforcing monotonicity in neural networks

This repo contains the code for the paper:

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:

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}
}

About

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

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published