generalized-UMNN

Extension to multivariate unconstrained monotonic functions. Direct application of the underlying principle behind the Kolmogorov-Arnold theorem.

This repository is a sketch of the possibility to model functions that are monotonic with respect to more than one input variable.

For detailed information visit the UMNN git repository.

Cite

If you make use of this code in your own work, please cite our paper:

@inproceedings{wehenkel2019unconstrained,
  title={Unconstrained monotonic neural networks},
  author={Wehenkel, Antoine and Louppe, Gilles},
  booktitle={Advances in Neural Information Processing Systems},
  pages={1543--1553},
  year={2019}
}