/pi-snm-qde

Prediction Intervals: Split Normal Mixture from Quality-Driven Deep Ensembles. Published at Uncertainty in AI (UAI) 2020.

Primary LanguageJupyter NotebookMIT LicenseMIT

Prediction Intervals

Split Normal Mixture from Quality-Driven Deep Ensembles

https://arxiv.org/abs/2007.09670


This repository is accompanying the paper "Prediction Intervals: Split Normal Mixture from Quality-Driven Deep Ensembles" published at UAI 2020. The implementation contains the proposed method SNM-QD+, and it allows to reproduce the presented results.

Requirements

pip install -r requirements.txt

or

conda env create --file environment.yml

Usage

Experiments:

source exp_{qdp2|qd2|mve2|mse2}.sh

Hyper-parameter search (HPS):

source hps_{qdp2|qd2|mve2|mse2}.sh

To learn how to execute/reproduce an experiment or HPS on a single dataset, please have a look inside the shell scripts (*.sh). All output artifacts are stored in an automatically created runs directory.

Feedback

For questions, suggestions, feedback or bug reports, please open an issue or contact us via email (to be found in the paper).

Citation

@InProceedings{pi-snm-qde:2020,
  title = {Prediction Intervals: Split Normal Mixture from Quality-Driven Deep Ensembles},
  author = {S. Salem, T{\'a}rik and Langseth, Helge and Ramampiaro, Heri},
  booktitle = {Proceedings of the 36th Conference on Uncertainty in Artificial Intelligence (UAI)},
  year = {2020},
  volume = {124},
  pages = {1179--1187},
  editor = {Jonas Peters and David Sontag},
  series = {Proceedings of Machine Learning Research},
  publisher = {PMLR},
  url = {http://proceedings.mlr.press/v124/saleh-salem20a.html}
}