Awesome

rio-paper

Code and supporting materials for the ICLR 2020 RIO paper

This repository contains all the source codes to reproduce the experimental results reported in paper "Quantifying Point-Prediction Uncertainty in Neural Networks via Residual Estimation with an I/O Kernel", which is published in ICLR 2020. (Arxiv Link: https://arxiv.org/abs/1906.00588)

Before running the codes, three directories need to be created under the current path:
./Datasets/ - contains the original datasets downloaded from UCI Machine Learning Repository (https://archive.ics.uci.edu/ml/datasets.php?format=&task=reg&att=&area=&numAtt=&numIns=&type=&sort=nameUp&view=table)
./Plots/ - for storing generated figures
./Results/ - for storing experimental results

Usages of each python file:

main_experiments_RIO_variants.py - main file to run tests for all the RIO variants on all the datasets
util.py - contains functions to read data and run RIO variants
Results_Table1.py - file for post-processing all the experimental results for Table 1 (in the main paper)
Results_Figure_RMSE.py - file for plotting all the figures in Figure 3
Results_Figure_CI.py - file for plotting all the figures in Figure 4 and Figure 5
Results_Spearman_correlation.py - file for calculating Spearman's rank correlation between RMSE and noise variance

Citation

If you use RIO in your research, please cite it using the following BibTeX entry

@inproceedings{qiu:iclr20,
title={Quantifying Point-Prediction Uncertainty in Neural Networks via Residual Estimation with an I/O Kernel},
author={Xin Qiu and Elliot Meyerson and Risto Miikkulainen},
booktitle={International Conference on Learning Representations},
year={2020}
}