ICML 2018oral2 citations
A Boo(n) for Evaluating Architecture Performance
Ondrej Bajgar, Rudolf Kadlec, Jan Kleindienst
Abstract
We point out important problems with the common practice of using the best single model performance for comparing deep learning architectures, and we propose a method that corrects these flaws. Each time a model is trained, one gets a different result due to random factors in the training process, which include random parameter initialization and random data shuffling. Reporting the best single model performance does not appropriately address this stochasticity. We propose a normalized expected best-out-of-$n$ performance ($\text{Boo}_n$) as a way to correct these problems.
BibTeX
@InProceedings{pmlr-v80-bajgar18a,
title = {A Boo(n) for Evaluating Architecture Performance},
author = {Bajgar, Ondrej and Kadlec, Rudolf and Kleindienst, Jan},
booktitle = {Proceedings of the 35th International Conference on Machine Learning},
pages = {334--343},
year = {2018},
editor = {Dy, Jennifer and Krause, Andreas},
volume = {80},
series = {Proceedings of Machine Learning Research},
month = {10--15 Jul},
publisher = {PMLR},
pdf = {http://proceedings.mlr.press/v80/bajgar18a/bajgar18a.pdf},
url = {https://proceedings.mlr.press/v80/bajgar18a.html},
abstract = {We point out important problems with the common practice of using the best single model performance for comparing deep learning architectures, and we propose a method that corrects these flaws. Each time a model is trained, one gets a different result due to random factors in the training process, which include random parameter initialization and random data shuffling. Reporting the best single model performance does not appropriately address this stochasticity. We propose a normalized expected best-out-of-$n$ performance ($\text{Boo}_n$) as a way to correct these problems.}
}