Out of Distribution Data Detection Using Dropout Bayesian Neural Networks
Abstract
We explore the utility of information contained within a dropout based Bayesian neural network (BNN) for the task of detecting out of distribution (OOD) data. We first show how previous attempts to leverage the randomized embeddings induced by the intermediate layers of a dropout BNN can fail due to the distance metric used. We introduce an alternative approach to measuring embedding uncertainty, and demonstrate how incorporating embedding uncertainty improves OOD data identification across three tasks: image classification, language classification, and malware detection.
Cite
Text
Nguyen et al. "Out of Distribution Data Detection Using Dropout Bayesian Neural Networks." AAAI Conference on Artificial Intelligence, 2022. doi:10.1609/AAAI.V36I7.20757Markdown
[Nguyen et al. "Out of Distribution Data Detection Using Dropout Bayesian Neural Networks." AAAI Conference on Artificial Intelligence, 2022.](https://mlanthology.org/aaai/2022/nguyen2022aaai-out/) doi:10.1609/AAAI.V36I7.20757BibTeX
@inproceedings{nguyen2022aaai-out,
title = {{Out of Distribution Data Detection Using Dropout Bayesian Neural Networks}},
author = {Nguyen, André T. and Lu, Fred and Munoz, Gary Lopez and Raff, Edward and Nicholas, Charles and Holt, James},
booktitle = {AAAI Conference on Artificial Intelligence},
year = {2022},
pages = {7877-7885},
doi = {10.1609/AAAI.V36I7.20757},
url = {https://mlanthology.org/aaai/2022/nguyen2022aaai-out/}
}