Epistemic Uncertainty-Weighted Loss for Visual Bias Mitigation
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2022
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866910470923878400 |
|---|---|
| author | Stone, Rebecca S Ravikumar, Nishant Bulpitt, Andrew J Hogg, David C |
| author_facet | Stone, Rebecca S Ravikumar, Nishant Bulpitt, Andrew J Hogg, David C |
| contents | Deep neural networks are highly susceptible to learning biases in visual data. While various methods have been proposed to mitigate such bias, the majority require explicit knowledge of the biases present in the training data in order to mitigate. We argue the relevance of exploring methods which are completely ignorant of the presence of any bias, but are capable of identifying and mitigating them. Furthermore, we propose using Bayesian neural networks with a predictive uncertainty-weighted loss function to dynamically identify potential bias in individual training samples and to weight them during training. We find a positive correlation between samples subject to bias and higher epistemic uncertainties. Finally, we show the method has potential to mitigate visual bias on a bias benchmark dataset and on a real-world face detection problem, and we consider the merits and weaknesses of our approach. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2204_09389 |
| institution | arXiv |
| publishDate | 2022 |
| record_format | arxiv |
| spellingShingle | Epistemic Uncertainty-Weighted Loss for Visual Bias Mitigation Stone, Rebecca S Ravikumar, Nishant Bulpitt, Andrew J Hogg, David C Computer Vision and Pattern Recognition Artificial Intelligence Computers and Society Machine Learning Deep neural networks are highly susceptible to learning biases in visual data. While various methods have been proposed to mitigate such bias, the majority require explicit knowledge of the biases present in the training data in order to mitigate. We argue the relevance of exploring methods which are completely ignorant of the presence of any bias, but are capable of identifying and mitigating them. Furthermore, we propose using Bayesian neural networks with a predictive uncertainty-weighted loss function to dynamically identify potential bias in individual training samples and to weight them during training. We find a positive correlation between samples subject to bias and higher epistemic uncertainties. Finally, we show the method has potential to mitigate visual bias on a bias benchmark dataset and on a real-world face detection problem, and we consider the merits and weaknesses of our approach. |
| title | Epistemic Uncertainty-Weighted Loss for Visual Bias Mitigation |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence Computers and Society Machine Learning |
| url | https://arxiv.org/abs/2204.09389 |