Utilizing Adversarial Examples for Bias Mitigation and Accuracy Enhancement
Fuente:
arXiv
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866916304639754240 |
|---|---|
| author | Shukla, Pushkar Srikanth, Dhruv Cohen, Lee Turk, Matthew |
| author_facet | Shukla, Pushkar Srikanth, Dhruv Cohen, Lee Turk, Matthew |
| contents | We propose a novel approach to mitigate biases in computer vision models by utilizing counterfactual generation and fine-tuning. While counterfactuals have been used to analyze and address biases in DNN models, the counterfactuals themselves are often generated from biased generative models, which can introduce additional biases or spurious correlations. To address this issue, we propose using adversarial images, that is images that deceive a deep neural network but not humans, as counterfactuals for fair model training. Our approach leverages a curriculum learning framework combined with a fine-grained adversarial loss to fine-tune the model using adversarial examples. By incorporating adversarial images into the training data, we aim to prevent biases from propagating through the pipeline. We validate our approach through both qualitative and quantitative assessments, demonstrating improved bias mitigation and accuracy compared to existing methods. Qualitatively, our results indicate that post-training, the decisions made by the model are less dependent on the sensitive attribute and our model better disentangles the relationship between sensitive attributes and classification variables. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2404_11819 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Utilizing Adversarial Examples for Bias Mitigation and Accuracy Enhancement Shukla, Pushkar Srikanth, Dhruv Cohen, Lee Turk, Matthew Computer Vision and Pattern Recognition We propose a novel approach to mitigate biases in computer vision models by utilizing counterfactual generation and fine-tuning. While counterfactuals have been used to analyze and address biases in DNN models, the counterfactuals themselves are often generated from biased generative models, which can introduce additional biases or spurious correlations. To address this issue, we propose using adversarial images, that is images that deceive a deep neural network but not humans, as counterfactuals for fair model training. Our approach leverages a curriculum learning framework combined with a fine-grained adversarial loss to fine-tune the model using adversarial examples. By incorporating adversarial images into the training data, we aim to prevent biases from propagating through the pipeline. We validate our approach through both qualitative and quantitative assessments, demonstrating improved bias mitigation and accuracy compared to existing methods. Qualitatively, our results indicate that post-training, the decisions made by the model are less dependent on the sensitive attribute and our model better disentangles the relationship between sensitive attributes and classification variables. |
| title | Utilizing Adversarial Examples for Bias Mitigation and Accuracy Enhancement |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2404.11819 |