Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2402.05400 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866909216931840000 |
|---|---|
| author | Lieberman, Kelsey Yuan, Shuai Ravindran, Swarna Kamlam Tomasi, Carlo |
| author_facet | Lieberman, Kelsey Yuan, Shuai Ravindran, Swarna Kamlam Tomasi, Carlo |
| contents | Although binary classification is a well-studied problem in computer vision, training reliable classifiers under severe class imbalance remains a challenging problem. Recent work has proposed techniques that mitigate the effects of training under imbalance by modifying the loss functions or optimization methods. While this work has led to significant improvements in the overall accuracy in the multi-class case, we observe that slight changes in hyperparameter values of these methods can result in highly variable performance in terms of Receiver Operating Characteristic (ROC) curves on binary problems with severe imbalance. To reduce the sensitivity to hyperparameter choices and train more general models, we propose training over a family of loss functions, instead of a single loss function. We develop a method for applying Loss Conditional Training (LCT) to an imbalanced classification problem. Extensive experiment results, on both CIFAR and Kaggle competition datasets, show that our method improves model performance and is more robust to hyperparameter choices. Code is available at https://github.com/klieberman/roc_lct. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2402_05400 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Optimizing for ROC Curves on Class-Imbalanced Data by Training over a Family of Loss Functions Lieberman, Kelsey Yuan, Shuai Ravindran, Swarna Kamlam Tomasi, Carlo Machine Learning Computer Vision and Pattern Recognition Although binary classification is a well-studied problem in computer vision, training reliable classifiers under severe class imbalance remains a challenging problem. Recent work has proposed techniques that mitigate the effects of training under imbalance by modifying the loss functions or optimization methods. While this work has led to significant improvements in the overall accuracy in the multi-class case, we observe that slight changes in hyperparameter values of these methods can result in highly variable performance in terms of Receiver Operating Characteristic (ROC) curves on binary problems with severe imbalance. To reduce the sensitivity to hyperparameter choices and train more general models, we propose training over a family of loss functions, instead of a single loss function. We develop a method for applying Loss Conditional Training (LCT) to an imbalanced classification problem. Extensive experiment results, on both CIFAR and Kaggle competition datasets, show that our method improves model performance and is more robust to hyperparameter choices. Code is available at https://github.com/klieberman/roc_lct. |
| title | Optimizing for ROC Curves on Class-Imbalanced Data by Training over a Family of Loss Functions |
| topic | Machine Learning Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2402.05400 |