Test optimization in DNN testing refers to the activities that focus on alleviating the
cost of data labeling during the DNN testing phase. Researchers try to select and label
a subset of the test data or directly use the unlabeled test data to satisfy testing
requirements. Related methods that need to label a part of test data are called
sampling-based testing methods, while others that need no labeling effort are named
labeling-free test methods.
Survey paper
Each paper could cover more than one topics.
In the paper
| All Papers: | 90 |
| Fault Detection: | 47 |
| Sampling-Based Model Retraining: | 20 |
| Model Selection: | 3 |
| Performance Estimation: | 33 |
Up to date
| All Papers: | 128 |
| Fault Detection: | 57 |
| Sampling-Based Model Retraining: | 22 |
| Model Selection: | 4 |
| Performance Estimation: | 54 |
@article{10.1145/3643678,
title={Test Optimization in DNN Testing: A Survey},
author={Hu, Qiang and Guo, Yuejun and Xie, Xiaofei and Cordy, Maxime and Ma, Lei and Papadakis, Mike and Le Traon, Yves},
journal={ACM Trans. Softw. Eng. Methodol.},
year={2024}
}