Research on Adjusting the Proportion of Adversarial Samples
Date:
Supervisor: Gerald Friedland
- Aim to determining the proportion of adversarial training to make the model the robust.
- Try to infering the weakness of the original samples in order to make the classification model stronger.