DeepPAC is a tool to analyse robustness for deep neural networks (DNNs) with the probably approximately correct (PAC) guarantee, which is an an implementation of the algorithms in our technical article titled "Towards Practical Robustness Analysis for DNNs based on PAC-Model Learning".
DeepPAC is now open source under the Apache-2.0 License. We intend to apply the Reusable & Available Badges in the ICSE 2022 artifact evaluation.
Please read the ICSE_Artifact_Evaluation.pdf and ReadMe.pdf for details.
We put large DNN models in release. Please download and put them in models folder manually.
Cifar_ResNet152_adv_trim.pth