Can Bad Teaching Induce Forgetting? Unlearning in Deep Networks using an Incompetent Teacher

This code is official implementation of Can Bad Teaching Induce Forgetting? Unlearning in Deep Networks using an Incompetent Teacher.

By Vikram S Chundawat, Ayush K Tarun, Murari Mandal, Mohan Kankanhalli

Unlearning Methods Supported:

1. Unlearning using an Incompetent Teacher (presented in this work)
2. Amnesiac Unlearning [presented as Unlearning in Graves, Nagisetty, and Ganesh 2021 (https://ojs.aaai.org/index.php/AAAI/article/view/17371/17178)]
3. UNSIR [presented as in Tarun et al. 2021 (https://arxiv.org/pdf/2111.08947.pdf)]

Models supported:

1. ResNet18
2. Vision Transformer (ViT) (has some issues with older pytorch versions)
3. AllCNN (comment transforms.Resize in datasets.py to work)

Datasets supported:

1. CIFAR-Super20 (mentioned in the paper)
2. CIFAR-10 (works similar to CIFAR-Super20, can be imported from torchvision.datasets and used)
3. CIFAR-100 (works similar to CIFAR-Super20, can be imported from torchvision.datasets and used)

Metrics Supported:

1. Performance (Accuracy)
2. Activation Distance
3. Membership Inference Attack