Machine Unlearning

(Released on March 27, 2025)

Introduction

This repository contains implementations of several popular machine unlearning algorithms in PyTorch, as listed below:

Algorithm	Paper	Original Github Repository
gradient_ascent	Unrolling SGD: Understanding Factors Influencing Machine Unlearning	https://github.com/cleverhans-lab/unrolling-sgd
bad_teacher	Can Bad Teaching Induce Forgetting? Unlearning in Deep Networks using an Incompetent Teacher	https://github.com/vikram2000b/bad-teaching-unlearning
scrub	Towards Unbounded Machine Unlearning	https://github.com/meghdadk/SCRUB/tree/main
amnesiac	Amnesiac Machine Learning	https://github.com/lmgraves/AmnesiacML
boundary	Boundary Unlearning: Rapid Forgetting of Deep Networks via Shifting the Decision Boundary	https://github.com/TY-LEE-KR/Boundary-Unlearning-Code
ntk	Forgetting Outside the Box: Scrubbing Deep Networks of Information Accessible from Input-Output Observations	https://github.com/AdityaGolatkar/SelectiveForgetting
fisher	Eternal Sunshine of the Spotless Net: Selective Forgetting in Deep Networks	https://github.com/AdityaGolatkar/SelectiveForgetting
unsir	Fast Yet Effective Machine Unlearning	https://github.com/vikram2000b/Fast-Machine-Unlearning
ssd	Fast Machine Unlearning Without Retraining Through Selective Synaptic Dampening	https://github.com/if-loops/selective-synaptic-dampening

Sincere appreciation to the authors of these popular machine unlearning algorithms for open-sourcing their code, greatly contributing to the success of this repository.

Getting started

Preparation

Before executing the project code, please prepare the Python environment according to the requirement.txt file. We set up the environment with python 3.9.12 and torch 2.0.0.

pip install -r requirement.txt

How to run

1. Model Training

python train_main.py -dataset Cifar10 

2. Unlearning

python unlearn_main.py -gpu -dataset Cifar10 -unlearn_class 0 -unlearn_method retrain -model_path 

Feedback

Suggestions and opinions on this work (both positive and negative) are greatly welcomed. Please contact the author by sending an email to winkent.ong at um.edu.my.

License and Copyright

The project is open source under BSD-3 license (see the LICENSE file).

©2024 Universiti Malaya.