Skip to content

zibinpan/FedOSD

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

7 Commits
 
 
 
 
 
 
 
 

Repository files navigation

Code of the AAAI-25 paper

image

How to run the code

First, run the following command to train a FL global model as the original model for unlearning experiments.

/usr/bin/python run_UnlearningTask.py --seed 1 --device 0 --module CNN_CIFAR10 --algorithm FedAvg --dataloader DataLoader_cifar10_pat --N 10 --NC 2 --balance True --B 200 --C 1.0 --R 2000 --E 1 --lr 0.05 --decay 0.999 --step_type bgd --unlearn_cn 1 --unlearn_pretrain True --save_model True

Next, execute the following command to run the unlearning algorithm.

/usr/bin/python run_UnlearningTask.py --seed 1 --device 0 --module CNN_CIFAR10 --algorithm FedOSD --dataloader DataLoader_cifar10_pat --N 10 --NC 2 --balance True --B 200 --C 1.0 --R 200 --UR 100 --E 1 --decay 0.999 --step_type bgd --unlearn_cn 1 --save_model True --lr 0.0004 --r_lr 1e-6

Background

image

image

image

image

Key Challenges to FU

image

The Proposed FedOSD

image

Unlearning Cross-Entropy (UCE) Loss

image

Orthogonal Steepest Descent Direction

image

image

Model Reverting Issue

image

Experiments

image

image

image

image

image

Conmunication

Welcome to our session during the conference time. We will be there on March 1, 12:00 - 14:30.

Cite

Pan Z, Wang Z, Li C, et al. Federated unlearning with gradient descent and conflict mitigation[C]//Proceedings of the AAAI Conference on Artificial Intelligence. 2025, 39(19): 19804-19812.

About

Code of the AAAI-25 paper

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages