Redefining Machine Unlearning: A Conformal Prediction-Motivated Approach

This is the official repo for the paper: Redefining Machine Unlearning: A Conformal Prediction-Motivated Approach.

News

• [2025.06] 👏👏 Accepted by MuGen @ ICML 2025.
• [2025.01] 🚀🚀 Release the paper.

Abstract

Machine unlearning aims to eliminate the impact of specific data on a trained model. Although metrics like unlearning accuracy (UA) and membership inference attack (MIA) are commonly used to evaluate forgetting quality, they fall short in capturing the reliability of forgetting. In this work, we observe that even when data are misclassified according to UA and MIA, their ground truth labels can still remain within the predictive set from an uncertainty quantification perspective, revealing a fake unlearning issue. To better assess forgetting quality, we propose two novel metrics inspired by conformal prediction that offer a more faithful evaluation of forgetting reliability. Building upon these insights, we further introduce a conformal prediction-guided unlearning framework that integrates the Carlini & Wagner adversarial loss. This framework effectively encourages the exclusion of ground truth labels from the conformal prediction set. Extensive experiments on image classification tasks demonstrate the effectiveness of our proposed metrics. By incorporating a tailored loss term, our unlearning framework improves the UA of existing unlearning methods by an average of 6.6%.

File Tree

Project file structure and description:

Conformal-Prediction-Unlearning
├─ README.md
├─ requirements.txt
├─ metrics	# package of our metrics (CR and MIACR)
│    ├─ CR.py
│    ├─ MIACR.py
├─ models	# package of models (ResNet-18 and Vit)
│    ├─ resnet.py
│    ├─ vit.py
├─ main_original_model.py
├─ main_unlearn.py
├─ main_unlearn_cpu.py
├─ main_evaluate.py
├─ unlearn.py
├─ unlearn_cpu.py
└─ utils.py

Setup

Installation requirements are described in requirements.txt.

• Use pip:

  pip install -r requirements.txt
  

• Use anaconda:

  conda install --file requirements.txt
  

Getting Started

Get an original model with the ResNet-18 or ViT architecture:

python main_original_model.py --model_name resnet18 --data_name cifar10 --data_dir ./data --batch_size 64 --num_epochs 200 --learning_rate 0.1 --num_classes 10 

To use the implemented logging, you’ll need a wandb.ai account. Alternatively, you can replace it with any logger of your preference.

To get an unlearning model with one of the existing unlearning methods, use the following command:

python main_unlearn.py --unlearn_name retrain --unlearn_type random --model_name resnet18 --data_name cifar10 --data_dir ./data --model_dir original_model.pth --num_epochs 200 --num_classes 10 --retain_ratio 0.9 --learning_rate 0.01

To get an unlearning model with our unlearning framework CPU, use the following command:

python main_unlearn_cpu.py --unlearn_name finetune --unlearn_type random --model_name resnet18 --data_name cifar10 --data_dir ./data --model_dir original_model.pth --num_epochs 20 --num_classes 10 --retain_ratio 0.9 --learning_rate 0.1 --delta 0.01 --alpha 0.05 --lamda 0.5

After unlearning the forget data, use main_evaluate.py to measure the unlearning model's performance by CR and MIACR metrics:

python main_evaluate.py --unlearn_name retrain --unlearn_type random --model_name resnet18 --data_name cifar10 --data_dir ./data --model_dir unlearning_model.pth --num_classes 10 --retain_ratio 0.9 --alphas 0.05

How to Cite

@article{shi2025redefining,
  title={Redefining machine unlearning: A conformal prediction-motivated approach},
  author={Shi, Yingdan and Liu, Sijia and Wang, Ren},
  journal={arXiv preprint arXiv:2501.19403},
  year={2025}
}