A novel approach to tackle machine unlearning problems in fine-trained DNN models by creating multiple conflicting objectives to optimize. In order to mitigate the huge-scale problem in the EA algorithms, the histogram-based blocking approach is used to reduce the search space from 11M to 31 with a 5.34 compression rate in memory capacity. In this project, the CIFAR10 dataset is studied on the ResNet-18 model.
Let's create virtual environment (venv) in the project and install packages using pip.
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
Simply run the main.py to unlearn 500 data in CIFAR10 data.
python main.py --model_path ".\input\models\resnet_cifar10_epochs10_state_dict" --block_path ".\input\blocks\solution_197bins_remerged.pickle"
- Fine-trained DNN
- Multiple Conflicted Objectives
- Unlearning using MOO algorithms
- Pareto Frontier
- Selection
Distributed under the MIT License. See LICENSE.txt for more information.
Rasa Khosrowshahli - rkhosrowshahli@brocku.ca