fsoft-aic/cvd

[Springer Machine Learning Journal] Continual Variational Dropout: A View of Auxiliary Local Variables in Continual Learning

Continual Variational Dropout

This repository contains all of our code for Continual Variational Dropout: A View of Auxiliary Local Variables in Continual Learning.

Dataset

To download the Omniglot dataset.

Perform Training

$ python3 main.py --experiment [dataset] --approach [approach] --film --KL_weight [KL_weight] --prior_var [prior_var]

To perform the CVD on split-mnist, enter the following command:

$ python3 main.py --experiment split_mnist --approach gvclf_vd --film --KL_weight 0.01

To perform CVD on split-CIFAR100, enter the following command:

$ python3 main.py --experiment split_cifar100 --approach gvclf_vd --film --KL_weight 0.01 --conv_Dropout --prior_var 1

Acknowledgement

Our implementation is based on yolky/gvcl.

Citation

@article{hainam2023continual,
  title={Continual variational dropout: a view of auxiliary local variables in continual learning},
  author={Hai, Nam Le and Nguyen, Trang and Van, Linh Ngo and Nguyen, Thien Huu and Than, Khoat},
  journal={Machine Learning},
  pages={1--43},
  year={2023},
  publisher={Springer}
}