Class-Incremental Learning with Causal Relational Replay (CRR)

This repository hosts the codebase corresponding to our paper, titled 'Class-Incremental Learning with Causal Relational Replay', published at Expert Systems With Applications.

Installation and Usage.

1. Install the dependencies

pip install -r requirements.txt

2. Use ./utils/main.py to run experiments.

python ./utils/main.py --model crr --seed 2711 --end_lr 0.4 --dataset seq-cifar10 --buffer_size 600 --lr 0.9 --batch_size 64 --minibatch_size 64 --batch_size_test 128 --n_epochs 60 --alpha 0.75 --beta 1.75 --gamma 1.25 --csv_log

python ./utils/main.py --model crr --seed 2711 --end_lr 0.32 --dataset seq-cifar100 --buffer_size 600 --lr 0.9 --batch_size 64 --minibatch_size 64 --batch_size_test 128 --n_epochs 60 --alpha 0.75 --beta 1.75 --gamma 1.25 --csv_log

python ./utils/main.py --model crr --seed 2711 --end_lr 0.36 --dataset seq-core50 --buffer_size 600 --lr 0.9 --batch_size 48 --minibatch_size 48 --batch_size_test 48 --n_epochs 20 --alpha 0.75 --beta 1.75 --gamma 1.25 --csv_log

Acknowledge:

We have built our implementation using the Mammoth toolbox and express our gratitude to the authors for providing an excellent repository:

• Mammoth - An Extendible (General) Continual Learning Framework for Pytorch (link).

Citation

If you employ the codes or datasets provided in this repository or utilise our proposed method as comparison baselines in your experiments, please cite our paper. Again, thank you for your interest!

@article{nguyen2024class,
  title={Class-incremental learning with causal relational replay},
  author={Nguyen, Toan and Kieu, Duc and Duong, Bao and Kieu, Tung and Do, Kien and Nguyen, Thin and Le, Bac},
  journal={Expert Systems with Applications},
  volume={250},
  pages={123901},
  year={2024},
  publisher={Elsevier}
}