Patch-Based Contrastive Learning and Memory Consolidation for Online Unsupervised Continual Learning

PyTorch code for the COLLAS 2024 paper:
Patch-Based Contrastive Learning and Memory Consolidation for Online Unsupervised Continual Learning
Cameron Taylor, Vassilis Vassiliades, Constantine Dovrolis
International Joint Conference on Artificial Intelligence (IJCAI), 2021\

Requirements

To install Python 3 requirements:

python3 -m venv .pcmc
source .pcmc/bin/activate
pip install -r requirements.txt

Basic Experiments

The bash scripts to run each model are included in the run_scripts folder. An example run for PCMC on the imagenet40 dataset.

python3 main.py log=pcmc/$1 model=pcmc dataset=imagenet40-long scenario=incremental seed=$2 \
        model.arch=resnet18 model.layers.layer0.feat_size=512 \
        load_pretrain=True pretrain_log=pcmc/seed$2 model.encoder_type=simclr model.sleep_on=True model.pretrained=False \
        scenario.eval_freq=4 model.mem_update=reduce_mem model.update_use=1 model.init_epochs=300 plot=True \
        dataset.super_size=100 dataset.test_size=100 dataset.stream_size=1000 dataset.t0_factor=1.0

Contributing

MIT License

Citation

If you found our work useful for your research, please cite our work:

    @article{taylor2024patch,
         title={Patch-Based Contrastive Learning and Memory Consolidation for Online Unsupervised Continual Learning},
         author={Taylor, Cameron  and Vassiliades, Vassilis  and Dovrolis, Constantine},
         journal={Proceedings of The 3rd Conference on Lifelong Learning Agents},
         year={2024},
    }