CORE: Mitigating Catastrophic Forgetting in Continual Learning through Cognitive Replay

(Accepted by COGSCI 2024 Oral)

Framework

This is a PyTorch implementation of the CORE

Installation & requirements

The current version of the code has been tested with Python 3.10.4 on a Fedora operating system with the following versions of PyTorch and Torchvision:

• pytorch 1.11.0
• torchvision 0.12.0

Further Python-packages used are listed in requirements.txt. Assuming Python and pip are set up, these packages can be installed using:

pip install -r requirements.txt

Run

python main.py --replay=buffer --target=True

Settings

Custom individual experiments in the academic continual learning setting can be run with main.py. The main options of this script are:

• --experiment: how to construct the context set? (splitMNIST|CIFAR10|CIFAR100)

• --contexts: how many contexts?

• --seed: seed count