/HindsightAnchor

Repository to contain Hindsight Anchor Learning Code

Primary LanguagePythonMIT LicenseMIT

Using Hindsight to Anchor Past Knowledge in Continual Learning

This is the official implementation of the Using Hindsight to Anchor Past Knowledge in Continual Learning in Tensorflow v1. The code is built on top of AGEM repository.

@inproceedings{ChaudhryHAL,
    title={Using Hindsight to Anchor Past Knowledge in Continual Learning},
    author={Chaudhry, Arslan and Gordo, Albert and Dokania, Puneet K and Torr, Philip HS and Lopez-Paz, David},
    booktitle={AAAI},
    year={2021}
}

To replicate the results of MNIST Permutations experiment, execute the following script:

$ ./replicate_mnist.sh

License

This source code is released under The MIT License found in the LICENSE file in the root directory of this source tree.