juntingzh / incremental-learning-baselines

Efficient Lifelong Learning with A-GEM

This repo is forked from the official implementation of RWalk and AGEM and modified to enable single-headed evaluation for incremental learning.

It is used to generate baseline results for EWC++, SI, MAS, and RWalk reported in our WACV 2020 paper Class-incremental Learning via Deep Model Consolidation.

Requirements

TensorFlow >= v1.9.0.

Training

To replicate the results of the paper on a particular dataset, execute (see the Note below for downloading the CUB and AWA datasets):

$ ./replicate_results.sh <DATASET> <THREAD-ID> <JE>

Example runs are:

$ ./replicate_results.sh MNIST 3      /* Train PNN and A-GEM on MNIST */
$ ./replicate_results.sh CUB 1 1      /* Train JE models of RWALK and A-GEM on CUB */

Note

For CUB and AWA experiments, download the dataset prior to running the above script. Run following for downloading the datasets:

$ ./download_cub_awa.sh

The plotting code is provided under the folder plotting_code/. Update the paths in the plotting code accordingly.

When using this code, please consider cite our paper:

@inproceedings{zhang2020class,
	title={Class-incremental learning via deep model consolidation},
	author={Zhang, Junting and Zhang, Jie and Ghosh, Shalini and Li, Dawei and Tasci, Serafettin and Heck, Larry and Zhang, Heming and Kuo, C-C Jay},
	booktitle={2020 IEEE Winter Conference on Applications of Computer Vision (WACV)},
	year={2020},
	organization={IEEE}
}

and the papers by the original author:
@inproceedings{AGEM,
  title={Efficient Lifelong Learning with A-GEM},
  author={Chaudhry, Arslan and Ranzato, Marc’Aurelio and Rohrbach, Marcus and Elhoseiny, Mohamed},
  booktitle={ICLR},
  year={2019}
}

@inproceedings{chaudhry2018riemannian,
  title={Riemannian Walk for Incremental Learning: Understanding Forgetting and Intransigence},
  author={Chaudhry, Arslan and Dokania, Puneet K and Ajanthan, Thalaiyasingam and Torr, Philip HS},
  booktitle={ECCV},
  year={2018}
}

License

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