Residual Continual Learning

Authors

  • Janghyeon Lee KAIST
  • Donggyu Joo KAIST
  • Hyeong Gwon Hong KAIST
  • Junmo Kim KAIST

DOI:

https://doi.org/10.1609/aaai.v34i04.5884

Abstract

We propose a novel continual learning method called Residual Continual Learning (ResCL). Our method can prevent the catastrophic forgetting phenomenon in sequential learning of multiple tasks, without any source task information except the original network. ResCL reparameterizes network parameters by linearly combining each layer of the original network and a fine-tuned network; therefore, the size of the network does not increase at all. To apply the proposed method to general convolutional neural networks, the effects of batch normalization layers are also considered. By utilizing residual-learning-like reparameterization and a special weight decay loss, the trade-off between source and target performance is effectively controlled. The proposed method exhibits state-of-the-art performance in various continual learning scenarios.

Downloads

Published

2020-04-03

How to Cite

Lee, J., Joo, D., Hong, H. G., & Kim, J. (2020). Residual Continual Learning. Proceedings of the AAAI Conference on Artificial Intelligence, 34(04), 4553-4560. https://doi.org/10.1609/aaai.v34i04.5884

Issue

Section

AAAI Technical Track: Machine Learning