English

F-MALLOC: Feed-forward Memory Allocation for Continual Learning in Neural Machine Translation

Computation and Language 2024-10-23 v2

Abstract

In the evolving landscape of Neural Machine Translation (NMT), the pretrain-then-finetune paradigm has yielded impressive results. However, the persistent challenge of Catastrophic Forgetting (CF) remains a hurdle. While previous work has introduced Continual Learning (CL) methods to address CF, these approaches grapple with the delicate balance between avoiding forgetting and maintaining system extensibility. To address this, we propose a CL method, named F-MALLOC\textbf{F-MALLOC} (F\textbf{F}eed-forward M\textbf{M}emory ALLOCation)\textbf{ALLOC}ation). F-MALLOC is inspired by recent insights highlighting that feed-forward layers emulate neural memories and encapsulate crucial translation knowledge. It decomposes feed-forward layers into discrete memory cells and allocates these memories to different tasks. By learning to allocate and safeguard these memories, our method effectively alleviates CF while ensuring robust extendability. Besides, we propose a comprehensive assessment protocol for multi-stage CL of NMT systems. Experiments conducted following this new protocol showcase the superior performance of F-MALLOC, evidenced by higher BLEU scores and almost zero forgetting.

Keywords

Cite

@article{arxiv.2404.04846,
  title  = {F-MALLOC: Feed-forward Memory Allocation for Continual Learning in Neural Machine Translation},
  author = {Junhong Wu and Yuchen Liu and Chengqing Zong},
  journal= {arXiv preprint arXiv:2404.04846},
  year   = {2024}
}

Comments

Accepted to the main conference of NAACL 2024

R2 v1 2026-06-28T15:46:21.900Z