Overcoming Catastrophic Forgetting beyond Continual Learning: Balanced Training for Neural Machine Translation
This work tackles a fundamental issue in neural network training for machine translation, though it is incremental as it builds on existing knowledge distillation techniques.
The paper identifies that neural machine translation models suffer from catastrophic forgetting even in static training due to imbalanced attention to samples, and proposes Complementary Online Knowledge Distillation to address this, achieving substantial improvements over baselines.
Neural networks tend to gradually forget the previously learned knowledge when learning multiple tasks sequentially from dynamic data distributions. This problem is called \textit{catastrophic forgetting}, which is a fundamental challenge in the continual learning of neural networks. In this work, we observe that catastrophic forgetting not only occurs in continual learning but also affects the traditional static training. Neural networks, especially neural machine translation models, suffer from catastrophic forgetting even if they learn from a static training set. To be specific, the final model pays imbalanced attention to training samples, where recently exposed samples attract more attention than earlier samples. The underlying cause is that training samples do not get balanced training in each model update, so we name this problem \textit{imbalanced training}. To alleviate this problem, we propose Complementary Online Knowledge Distillation (COKD), which uses dynamically updated teacher models trained on specific data orders to iteratively provide complementary knowledge to the student model. Experimental results on multiple machine translation tasks show that our method successfully alleviates the problem of imbalanced training and achieves substantial improvements over strong baseline systems.