In contrast to the natural capabilities of humans to learn new tasks in a sequential fashion, neural networks are known to suffer from catastrophic forgetting, where the model's performances on old tasks drop dramatically after being optimized for a new task. Since then, the continual learning (CL) community has proposed several solutions aiming to equip the neural network with the ability to learn the current task (plasticity) while still achieving high accuracy on the previous tasks (stability). Despite remarkable improvements, the plasticity-stability trade-off is still far from being solved and its underlying mechanism is poorly understood. In this work, we propose Auxiliary Network Continual Learning (ANCL), a novel method that applies an additional auxiliary network which promotes plasticity to the continually learned model which mainly focuses on stability. More concretely, the proposed framework materializes in a regularizer that naturally interpolates between plasticity and stability, surpassing strong baselines on task incremental and class incremental scenarios. Through extensive analyses on ANCL solutions, we identify some essential principles beneath the stability-plasticity trade-off.
翻译:与人类在顺序学习新任务时的自然能力相比,神经网络往往会遭受灾难性的遗忘。也就是说,在经过新任务的优化后,模型在旧任务上的表现会急剧下降。自那时以来,连续学习(CL)社区提出了几种解决方案,旨在为神经网络提供在学习当前任务(可塑性)的同时仍在之前的任务上实现高精度(稳定性)的能力。尽管取得了显着的进展,但稳定性和可塑性之间的权衡仍然远未得到解决,其潜在机制也知之甚少。在这项工作中,我们提出了基于辅助网络连续学习(ANCL)的新方法,将另一个能够促进可塑性的辅助网络应用于主要关注稳定性的连续学习模型中。具体而言,所提出的框架采用一个自然插值可塑性和稳定性的正则化器,优于任务增量和类增量场景的强基线。通过对ANCL解决方案的广泛分析,我们确定了稳定性-可塑性权衡之间的一些基本原则。