Alleviating catastrophic forgetting using context-dependent gating and synaptic stabilization
Alleviating catastrophic forgetting using context-dependent gating and synaptic stabilization
复制标题
DOI：
10.1073/pnas.1803839115
复制
发表时间：
2018-10-30
影响因子：
11.1
通讯作者：
Freedman, David J.
中科院分区：
文献类型：
--
作者：
Masse, Nicolas Y.;Grant, Gregory D.;Freedman, David J.
人类和大多数动物都可以在不忘记旧任务的情况下学习新任务。然而，在新任务上训练人工神经网络（ANN）通常会导致它们忘记以前学习的任务。这种现象是“灾难性遗忘”的结果，在这种情况下，训练ANN会破坏对解决先前任务很重要的连接权重，从而降低任务性能。最近的几项研究提出了稳定人工神经网络连接权重的方法，这些方法被认为对解决任务最重要，这有助于减轻灾难性遗忘。在这里，从被认为是在体内实现的算法的灵感，我们提出了一个补充的方法：添加一个上下文相关的门控信号，这样，只有稀疏的，大多是不重叠的模式的单位是活跃的任何一个任务。这种方法易于实现，需要很少的计算开销，并允许ANN在大量顺序呈现的任务中保持高性能，特别是当与权重稳定相结合时。我们证明了这种方法适用于前馈和递归网络架构，使用监督或基于递归的学习进行训练。这表明，使用多种互补的方法，类似于人们认为发生在大脑中的方法，可以成为支持持续学习的高效策略。
Humans and most animals can learn new tasks without forgetting old ones. However, training artificial neural networks (ANNs) on new tasks typically causes them to forget previously learned tasks. This phenomenon is the result of "catastrophic forgetting," in which training an ANN disrupts connection weights that were important for solving previous tasks, degrading task performance. Several recent studies have proposed methods to stabilize connection weights of ANNs that are deemed most important for solving a task, which helps alleviate catastrophic forgetting. Here, drawing inspiration from algorithms that are believed to be implemented in vivo, we propose a complementary method: adding a context-dependent gating signal, such that only sparse, mostly nonoverlapping patterns of units are active for any one task. This method is easy to implement, requires little computational overhead, and allows ANNs to maintain high performance across large numbers of sequentially presented tasks, particularly when combined with weight stabilization. We show that this method works for both feedforward and recurrent network architectures, trained using either supervised or reinforcement-based learning. This suggests that using multiple, complementary methods, akin to what is believed to occur in the brain, can be a highly effective strategy to support continual learning.