life long learning 相关文章及论文

神经网络之灾难性遗忘问题

 

利用突触智能实现连续学习|从原理到代码解析

Existing approaches can be categorized into three major families based on how the information regarding previous task data is stored and used to mitigate forgetting and potentially support the learning of new tasks. These include replay-based [8, 30] methods which store prior samples, dynamic architectures [34, 37] which add and remove components and prior-focused [18, 39, 9, 7] methods that rely on regularization.