期刊
PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA
卷 114, 期 13, 页码 3521-3526出版社
NATL ACAD SCIENCES
DOI: 10.1073/pnas.1611835114
关键词
synaptic consolidation; artificial intelligence; stability plasticity; continual learning; deep learning
资金
- Wellcome Trust
- Engineering and Physical Sciences Research Council
- Google Faculty Award
- Engineering and Physical Sciences Research Council [EP/M019780/1] Funding Source: researchfish
- EPSRC [EP/M019780/1] Funding Source: UKRI
The ability to learn tasks in a sequential fashion is crucial to the development of artificial intelligence. Until now neural networks have not been capable of this and it has been widely thought that catastrophic forgetting is an inevitable feature of connectionist models. We show that it is possible to overcome this limitation and train networks that can maintain expertise on tasks that they have not experienced for a long time. Our approach remembers old tasks by selectively slowing down learning on the weights important for those tasks. We demonstrate our approach is scalable and effective by solving a set of classification tasks based on a hand-written digit dataset and by learning several Atari 2600 games sequentially.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据