深度学习论文Lifelong Machine Learning Systems Beyond Learning Algorithms_20180118194149.pdfVIP

Lifelong Machine Learning Systems: Beyond Learning Algorithms Daniel L. Silver Qiang Yang and Lianghao Li Jodrey School of Computer Science Department of Computer Science and Engineering, Acadia University, Hong Kong University of Science and Technology, Wolfville, Nova Scotia, Canada B4P 2R6 Clearwater Bay, Kowloon, Hong Kong Abstract web agents and robotics. And our computing and communi- cation systems now have the capacity to implement and test Lifelong Machine Learning, or LML, considers sys- LML systems. tems that can learn many tasks from one or more do- This paper reviews prior work on LML that uses su- mains over its lifetime. The goal is to sequentially re- pervised, unsupervised or reinforcement learning methods. tain learned knowledge and to selectively transfer that knowledge when learning a new task so as to develop This work has gone by names such as constructive induc- more accurate hypotheses or policies. Following a re- tion, incremental and continual learning, explanation-based view of prior work on LML, we propose that it is learning, sequential task learning, never ending learning, now appropriate for the AI community to move beyond and most recently learning with deep architectures. We then learning algorithms to more seriously consider the na- present our position on the move beyond learning algorithms