In the field of artificial intelligence, lifelong learning refers to the ability of a machine to continuously update and improve its knowledge base and models by continuously receiving new data and experience. This learning method mimics the characteristics of human learning, that is, gradually improving the ability to understand and solve problems through continuous learning and experience accumulation over time. In artificial intelligence, lifelong learning pays special attention to solving problems such as catastrophic forgetting, that is, not losing memory and knowledge of old tasks when learning new tasks.

The concept of lifelong learning was proposed by Thrun and Mitchell in 1995. They mainly studied four research directions of lifelong learning in machine learning: lifelong supervised learning, lifelong unsupervised learning, lifelong semi-supervised learning and lifelong reinforcement learning. In addition, the team led by Associate Professor Fang Lu from the Department of Electronic Engineering at Tsinghua University pioneered the intelligent optical computing lifelong learning architecture, breaking through the single function limitation of optical networks and supporting brain-like parallel multi-task learning.

References

【1】Thrun S, Mitchell T M. Lifelong robot learning. In: Steels L,ed. The Biology and Technology of Intelligent Autonomous Agents. Berlin: Springer,1995, 165–196