AS Seminar: N-to-1 Knowledge Transfer in Reinforcement learning tasks

Deep reinforcement learning has shown an ability to achieve super-human performance in solving complex reinforcement learning tasks only from raw-pixels. However, it fails to reuse knowledge from previously learnt tasks to solve new, unseen ones. To generalize and reuse knowledge is one of the fundamental requirements for creating a truly intelligent agent. In this seminar, I’ll discuss my ongoing work on N-to-1 knowledge transfer, where an agent leverages insights from N previously solved tasks to tackle new challenges.