Sparse Multi-Task Reinforcement Learning

Neural Information Processing Systems 

In multi-task reinforcement learning (MTRL), the objective is to simultaneously learn multiple tasks and exploit their similarity to improve the performance w.r.t.