Safe Policy Search for Lifelong Reinforcement Learning with Sublinear Regret

Lifelong reinforcement learning provides a promising framework for developing versatile agents that can accumulate knowledge over a lifetime of experience and rapidly learn new tasks by building upon