Learning from prior tasks and transferring that experience to improve future
performance is critical for building lifelong learning agents. Although results
in supervised and reinforcement learning show that transfer may significantly
improve the learning performance, most of the liter