Policy Gradient from Demonstration and Curiosity

Jie Chen, Wenjun Xu

With reinforcement learning, an agent could learn complex behaviors from high-level abstractions of the task. However, exploration and reward shaping remained challenging for exist…