
Jump-Start Reinforcement Learning
Preprint
Jump-Start Reinforcement Learning (JSRL) enables any pre-existing policy to form a curriculum for an on-policy RL algorithm
Jump-Start Reinforcement Learning
Preprint
JSRL is a meta-algorithm that uses a pre-existing guide-policy along with a learned exploration policy to quickly explore and solve new tasks.
By using the guide-policy to form a curriculum of starting states for the exploration-policy, we are able to efficiently improve performance on a set of simulated robotic tasks.