Michał Bortkiewicz

Examining ways to maintain skills in RL during fine-tuning.

Research shows general regularization methods boost off-policy RL agent performance across tasks.

Researchers propose new methods to help learning systems adapt continuously.