Configuration Path Control

Sergey Pankov

Reinforcement learning methods often produce brittle policies -- policies that perform well during training, but generalize poorly beyond their direct training experience, thus bec…