24 ms·Hierarchical Deep Reinforcement Learning vs. Montezuma's Revenge1 points by dontreact 10y agodontreact 10y agoThe authors were able to do well on one of the few games previously too difficult for Deep Reinforcement learning. The agent finds subgoals based on object detection.dontreact 10y agoHere is a gif of a few runs https://goo.gl/3Z64Ji https://goo.gl/3Z64Ji