3 ms·Reinforcement Learning Upside Down: Don't Predict Rewards, Map to Actions12 points by wyldfire 7y ago