cramer-distance

This paper introduces C-DSAC, a new distributional reinforcement learning algorithm that uses the Cramér distance to improve performance and stability in robotic benchmarks compared to standard SAC.