Sampled Muzero with the EfficientZero and the stochastic muzero to introduce random events into the continuous action space

[LiuDongyang39]

Based on how you combined Sampled Muzero with EfficientZero, my latest idea is to combine Sampled Muzero with the EfficientZero and the stochastic muzero to introduce random events into the continuous action space. My question is is it feasible? Looking forward to your answer. Thank you very much!

[puyuan1996]

Based on my analysis, this approach is highly feasible. Expanding the action space (Sampled MuZero), addressing stochasticity (Stochastic MuZero), and enhancing sample efficiency (EfficientZero) represent orthogonal improvements that can be seamlessly integrated. By combining these advancements, the approach is expected to support the modeling of continuous actions and stochastic events while maintaining strong learning efficiency. We are eagerly looking forward to seeing the corresponding implementation and would like to express our sincere gratitude for your valuable contributions. If you have any questions, please feel free to continue the discussion here at any time!