I tried the 2017 competition, and DDPG never converges for me. Since the 30+ dimensional state space is so large, I wonder are there some ML techniques that deal with dimensionality reduction or just large dimensions in general?