Bipedalwalker V2 With A3c Algorithm

Summary

Agent trained about 30k episodes per worker in ~21h on a single CPU, with 4 workers. drive.google.com/open?id=1r-M1rMtmBe0E3hTQiRZiPrH-r84PO3ZS. deeplearning Source code can be seen here: github.com/Amegatron/ This is one iteration of the walker I trained that solved the task(100 consecutive 300+). The writeup, code and checkpoint file can ... Full implementation - github.com/KaleabTessera/Policy-Gradient. Additional Details can be found at github.com/vinits5/augmented-random-search. Pre-trained Soft Actor Critic agent playing No - Batch Norm Yes - DDPG Yes - Densely Connected Network Yes - Epsilon Greedy. This compiles some snapshot runs during training for solving the