HN Theater @HNTheaterMonth

This video explains AlphaGo Zero! AlphaGo Zero uses less prior information about Go than AlphaGo. Whereas AlphaGo is initialized by supervised learning on human experts mappings from state to action; AlphaGo Zero is trained from scratch through self-play. AlphaGo Zero achieves this by combining the policy and value networks into a single, residual neural network. AlphaGo Zero also enhances the self-play training loop with MCTS and uses the MCTS action distributed to train its own policy / value network!
Thanks for watching! Please subscribe!
Paper Link:
DeepMind Blog Post:

This video is a member of the series "From AlphaGo to MuZero" covering the progression of DeepMind's board game agents, inspired by the Kaggle Connect X competition!