hyu2000

hyu2000.github.io

My journey on AlphaGo experiments. First on 5x5 board:

• C2 Opening and intracacies
• 5x5 Go complexity
• Dissecting “super-human performance” of AlphaGo

If 5x5 Go is a puzzle, 9x9 is more like the real deal. And you can finish a game in several minutes instead of hours!

• Dispatch from the trenches: from 5x5 to 9x9
• Training bot to play endgames
• Too many bots
• Some super-human 9x9 games and my attempt at an analysis

More to come …

• 5x5 Go through the lens of KataGo
• The Split-Brain problem, mini-max, and a conservative strategy out of self-play
• AlphaGo lineage, Game theory, open-ended learning
• Computation power, efficiency
• Solving Go puzzles
• Selfplay Evolution

This site is open source. Improve this page.