Cmput 455 Resources
Links to external resources are provided as-is, for additional
reading, watching and listening, without any guarantees that the material
will stay accessible, be relevant for your needs, or match the
contents of our course.
Resources include introductory books, articles, online courses, blog posts,
demos and video clips.
Background on Algorithms
Online AI Courses
Appendix A: Python bootcamp
These notes cover some required background knowledge.
I expect many of the students to know much of this material,
so it is provided to enable you to fill in any "gaps".
Some of the activities and all of the assignments will
require this knowledge.
Lecture 1 - The game of Go and Computer Go
Go UI's, tools
- Gomill - Tools
and Go library written in Python by Matthew Woodcraft
Open Source Computer Go Programs
- Cmput 455 Go programs.
This series of simple Go programs is used to
illustrate many of the concepts taught in this course.
- Ray, Rayon, Rayn
by Yuandong Tian and Yan Zhu of Facebook. DCNN and MCTS.
Darkforest on github. Written in C and Lua.
- Fuego – C++ libraries and Go playing program.
One of the strongest open source MCTS programs before DCNN.
Developed mainly at Alberta. Currently not under
active development, but still used for research.
- Pachi - one of the strongest
open source MCTS programs before DCNN. Written in C.
Pachi on github.
Minimalistic Go MCTS Engine, by pachi author Petr Baudis.
Only about 500 lines of Python, MCTS and large patterns. No DCNN.
Minimalistic AlphaGo Zero reimplementation,
by pachi author Petr Baudis.
Based on Michi.
version with DCNN.
Written in C++.
One of the currently (as of 2020) strongest open source Go programs.
Trained using the AlphaZero algorithm on 28 GPUs, with a long list of Go specific and general enhancements to speed up the training process and improve playing strength.
Paper describing most of the main improvements can be read here.
Decision-Making Case Studies
How UPS Trucks Saved Millions of Dollars By Eliminating Left Turns
Chances and Risks of Algorithms
Lecture 7 - Search and heuristic search
Blind Search - Depth-first Search, Breadth-first search, etc.
DCNN for Move Prediction in Go
Neural Networks and Deep Learning
Lecture 21 Reinforcement Learning
- 2015 vs Fan Hui
- 2016 vs Lee Sedol:
Alpha Go movie
January 2017, fast games on internet against almost all of the top 20 human professionals. 60 wins no losses for AlphaGo.
Humans Mourn Loss After Google Is Unmasked as China’s Go MasterArticle in WSJ (needs registration or subscription)
- 2017 vs Ke Jie
- 2017 self play games
Article by Carl Johan Ragnarsson on medium.com
Programs giving handicap to professional players
Current Go Programs, mostly Based on AlphaGo Zero Ideas
Recent Computer Go Events
Created: Jul 18, 2016 Last modified: Dec 3, 2020
Martin Müller and Ting-Han Wei