EPSRC Winter School in Mathematics for Data Modelling, Sheffield 2008

Applications of Machine Learning to the Game of Go

author: David Stern, Microsoft Research
published: Feb. 5, 2008, recorded: January 2008, views: 630

Slides

Slides
0:00	Applications of Machine Learning to the Game of Go
0:34	Contents
1:13	The Game of Go
2:32	One Eye = Death
2:56	Two Eyes = Life
3:17	Computer Go - 1
4:07	Computer Go - 2
5:27	Use of Knowledge in Computer Go
6:48	Uncertainty in Go
7:36	Move Prediction
8:12	Pattern Matching for Move Prediction
8:38	Patterns - 1
8:45	Patterns - 2
8:58	Patterns - 3
8:59	Patterns - 4
9:00	Patterns - 5
9:02	Patterns - 6
9:07	Patterns - 7
9:28	Pattern Matching
10:09	Pattern Hash Key
11:04	Harvesting
12:38	Relative Frequencies of Pattern Sizes
13:30	Training
14:30	Matching Largest Pattern
15:02	Bayesian Ranking Model - 1
16:23	Bayesian Ranking Model - 2
17:38	Bayesian Ranking Model - 3
17:50	Bayesian Ranking Model - 4
18:10	Message Passing
18:33	Marginal Calculation
18:41	- Questions
21:12	Gaussian Message Passing - 2
21:59	Gaussian Message Passing - 3
22:17	Gaussian Message Passing - 4
22:40	Move Prediction Performance
25:00	Rank Error vs Game Phase
25:18	Rank Error vs Pattern Size
25:28	Hierarchical Gaussian Model of Move Values
26:26	Pattern Hierarchy
26:56	Hierarchical Gaussian Model
27:18	Move Prediction Performance
27:25	Predictive Probability
27:53	Territory Prediction
28:25	Territory
30:50	Predicting Territory - 1
31:11	- Questions
31:46	Predicting Territory - 3
31:51	Predicting Territory - 4
31:53	Predicting Territory - 3
31:57	Predicting Territory - 4
33:30	- Questions
35:06	Game Records
35:07	1000,000 Expert Games…
35:18	Final Position
35:24	Final Position + Territory
35:29	Position + Final Territory: Train CRF by Maximum Likelihood
35:43	Position + Boltzmann Sample - 1
36:45	Position + Boltzmann Sample - 2
36:51	Boltzmann Machine – Expectation over Swendsen-Wang Samples
37:30	Position + Boltzmann Sample - 3
38:37	- Questions
39:42	Monte Carlo Go - 1
39:56	Monte Carlo Go - 2
41:16	Monte Carlo Go - 3
41:34	Adaptive Monte Carlo Planning
43:01	Adaptive Monte Carlo Go - 1
44:01	Adaptive Monte Carlo Go - 2
44:21	Adaptive Monte Carlo Go - 3
44:26	Adaptive Monte Carlo Go - 4
44:51	Bayesian Model For Policy
46:34	Bayesian’ Adaptive MC
47:10	‘Bayesian’ Adaptive MC (BMC)
47:50	Exploitation vs Exploration - 1
48:23	- Questions
50:05	MC Planning on P-Game Trees
53:20	- Questions

Report a problem or upload files

If you have found a problem with this lecture or would like to send us extra material, articles, exercises, etc., please use our ticket system to describe your request and upload the data.
Enter your e-mail into the 'Cc' field, and we will keep you updated with your request's status.

Lecture popularity: You need to login to cast your vote.