Dynamic Analysis of Multiagent Q-learning with e-greedy Exploration

author:Eduardo Rodrigues Gomes, Swinburne University of Technology
published: Aug. 26, 2009, recorded: June 2009, views: 69

Turn off the lights

You might be experiencing some problems with Your Video player.

Slides

Slides
0:00	Dynamic Analysis of Multiagent Q-learning with e-greedy Exploration
0:02	Motivation (1)
0:25	Motivation (2)
0:34	Motivation (3)
0:49	Motivation (4)
0:59	Motivation (5)
1:05	Motivation (6)
1:07	Motivation (7)
1:37	RL and Evolutionary Game Theory (1)
1:38	RL and Evolutionary Game Theory (2)
2:22	RL and Evolutionary Game Theory (3)
2:30	RL and Evolutionary Game Theory (4)
2:40	Background (1)
3:07	Background (2)
3:51	Action-selection mechanism (1)
4:26	Action-selection mechanism (2)
4:56	Modelling the algorithm (1)
4:58	Modelling the algorithm (2)
5:04	Modelling the algorithm (3)
5:08	Modelling the algorithm (4)
5:14	Modelling the algorithm (5)
5:20	Modelling the algorithm (6)
5:30	Notation (1)
5:32	Notation (2)
5:43	Notation (3)
5:56	Notation (4)
6:41	Continuous-time version (1)
6:45	Continuous-time version (2)
7:08	Continuous-time version (3)
7:13	Continuous-time version (4)
7:17	Continuous-time version (5)
7:28	Continuous-time version (5)
7:34	Continuous-time version (6)
7:47	Limit of the equation (1)
7:48	Limit of the equation (2)
8:00	Limit of the equation (3)
8:16	Non-learning adversary with pure strategy (1)
8:46	Non-learning adversary with pure strategy (2)
9:10	Non-learning adversary with mixed strategy (1)
9:14	Non-learning adversary with mixed strategy (2)
9:28	Non-learning adversary with mixed strategy (3)
9:33	Non-learning adversary with mixed strategy (4)
9:41	Learning adversary (1)
9:43	Learning adversary (2)
10:03	Learning adversary (3)
10:10	Learning adversary (4)
10:21	Learning adversary (5)
10:32	Learning adversary (6)
10:44	Learning adversary (7)
10:51	Learning adversary (8)
11:32	The effects of the e-greedy (1)
11:34	The effects of the e-greedy (2)
11:50	The effects of the e-greedy (3)
12:02	The effects of the e-greedy (4)
12:18	The effects of the e-greedy (5)
12:23	The effects of the e-greedy (6)
12:44	Summary of the analysis (roughly speaking) (1)
12:46	Summary of the analysis (roughly speaking) (2)
12:47	Summary of the analysis (roughly speaking) (3)
12:48	Summary of the analysis (roughly speaking) (4)
13:18	System of difference equations (1)
13:19	System of difference equations (2)
13:24	System of difference equations (3)
13:54	Prisoner’s Dilemma (1)
13:55	Prisoner’s Dilemma (2)
13:57	Prisoner’s Dilemma (3)
14:54	Prisoner’s Dilemma (4)
15:36	Battle of the Sexes (1)
15:46	Battle of the Sexes (2)
15:50	A game with no equilibrium (1)
15:55	A game with no equilibrium (2)
16:02	A game with no equilibrium (3)
16:38	A game with no equilibrium (4)
16:57	Conclusions (1)
16:59	Conclusions (2)
17:05	Conclusions (3)
17:11	Future Works
17:40	- 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.

Description

The development of mechanisms to understand and model the expected behaviour of multiagent learners is becoming increasingly important as the area rapidly find application in a variety of domains. In this paper we present a framework to model the behaviour of Q-learning agents using the e-greedy exploration mechanism. For this, we analyse a continuous-time version of the Q-learning update rule and study how the presence of other agents and the e-greedy mechanism affect it. We then model the problem as a system of difference equations which is used to theoretically analyse the expected behaviour of the agents. The applicability of the framework is tested through experiments in typical games selected from the literature.