darylchang / td_bs Goto Github PK

Training BS Agents Using TD Learning
Addison Weiler and Daryl Chang

Description of files:

run.py - The main file that executes TD learning, trains, and tests agents.
agent.py - Contains implementation of the different agents.
game.py - Implementation of the Game class.
player.py - Implementation of a player (used in the Game class).
evaluation.py - Contains implementation of the simple evaluation function.
test.py - Unit tests to test extractFeatures and simpleEvaluation.
results.xlsx - Raw data for all test results.

--------------------------------------------------------------------------
Please use 'python run.py' to run a series of games.
Command line options:

-train [# of training games]. Default 50.

-run [# of runs]. Default 10.

-p [# of players]. Default 3.

-test [# of test games]. Default 100.

-agent [random/simple/honest/dishonest/bs/model/reflex] (i.e. type of agent). Default model-based reflex agent.

-opp [random/simple/honest/dishonest/bs] (i.e. type of opponents). Default simple reflex agent.

-tournament [y/n] (whether or not to use tournament-style training). Default 'n'.

-human [y/n] (y if you want to play a game against the opponents yourself). Default 'n'.