DEEP RED An Intelligent Approach to Chinese Checkers.

9
DEEP RED An Intelligent Approach to Chinese Checkers

Transcript of DEEP RED An Intelligent Approach to Chinese Checkers.

Page 1: DEEP RED An Intelligent Approach to Chinese Checkers.

DEEP RED

An Intelligent Approach to Chinese Checkers

Page 2: DEEP RED An Intelligent Approach to Chinese Checkers.

The Problem

• Create an Artificial Intelligence that can successfully play Chinese Checkers, with a random board.

Page 3: DEEP RED An Intelligent Approach to Chinese Checkers.

Strategy Overview

• Complicated games like chess and the modified game of Chinese Checkers have complicated strategies

• We will likely not be able to experience all the nuances of this strategy in the class period

• We want to use Machine Learning, so that the machine might learn complicated winning strategies

Page 4: DEEP RED An Intelligent Approach to Chinese Checkers.

Reading / External Sources

• We will be using the sections on machine learning (18-20) in R+N

• Possibly other data sources, depending on complexity of theory behind implementation

Page 5: DEEP RED An Intelligent Approach to Chinese Checkers.

I/O

• Inputo(0, initialization of board)o(1, move request)o(2, enemy move)o(3, error number)o(4, end game)

Page 6: DEEP RED An Intelligent Approach to Chinese Checkers.

System Architecture

We will use as data sources the games that both the heuristic and learning bot play

Page 7: DEEP RED An Intelligent Approach to Chinese Checkers.

Data Sources

• Our learning agent will provide data collected over the course of design

• If other data sources are used, we may incorporate them, if they pass a sanity check

Page 8: DEEP RED An Intelligent Approach to Chinese Checkers.

Evaluation Plans

• Initially our AI will compete with a heuristic bot• After it learns enough to best the heuristic bot most

of the time (90%, optimistically), we will train it against itself

• At all times, the data collected will go towards the learning algorithm

Page 9: DEEP RED An Intelligent Approach to Chinese Checkers.

I/O

• Inputo(0, initialization of board)o(1, move request)o(2, enemy move)o(3, error number)o(4, end game)