80 likes | 321 Views
CS 4701 – Project Proposal. Jane Park (jp624) Ran Zhao (rz54). Problem Statement and Motivation Create an efficient and powerful Chinese Checkers AI Challenges: Irregular hexagonal board High branching factor Greedy solutions are suboptimal
E N D
CS 4701 – Project Proposal Jane Park (jp624) Ran Zhao (rz54)
Problem Statement and Motivation • Create an efficient and powerful Chinese Checkers AI • Challenges: • Irregular hexagonal board • High branching factor • Greedy solutions are suboptimal • Opponent plays an important role (can be leveraged) • Motivation: • TD Gammon – teaching itself by playing against itself • A heart-felt desire for academic enlightenment in the area of artificial intelligence
I/O Specification • Input: • Board configuration • Training data • Output: • Optimal move as determined by the AI System.
Background Reading • “Temporal Difference Learning and TD-Gammon” by Gerald Tesauro • “Algorithm Design” by Jon Kleinberg • “Machine Learning” by Tom Mitchell • “Effective Java” by Joshua Bloch • “Art of War” by Sun Tzu
General Approaches • Uncle Scrooge • Green Mushroom • Lime Soda • ThorstenJoachims • 中华人民共和国 • Hard Code
System Architecture and Work Plan • Components • Data collection and aggregation • Machine Learning system • Strategic Game Play system • Tasks • Basic AIs • Data Collection • Machine Learning • Parallel Computing • Distributed Caching • Marketing
Data sources • Non-deterministic AI Players • Human players • Hand generated
Evaluation Plans • Number of turns to win against the following opponents and the margin of victory: • No opponent • JP624 • Greedy AI • RZ54 • Self • Win/loss ratio against random human players