Event Title

Comparing Efficiency of the Minimax Algorithm to Alternatives

Session Number

F13

Advisor(s)

Phadmakar Patankar, Illinois Mathematics and Science Academy

Location

A-115

Start Date

28-4-2016 8:50 AM

End Date

28-4-2016 9:15 AM

Abstract

Digital technology has become increasingly important in our lives. While many programs have a set capability, we have begun using machines that learn using adaptive algorithms, allowing computers to imitate the flexibility of the human brain, reacting quicker and more effectively. This investigation could provide rationale for alternatives to brute force algorithms. We started by coding a connect four game in Java, implementing a brute force algorithm called Minimax, and developed a version that used limited resources as a foundation for our AI (artificial intelligence). Our data analysis compares the win-loss ratio of the AI to the hypothetical perfect record of Minimax. Preliminary testing suggests that Minimax is far less efficient than AI alternatives; Minimax was unable to calculate a single move in fourteen hours, while the AI takes significantly less time. One of our alternatives to brute force is an algorithm that uses large quantities of random sampling to determine the most effective turn, rather than simulating all possible combinations; This algorithm is slower than one that makes random moves, but more effective. The improved performance demonstrated by our AI could show alternatives to brute force algorithms in real world applications, trading calculation time and effectiveness for increased efficiency.


Share

COinS
 
Apr 28th, 8:50 AM Apr 28th, 9:15 AM

Comparing Efficiency of the Minimax Algorithm to Alternatives

A-115

Digital technology has become increasingly important in our lives. While many programs have a set capability, we have begun using machines that learn using adaptive algorithms, allowing computers to imitate the flexibility of the human brain, reacting quicker and more effectively. This investigation could provide rationale for alternatives to brute force algorithms. We started by coding a connect four game in Java, implementing a brute force algorithm called Minimax, and developed a version that used limited resources as a foundation for our AI (artificial intelligence). Our data analysis compares the win-loss ratio of the AI to the hypothetical perfect record of Minimax. Preliminary testing suggests that Minimax is far less efficient than AI alternatives; Minimax was unable to calculate a single move in fourteen hours, while the AI takes significantly less time. One of our alternatives to brute force is an algorithm that uses large quantities of random sampling to determine the most effective turn, rather than simulating all possible combinations; This algorithm is slower than one that makes random moves, but more effective. The improved performance demonstrated by our AI could show alternatives to brute force algorithms in real world applications, trading calculation time and effectiveness for increased efficiency.