Event Title

Quantifying How Long Learning Takes: A Look at Machine Learning and Blackjack

Location

CSU Ballroom

Start Date

20-4-2015 2:00 PM

End Date

20-4-2015 3:30 PM

Student's Major

Integrated Engineering

Student's College

Science, Engineering and Technology

Mentor's Name

Rebecca Bates

Mentor's Email Address

rebecca.bates@mnsu.edu

Mentor's Department

Integrated Engineering

Mentor's College

Science, Engineering and Technology

Second Mentor's Name

Dean Kelley

Second Mentor's Email Address

dean.kelley@mnsu.edu

Second Mentor's Department

Computer Information Science

Second Mentor's College

Science, Engineering and Technology

Third Mentor's Name

Jennifer Veltsos

Third Mentor's Email Address

jennifer.veltsos@mnsu.edu

Third Mentor's Deparment

English

Third Mentor's College

Arts and Humanities

Description

Machine learning is a form of artificial intelligence that allows a computer program to learn the most optimal way to act, instead of following explicitly programmed instructions. It is a form of pattern recognition that is being implemented and experimented with in many different fields and areas. The overall goal of this work is to find out how long it takes for machine learning to perform optimally in the simple game of black jack. We expect to distinguish the relationship between number of training hands played and how well the agent performs. To do this we are working to design a machine learning agent that has no knowledge of the game of blackjack. The agent is given X amount of training hands to play. These hands are randomly generated from a shoe of 6 decks and each deck is re-shuffled upon reaching the near end of the shoe. Each hand the agent is dealt, along with the dealer’s up card, is stored in a data structure that represents this pairing. The agent will then hit or stay randomly and the result of whether they lost or not is tracked. After these X training hands have been completed, the agent then plays a large number of new hands using the statistics of which option (hitting or staying) failed the least during the training hands to make its decisions. We expect that as the number of training hands an agent has played rises, so will percentage of hands it wins.

This document is currently not available here.

Share

COinS
 
Apr 20th, 2:00 PM Apr 20th, 3:30 PM

Quantifying How Long Learning Takes: A Look at Machine Learning and Blackjack

CSU Ballroom

Machine learning is a form of artificial intelligence that allows a computer program to learn the most optimal way to act, instead of following explicitly programmed instructions. It is a form of pattern recognition that is being implemented and experimented with in many different fields and areas. The overall goal of this work is to find out how long it takes for machine learning to perform optimally in the simple game of black jack. We expect to distinguish the relationship between number of training hands played and how well the agent performs. To do this we are working to design a machine learning agent that has no knowledge of the game of blackjack. The agent is given X amount of training hands to play. These hands are randomly generated from a shoe of 6 decks and each deck is re-shuffled upon reaching the near end of the shoe. Each hand the agent is dealt, along with the dealer’s up card, is stored in a data structure that represents this pairing. The agent will then hit or stay randomly and the result of whether they lost or not is tracked. After these X training hands have been completed, the agent then plays a large number of new hands using the statistics of which option (hitting or staying) failed the least during the training hands to make its decisions. We expect that as the number of training hands an agent has played rises, so will percentage of hands it wins.

Recommended Citation

Lindquist, Dustin. "Quantifying How Long Learning Takes: A Look at Machine Learning and Blackjack." Undergraduate Research Symposium, Mankato, MN, April 20, 2015.
http://cornerstone.lib.mnsu.edu/urs/2015/poster_session_B/29