Paled - Artificial intelligence in poker as simulation for real life processes
Rapp, Stefan (2010)
2010
All rights reserved. This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-202401231713
https://urn.fi/URN:NBN:fi:amk-202401231713
Tiivistelmä
Development of the program PALED, an artificial intelligence (AI) based software for simulation of stochastic processes, is presented in the thesis. Poker as the model of stochastic process and case-base reasoning (CBR) as the AI tool where used in the work.
Case-based reasoning (CBR) is driven by to motivation to imitate human behavior and developed technology for AI systems. It is the method to learn from previous problems. PALED uses a case-base of 15, 000 played hands and the K-nearest neighbor algorithm to find the closest match cases. Based on the solution and outcome of the closest match cases PALED evaluates a solution for the current hand.
The first version of PALED used CBR AI technology to solve the problem of poker. After 10 000 played hands against other AI PALED lose money but the play analysis showed that PALED was able to play poker well.
The second version of PALED was extended with expert knowledge based on the weaknesses of the first version. The testing phase against other AI showed that PALED won money over 8, 000 played hands. This version was also tested against human players where it won money on lower limits such as 0.05/0.10 $. On higher limits such as 0.25/0.50 it lose money because the opponents started to exploit its weaknesses.
In conclusion, CBR built a strong decision framework which can be extended easily with other AI technology.
Case-based reasoning (CBR) is driven by to motivation to imitate human behavior and developed technology for AI systems. It is the method to learn from previous problems. PALED uses a case-base of 15, 000 played hands and the K-nearest neighbor algorithm to find the closest match cases. Based on the solution and outcome of the closest match cases PALED evaluates a solution for the current hand.
The first version of PALED used CBR AI technology to solve the problem of poker. After 10 000 played hands against other AI PALED lose money but the play analysis showed that PALED was able to play poker well.
The second version of PALED was extended with expert knowledge based on the weaknesses of the first version. The testing phase against other AI showed that PALED won money over 8, 000 played hands. This version was also tested against human players where it won money on lower limits such as 0.05/0.10 $. On higher limits such as 0.25/0.50 it lose money because the opponents started to exploit its weaknesses.
In conclusion, CBR built a strong decision framework which can be extended easily with other AI technology.