Javascript must be enabled for the correct page display

Learning to play Connect-Four using Evolutionary Neural Networks

Box, R.H. (2015) Learning to play Connect-Four using Evolutionary Neural Networks. Bachelor's Thesis, Artificial Intelligence.

[img]
Preview
Text
AI-BA-2015-RoburBox.pdf - Published Version

Download (399kB) | Preview
[img] Text
toestemming.pdf - Other
Restricted to Backend only

Download (24kB)

Abstract

This thesis describes the machine learning technique Evolutionary Neural Networks applied to learning to play connect-four. It introduces the concept of Evolution and Genetic Algorithms, and explains the process of optimizing neural networks. Genetic Algorithms can be used in many different ways, so the design decisions and parameter optimization plays a large role. A total of 941 hours of (single core) processing time was dedicated to evolving neural networks with various different evolution settings. These settings include, amongst others; population size, mutation rate, crossovers, etc. The thesis concludes that Evolutionary Neural Networks are suited to learn to play connect-four, compared to earlier research of Neural-Fitted Temporal Difference Learning.

Item Type: Thesis (Bachelor's Thesis)
Degree programme: Artificial Intelligence
Thesis type: Bachelor's Thesis
Language: English
Date Deposited: 15 Feb 2018 08:03
Last Modified: 15 Feb 2018 08:03
URI: http://fse.studenttheses.ub.rug.nl/id/eprint/12645

Actions (login required)

View Item View Item