Applying Hebbian Learning in System Control for Neuromorphic Computing

Kerssies, Jochem (2024) Applying Hebbian Learning in System Control for Neuromorphic Computing. Bachelor's Thesis, Artificial Intelligence.

Preview	Text Applying-Hebbian-Learning-in-System-Control-for-Neuromorphic-Computing.pdf Download (1MB) | Preview
	Text toestemming.pdf Restricted to Registered users only Download (400kB)

Abstract

This report explores Hebbian learning as an alternative to backpropagation. In previous research, Hebbian learning has been used often in combination with Hopfield networks. However, there is not a lot of research focused on the use of Hebbian learning for neuromorphic computing. In this report, an adaptation of the standard Hebbian learning rule called Oja's rule is used, which normalizes weight changes. Utilizing this approach, a single-layer perceptron is trained to play multiple adapted versions of the Chrome Dino game, testing the feasibility of Hebbian learning for simple control tasks. The network proved to be capable of learning some simplified versions of the game, but it did not succeed in learning the more complicated versions. Investigation of this resulted in a mathematical problem that is inherent to the design of the network and the way the state of the game is observed.

Item Type:	Thesis (Bachelor's Thesis)
Supervisor name:	Timmermans, J.J.M.A.
Degree programme:	Artificial Intelligence
Thesis type:	Bachelor's Thesis
Language:	English
Date Deposited:	24 Jul 2024 11:28
Last Modified:	24 Jul 2024 11:28
URI:	https://fse.studenttheses.ub.rug.nl/id/eprint/33672

Actions (login required)

View Item