The effect of dependencies in data on the accelerated argumentation-based learning algorithm

Scholz, Jonas (2023) The effect of dependencies in data on the accelerated argumentation-based learning algorithm. Bachelor's Thesis, Artificial Intelligence.

Preview	Text bAI_2023_ScholzJ.pdf Download (509kB) | Preview
	Text toestemming.pdf Restricted to Registered users only Download (127kB)

Abstract

With the increased usage of artificial intelligence(AI), there is an increase in the misuse of AI, as users do not understand how AI comes to its decisions. This means that there is a need for an AI that is explainable, meaning that humans can understand how the AI came to its decision. There are some explainable AI systems that are able to both learn as well as stay human-understandable, one such system is the accelerated argumentation-based learning(AABL) algorithm. This is a new learning algorithm that uses argumentation to determine the correct output. As the AABL algorithm is still new the limits of this algorithm are still unknown, as this algorithm seems promising it is important to determine where the limits lie. In this regard, this paper will explore the effects of dependencies in the data on the accuracy of the model in comparison to a decision tree and neural network. The data shows that the AABL algorithm outperforms the other machine learning algorithms eventually for higher dependencies, though this trend does not seem to continue for 21 dependencies. Furthermore, the AABL algorithm slows down considerably for higher dependencies compared to the other two algorithms. This, therefore, highlights that while this approach towards explainable AI works also for other scenarios compared to the original scenario it was tested on, it also shows that the algorithm still has some limitations that need to be addressed by derived algorithms.

Item Type:	Thesis (Bachelor's Thesis)
Supervisor name:	Verheij, H.B.
Degree programme:	Artificial Intelligence
Thesis type:	Bachelor's Thesis
Language:	English
Date Deposited:	12 Jul 2023 12:04
Last Modified:	12 Jul 2023 12:04
URI:	https://fse.studenttheses.ub.rug.nl/id/eprint/30581

Actions (login required)

View Item