Javascript must be enabled for the correct page display

How grounded defeasible rules and exceptions emerge: Conceptualizing the environment in a population of agents

Harders, W.H. (2010) How grounded defeasible rules and exceptions emerge: Conceptualizing the environment in a population of agents. Master's Thesis / Essay, Artificial Intelligence.

[img]
Preview
Text
ai-mai-2010-w.h.harders.pdf - Published Version

Download (6MB) | Preview

Abstract

We are able to use complex concepts that describe our environment. Exceptions are part of these concepts. Think for example about the platypus that we consider as a mammal but not as a bird, although the platypus has some of the typical properties of a bird. How are such exceptions created? In this work an agent model was developed to investigate how defeasible rules with exceptions can emerge in a population. The agents learn rules and exceptions about their environment by playing games. Using these rules a set of concepts is learnt that is grounded in the environment. The novelty of this model is the use of defeasible rules and exceptions. These rules have simple forms, with a conjunction of properties in the antecedent and a single concept as conclusion. Rules can be defeated under exceptional circumstances. In that case the conclusion of the rule cannot be reached anymore by using the defeated rule. A method was developed to iteratively learn and update such rules. Experiments with the model show that the agents are able to reach communicative success using the defeasible rules. Concepts that contain exceptions emerge in some occasions from the interactions.

Item Type: Thesis (Master's Thesis / Essay)
Degree programme: Artificial Intelligence
Thesis type: Master's Thesis / Essay
Language: English
Date Deposited: 15 Feb 2018 07:31
Last Modified: 15 Feb 2018 07:31
URI: http://fse.studenttheses.ub.rug.nl/id/eprint/9116

Actions (login required)

View Item View Item