Effects on Clustering Quality of Direct and Indirect Communication Among Agent in Ant-based Clustering Algorithms

Hdl Handle:
http://hdl.handle.net/11285/572397
Title:
Effects on Clustering Quality of Direct and Indirect Communication Among Agent in Ant-based Clustering Algorithms
Issue Date:
01/05/2005
Abstract:
Ant-based clustering algorithms are knowledge discovery tools inspired by the collective behavior of social insect colonies. In these algorithms, insects are modeled as software agents that communicate with each other indirectly through the environment. This particular kind of communication is known as stigmergic communication. In the classic ant-based clustering algorithm, a group of agents that exhibit the same behavior move randomly over a toroidal square grid. In the environment there are data objects that were initially scattered in a random fashion. The objects can be picked up, moved or dropped in any free location on the grid. An object is picked up with high probability if it is not surrounded by similar objects and is dropped with high probability if an agent's neighborhood is densely populated by other similar objects and its location is free. Here, stigmergy occurs when an object is placed next to another. The resultant structure is much more attractive to agents to drop other similar objects nearby. However, stigmergy is not the only way social insects interact with each other. In most species, trophallaxis or liquid food exchange among members of the same colony, plays a key role in their social organization. Consider the case of some termite species which require intestinal protozoa to derive benefits from cellulose. Their early instar nymphs are fed either by oral or anal trophallaxis. The latter infects them with symbiotic protozoa or bacteria contained in the proctodeal liquid. The subsocial association result of this codependence have evolved into a complex social and morphological structure. Inspired by the trophallaxis phenomenon observed in some ant and termite species, two different communication strategies among agents in ant-based clustering algorithms are investigated: (i) direct and (ii) indirect communication. The impact on the final clustering quality is evaluated by comparing the development of the clustering process generated by each strategy. It is shown that benefits on the final clustering are directly related to the usefulness of the exchanged information, its use, and on the number of participating agents.
Keywords:
Clustering Quality; Communications; Electronics; Clustering Algorithms; Computer Science
Advisors:
Dr. Leonardo Garrido Luna
Committee Member / Sinodal:
Dr. José Luis Aguirre Cervantes; Dr. Ramón Felipe Brena Pinero
Degree Level:
Maestro en Ciencias en Sistemas Inteligentes
School:
Electrónica, Computación, Información y Comunicaciones
Campus Program:
Campus Monterrey
Discipline:
Ingeniería y Ciencias Aplicadas / Engineering & Applied Sciences
Appears in Collections:
Ciencias Exactas

Full metadata record

DC FieldValue Language
dc.contributor.advisorDr. Leonardo Garrido Lunaes
dc.creatorMontes de Oca Roldán, Marco A.en
dc.date.accessioned2015-08-17T11:27:54Zen
dc.date.available2015-08-17T11:27:54Zen
dc.date.issued01/05/2005-
dc.identifier.urihttp://hdl.handle.net/11285/572397en
dc.description.abstractAnt-based clustering algorithms are knowledge discovery tools inspired by the collective behavior of social insect colonies. In these algorithms, insects are modeled as software agents that communicate with each other indirectly through the environment. This particular kind of communication is known as stigmergic communication. In the classic ant-based clustering algorithm, a group of agents that exhibit the same behavior move randomly over a toroidal square grid. In the environment there are data objects that were initially scattered in a random fashion. The objects can be picked up, moved or dropped in any free location on the grid. An object is picked up with high probability if it is not surrounded by similar objects and is dropped with high probability if an agent's neighborhood is densely populated by other similar objects and its location is free. Here, stigmergy occurs when an object is placed next to another. The resultant structure is much more attractive to agents to drop other similar objects nearby. However, stigmergy is not the only way social insects interact with each other. In most species, trophallaxis or liquid food exchange among members of the same colony, plays a key role in their social organization. Consider the case of some termite species which require intestinal protozoa to derive benefits from cellulose. Their early instar nymphs are fed either by oral or anal trophallaxis. The latter infects them with symbiotic protozoa or bacteria contained in the proctodeal liquid. The subsocial association result of this codependence have evolved into a complex social and morphological structure. Inspired by the trophallaxis phenomenon observed in some ant and termite species, two different communication strategies among agents in ant-based clustering algorithms are investigated: (i) direct and (ii) indirect communication. The impact on the final clustering quality is evaluated by comparing the development of the clustering process generated by each strategy. It is shown that benefits on the final clustering are directly related to the usefulness of the exchanged information, its use, and on the number of participating agents.es
dc.language.isoen-
dc.rightsOpen Accessen
dc.rights.urihttp://creativecommons.org/licenses/by-nc-nd/4.0/*
dc.titleEffects on Clustering Quality of Direct and Indirect Communication Among Agent in Ant-based Clustering Algorithmsen
dc.typeTesis de Maestríaes
thesis.degree.grantorInstituto Tecnológico y de Estudios Superiores de Monterreyes
thesis.degree.levelMaestro en Ciencias en Sistemas Inteligenteses
dc.contributor.committeememberDr. José Luis Aguirre Cervanteses
dc.contributor.committeememberDr. Ramón Felipe Brena Pineroes
thesis.degree.disciplineElectrónica, Computación, Información y Comunicacioneses
dc.subject.keywordClustering Qualityes
dc.subject.keywordCommunicationses
dc.subject.keywordElectronicses
dc.subject.keywordClustering Algorithmses
dc.subject.keywordComputer Sciencees
thesis.degree.programCampus Monterreyes
dc.subject.disciplineIngeniería y Ciencias Aplicadas / Engineering & Applied Sciencesen
All Items in REPOSITORIO DEL TECNOLOGICO DE MONTERREY are protected by copyright, with all rights reserved, unless otherwise indicated.