Canada Spends Logo

Explorez notre analyse du budget 2026

Grant: Discovery Grants Program - Individual

University of Guelph

$15,000.00

Modeling evolutionary algorithms for dynamic and stochastic optimization

EF 2011-12

Résumé
The aim of this research is to study and model the behavior of evolutionary algorithms (EA) in stochastic and dynamic environments, with an eye to better understand and so better control evolutionary behavior under these environments. The models used in this research are based on statistically significant observations of production-level EA systems as opposed to a bottom-up analysis of the mathematical (probabilistic) properties of a very simplified EA. Furthermore they are not just performance (output) models of the EA but also comprise the properties that make up the state and dynamics of the EA within a complex environment. The project can be divided into two broad areas: population modeling for dynamic environments (DE), and statistical modeling for stochastic environments (SE). With population modeling, we first aim to properly analyze the fitness landscape (FL). Current FLs are based on the topology of the problem, not the topology of the search space. Furthermore, the crossover operator does not readily lend itself to a distance measure, which is needed to define a topology. Once a new FL is developed, a DE provides an interesting framework for utilizing the new model. With it we should be able to discern the evolutionary motion towards the global optimum and factor it out of the population motion, thus exposing the environmental motion. The EA could then be modified to be predictive, evolving to where the global optimum is heading, as opposed to being reactive, evolving to the optimum's current location. The second thrust of this research is the use of statistical modeling to capture and possibly control the behavior of an EA system in a SE, i.e. a probabilistic fitness function. Recently there has there been a renewed interest in this area mostly focused on the similarity with DEs or with the goal of finding ""robust"" solutions. However, there are more ramifications than are commonly realized, especially in the areas of multi-objective optimization and automated program induction using Genetic Programming (GP) especially when applied to the problem of symbolic regression.
Programme
Discovery Grants Program - Individual
Bénéficiaire
Emplacement
Ontario, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
NSERC's Awards Data

Tel que publié

Application id
238777-2010
Name nom
Wineberg, Mark
Department département
Computing and Information Science, School of
Organization id
24
Institution établissement
University of Guelph
Province en
Ontario
Province fr
Ontario
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2011
Competition year année de concours
2010
Award amount
15000
Program id
RGPIN
Program nane en
Discovery Grants Program - Individual
Program name fr
Programme de subventions à la découverte - individuelles
Group en
Discovery grants programs
Group fr
Programmes de subventions à la découverte
Committee code
1507
Committee name en
Computer Science
Committee name fr
Sciences informatiques
Area of application code
1207
Area of application group en
Advancement of knowledge
Area of application group fr
Avancement des connaissances
Area of application en
Information, computer and communication technologies
Area of application fr
Technologie de l'information, des ordinateurs et des réseaux de communications
Research subject code
2800
Research subject group en
Artificial intelligence
Research subject group fr
Intelligence artificielle
Research subject en
Artificial intelligence (computer vision, use2603)
Research subject fr
Intelligence artificielle (vision artificielle, utiliser 2603)
Partie
2
Nb partie
5
Application title
Modeling evolutionary algorithms for dynamic and stochastic optimization
Keyword
NSERC
Application summary
The aim of this research is to study and model the behavior of evolutionary algorithms (EA) in stochastic and dynamic environments, with an eye to better understand and so better control evolutionary behavior under these environments. The models used in this research are based on statistically significant observations of production-level EA systems as opposed to a bottom-up analysis of the mathematical (probabilistic) properties of a very simplified EA. Furthermore they are not just performance (output) models of the EA but also comprise the properties that make up the state and dynamics of the EA within a complex environment. The project can be divided into two broad areas: population modeling for dynamic environments (DE), and statistical modeling for stochastic environments (SE). With population modeling, we first aim to properly analyze the fitness landscape (FL). Current FLs are based on the topology of the problem, not the topology of the search space. Furthermore, the crossover operator does not readily lend itself to a distance measure, which is needed to define a topology. Once a new FL is developed, a DE provides an interesting framework for utilizing the new model. With it we should be able to discern the evolutionary motion towards the global optimum and factor it out of the population motion, thus exposing the environmental motion. The EA could then be modified to be predictive, evolving to where the global optimum is heading, as opposed to being reactive, evolving to the optimum's current location. The second thrust of this research is the use of statistical modeling to capture and possibly control the behavior of an EA system in a SE, i.e. a probabilistic fitness function. Recently there has there been a renewed interest in this area mostly focused on the similarity with DEs or with the goal of finding ""robust"" solutions. However, there are more ramifications than are commonly realized, especially in the areas of multi-objective optimization and automated program induction using Genetic Programming (GP) especially when applied to the problem of symbolic regression.