Models of steady state evolutionary algorithms

Document Type

Presentation Abstract

Presentation Date

3-16-2000

Abstract

Evolutionary computation algorithms are based on the simulated evolution of a population of potential solutions to a problem. They have been applied to a very wide variety of practical problems. The behavior of these algorithms can be modeled both as dynamical systems and as Markov chains.

These algorithms can be categorized into generational, where the entire population is replaced at each time step, and steady state, where only a single individual is replaced per time step. Previous dynamical system models have been discrete-time models of generational evolutionary algorithms. I will describe how a continuous-time dynamical system model of a steady state evolutionary algorithm can be derived by letting the population size go to infinity while the time step goes to zero. It turns out that the discrete-time system is the Euler approximation to the continuous-time system. Then it is shown how the continuous-time system might be stable while the discrete-time system is unstable.

This talk represents joint work with Jonathon Rowe, De Montfort University, UK.

Additional Details

Thursday, 16 March 2000
4:10 p.m. in Math 109
Coffee/treats at 3:30 p.m. Math 104 (lounge)

This document is currently not available here.

Share

COinS