Curve Fitting Using Genetic Algorithms
Abstract
Genetic algorithms are search techniques based on the mechanics of natural selection. They have been used successfully in many applications because of their robustness and because of their ability to search in a noisy problem space. In particular, genetic algorithms are used in curve-fitting. The genetic algorithm selects the coefficients of a particular curve that most closely matches a given set of data. Candidate solutions are vectors of real numbers that represent the coefficients of the curve to be modeled. Thus, every candidate solution corresponds to a new function. As such, each candidate solution is evaluated using the sum of the squares of the residuals. The evaluation of each of these curves with respect to its fit of the data guides the genetic algorithm toward the solution with the greatest merit. Several examples of the application of genetic algorithms to curve-fitting problems are presented. Convergence to the optimal solution is rapid when knowledge of the coefficients is available. When little is known about coefficients, a degree of experimentation help obtain the optimal solution.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 1991
- Accession Number
- ADA247206
Entities
People
- K. Messa
- M. Lybanon
Organizations
- United States Naval Research Laboratory