Supervised and unsupervised discretization methods for evolutionary algorithms
This paper introduces simple model-building evolutionary algorithms (EAs) that operate on continuous domains. The algorithms are based on supervised and unsupervised discretization methods that have been used as preprocessing steps in machine learning. The basic idea is to discretize the continuous variables and use the discretization as a simple model of the solutions under consideration. The model is then used to generate new solutions directly, instead of using the usual operators based on sexual recombination and mutation. The algorithms presented here have fewer parameters than traditional and other model-building EAs. They expect that the proposed algorithms that use multivariate models scale up better to the dimensionality of the problem than existing EAs.
- Research Organization:
- Lawrence Livermore National Lab. (LLNL), Livermore, CA (United States)
- Sponsoring Organization:
- US Department of Energy (US)
- DOE Contract Number:
- W-7405-ENG-48
- OSTI ID:
- 15005541
- Report Number(s):
- UCRL-JC-142243; TRN: US200323%%315
- Resource Relation:
- Conference: Genetic and Evolutionary Computation Conference 2001, San Francisco, CA (US), 07/07/2001--07/11/2001; Other Information: PBD: 24 Jan 2001
- Country of Publication:
- United States
- Language:
- English
Similar Records
Grain boundary structure search by using an evolutionary algorithm with effective mutation methods
Evolutionary pattern search algorithms