Newsletter Signup
Where current and emerging technology trends meet.
TecTrendsInformation Sources, Inc.
  | About TecTrends | Email Signup | Contact Us
 Live Search:
Live Search | Articles | Companies | TecTerms | Products
  Loading TecTrends Live Search - please wait... 
View Noteworthy Articles      PRNewswire
 
Article

Title: Keep it Pithy: Parallel probabilistic model building...

Author: Weaver, Bill Article Type: Product Analysis
Source: Scientific Computing, v21 n11 p10(2) Publication Date: Oct 2004
  ISSN: 1524-2560
URL of Publication: http://www.scientificcomputing.com

An expert discusses the parallel probabilistic model building genetic algorithm. When research and development (R&D) centers have many people looking at multiple components of a large problem concurrently, management must coordinate investigations through analysis of intermediate results. Genetic algorithms also use that large-scale problem-solving concept when searching for a good solution to computationally complex problems. GAs, like other error space searching algorithms, propose candidates for the best solution, evaluate them to find the best solution, and generate more candidates in that region of minimum error space. Proposed solutions are fixed-length strings of variables and the whole collaboration of strings can be exemplified by a population of chromosomes. Sometimes error spaces are too large for a complete search for the true global minimum and instead offer best available solutions. A method that can be used to speed the process is use of a master processor to generate a large population and submit small groups of chromosomes to other processes. Fernando G. Lobo and his colleagues at the University of Algarve have shown a neat solution to bandwidth problems by replacing the traditional crossover and mutation GA with Probabilistic Model Building.

Products:
Genetic Algorithms Genetics

TecTerms:


[Get Copyright Permissions] Click here for copyright permissions!
Copyright 2004-2008 Information Sources Inc.
 


Home About TecTrends About Us Contact Us Privacy Statement Terms and Conditions

TecTrends | P.O. Box 8120 | Berkeley CA 94707 | (510) 525-6220 | Email: tectrends@tectrends.com
© 2006 INFORMATION SOURCES INC | All rights reserved.