Title :
Methods for decreasing the number of objective evaluations for independent computationally expensive objective problems
Author_Institution :
Georgia Tech Res. Inst., Atlanta, GA
Abstract :
In this paper, three new methods for pushing solutions toward a desired region of the objective space more quickly are explored; hypercube distance scaling, dynamic objective thresholding, and hypercube distance objective ordering. These methods are applicable for problems that do not require the entire Pareto front and that require an independent computationally expensive computation for each objective. The performance of these methods is evaluated with the multiple objective 0/1 knapsack problem.
Keywords :
Pareto optimisation; evolutionary computation; geometry; knapsack problems; Pareto front; dynamic objective thresholding; hypercube distance objective ordering; hypercube distance scaling; independent computationally expensive objective problems; knapsack problem; objective evaluations; Evolutionary computation; Hafnium; Hypercubes; Multidimensional systems; Optimization methods; Pareto optimization;
Conference_Titel :
Evolutionary Computation, 2008. CEC 2008. (IEEE World Congress on Computational Intelligence). IEEE Congress on
Conference_Location :
Hong Kong
Print_ISBN :
978-1-4244-1822-0
Electronic_ISBN :
978-1-4244-1823-7
DOI :
10.1109/CEC.2008.4631245