Title :
Predictive measures for problem difficulty
Author :
Reeves, Colin R.
Author_Institution :
Sch. of Math. & Inf. Sci., Coventry Univ., UK
Abstract :
In practical applications to instances of optimization problems it would be of great benefit if we could decide a priori which algorithm is suited to the problem; or at least, which of several candidates might be better than others. In some cases it is obvious: optimizing a linear function of variables subject to linear constraints for instance. However, such examples are rare, and this is certainly the case for potential applications of evolutionary algorithms (EAs). For this reason, various predictive measures have been suggested. These all sample the Universe of all potential solutions in order to compute some statistics that are thought to aid the decision. This paper describes some of these ideas and then discusses and evaluates them using an adversary argument
Keywords :
algorithm theory; computability; genetic algorithms; epistasis; evolutionary algorithms; genetic algorithms; optimization problems; predictive measures; problem difficulty; Biological cells; Constraint optimization; Design for experiments; Evolutionary computation; Genetic algorithms; History; Prediction algorithms; Search methods;
Conference_Titel :
Evolutionary Computation, 1999. CEC 99. Proceedings of the 1999 Congress on
Conference_Location :
Washington, DC
Print_ISBN :
0-7803-5536-9
DOI :
10.1109/CEC.1999.782006
