| Abstract |
Finding appropriate parameter values for evolutionary algorithms (EA) is one of the persisting grand challenges of the evolutionary computing (EC) field. On the one hand, all EC researchers and practitioners acknowledge that good parameter values are essential for good EA performance. On the other hand, even after 30 years of EC research there is very little known about the effect of EA parameters on EA performance. Users mostly rely on conventions (mutation rate should be low), ad hoc choices (why not use uniform crossover), and experimental comparisons on a limited scale (testing combinations of three different crossover rates and three different mutation rates). Hence, there is a striking gap between the widely acknowledged importance of good parameter values and the widely exhibited ignorance concerning principled approaches to tune EA parameters. In this talk we discuss the parameter issue from various angles, outline options for on-line control or off-line tuning of parameters and look in more details into the benefits algorithmic approaches to parameter tuning. |