On composing an (evolutionary) algorithm portfolio

In this paper, we propose a general methodology to automatically compose a good portfolio from a set of selected EAs. As a single EA is a degenerate portfolio, our method also provides an answer to when a portfolio of two or more EAs are beneficial. Our method has the nice property of being parameter-less; it does not introduce extra parameters. Hence there is no need for parameter control, which is well known to be a thorny research issue. To illustrate our idea, we show how a portfolio that is constructed by considering five state of the art EAs as candidates is automatically constructed from ten CEC 2005 benchmark functions. It is found that the resulting portfolio enjoys excellent, and equally importantly, stable ranking. Thus the new portfolio algorithm has the property of being a robust algorithm, which is a highly desirable property in practical applications.