Do multiple trials help Univariate methods?

Cooperative Coevolutionary Algorithms (CCEAs) and Univariate Estimation of Distribution Algorithms (Univariate ED As) are closely related algorithms in that both update marginal distributions/populations, and test samples of those distributions/populations by grouping them with collaborators drawn from elsewhere to form a complete solution. Thus the quality of these samples is context-sensitive and the algorithms assume low linkage among their variables. This results in well-known difficulties with these methods. While EDAs have commonly overcome these difficulties by examining multivariate linkage, CCEAs have instead examined basing the fitness of each marginal sample on the maximum of several trials. In this study we examine whether multiple-trial CCEA approach is really effective for difficult problems and large numbers of subpopulations; and whether this approach can be used to improve Univariate EDAs as well.