Multialgorithms for parallel computing: A new paradigm for optimization

Fixed-population and evolving-population multialgorithms are introduced. They constitute a new population-based paradigm for optimization at the confluence of traditional optimization techniques, evolutionary algorithms and parallel computing. A detailed illustration is given within the context of a new two-parameter family of nonlinear conjugate gradient algorithms and a set of four standard test problems. The illustration provides a platform for a discussion of several major themes that arise in the study of multialgorithms in general.