In the last twenty years, the design of efficient function optimizers has been a crucial topic of research work. Many theoretical and practical research problems involve combinatorial optimization, which is obtaining the optimal solution among a finite set of alternatives. Such optimization problems are notoriously difficult to solve, one of the primary reasons being that in most applications the number of alternatives is extremely large and only a fraction of them can be considered within a reasonable amount of time (Shi et al, 1999). Indeed, such problems are often non-differentiable or multimodal, causing difficulties for classical gradient and random search methods, which are limited to simple unimodal functions and which are thus inappropriate to apply to such difficult problems. The seminal methods to deal with this problem are called Deterministic Optimization Methods (DOMs) (Fletcher, 1980). There are two major drawbacks when using DOMs: their high computational complexity and the possibility of becoming trapped in a local optimum. Moreover, DOMs are difficult to program, as they require a deep mathematical knowledge.