Study of Genetic Algorithms Behavior for High Epitasis and High Dimensional Deceptive Functions

Optimization is a widespread notion: a large number of concrete problems require, during the solving process, optimizing a set of parameters (or variables) with respect to a given objective function (or fitness function). These variables can take integer or real values. Such problems for which variables can only take integer values, are called combinatorial optimization problems. This paper focuses on combinatorial optimization problems. The set of all possible combinations of values for the variables of the problem represents the search space of the problem. Constraint combinatorial optimization problems that is problems which define a set of constraints on the variables enabling a part of the search space are not considered here. Global combinatorial optimization problems for which the whole search space is available – are the main focus. In a large number of real problems, in particular in bioinformatics, the objective function is partially or entirely unknown. In this contest, it is however possible to calculate the value of the function in each point of the search space. This kind of problem is called a “black box” problem. Classical techniques of operations research are weakly or not at all fitted to black box problems. Consequently, evolutionist approaches of heuristic exploration strategies, have been developed specifically for black box problems. We are particularly interested in the complexity of such problems and in the efficiency of evolutionist methods to solve them.