Efficient-Discarding Fitness Functions

In the last 30 years, a lot of effort has been dedicated to develop robust optimization methods like Evolutionary Algorithms, Simulated Annealing, and recently the so-called Estimation Distribution Algorithms. All these algorithms have to evaluate an objective functions many times. This situation leads us to the problem of reducing the cost and number of function evaluations without affecting the performance of the methods. This paper investigates some conditions that allow the definition of an alternative fitness function called Eff icient-Discarding Fitness Function that substitutes the original function in a way that reduces the computational cost of each evaluation without affecting the quali ty of the search. This idea will permit to establish some paradigms leading to low-cost general optimization algorithms.