# The Black Art of Evolutionary Algorithm

Before we go any further, the first thing that we have to do is to define what’s black art itself. According to Oxford Dictionary, black art means a technique or practice considered mysterious and sinister. In the other word, the approaches are questioned how an outcome can be produced using the algorithm that its parameter is defined through trial and error process which is mysterious.

The Evolutionary Algorithm (EA) is a stochastic approach optimization which using a random number generator in the search of solutions. As a consequence, an algorithm in this class will produce different final solution every run the same problem. There is some type of EA, Genetic Algorithm, Evolution Strategies, Genetic Programming, and Evolutionary Programming.

Genetic Algorithm (GA) is one of the popular types of EA. The GA process consists of initialization, reproduction, evaluation, and selection. Each process has parameters to look for the best setting for a certain problem. In the initialization phase, we have to define the population size and generation size. In the reproduction phase, the crossover and mutation method defined and also the crossover and mutation rate. In the evaluation phase, the fitness function is determined. In the selection phase, the selection method is selected.

The determination of the appropriate setting and the right values of each parameter for each process based on the problem itself requires board experience. Some questions to determine the appropriate setting arise. Should we use high or low mutation rate and crossover rate? What generation size and population size should we use? What selection method should we use? What the suitable fitness function for the problem? In order to answer those questions, it particularly requires experimentation (trial and error) process which some people called this ‘black art’.

However, in real life, we know an extraordinary arrangement about sufficient values for the vast majority of the parameters of a basic GA. For instance, a basic variable for the achievement of a basic GA is the population size, and there is a theory that related to the problem length and difficulty of some class of functions to the population size.

Moreover, a lot of research has been provided with some solution to determine the appropriate setting. One of them is the determination of crossover rate and mutation rate. Some research applies adaptive crossover and mutation rate mechanism. While another research determines the best crossover rate and best mutation rate between certain ranges based on their experiment.

In conclusion, the questions about the appropriate setting which requires experimentation process (black art) are can be diminished by applying the theory that obtained by another researcher.

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.