Comparative study of the performance of different bio-inspired algorithms using the same cost function

Resumo

Evolutionary computing is a computational intelligence tool widely used for data analysis and cost

functions. In addition to being accessible in different languages, there are different types of techniques and con-
cepts that can be used for different types of functions. One of these techniques is the bio-inspired algorithms,

which have this name due to the fact that they are inspired by phenomena occurring in nature to search for results.
One of the situations where this type of algorithm is generally used is in minimizing or maximizing a value of
a cost function. There are several types and models of bio-inspired algorithms in the literature, each inspired by
a phenomenon and which works best on a certain type of problem. This work aims to explore different types
of these algorithms and thus perform a comparative study of which has the best performance for a function-cost
minimization problem, here represented by the function called Matyas, which has two dimensions and has no local

minimums , only the global. For this, 7 different types of algorithms were designed, which are: Simple Differen-
tial Evolution, Differential Evolution with Variant, Differential Evolution with Opposition, Simple PSO, PSO with

Constriction Factor, PSO with Inertia Factor and CLONALG. The results showed that, for the chosen optimization
cost-function, as the objective was to minimize the value of the variable, the Differential Evolution algorithms
obtained more satisfactory results in relation to the PSO and CLONALG algorithms, especially with opposition
learning.