Overview

It is typical for real-world genetic algorithms to define several types of genetic operators, whether unary, binary or multi-ary. In general, the rates of the operators do not have optimal values which are valid for an entire run of the genetic algorithm. More often than not, a strategy for optimally setting the rates is hard to find by the designer of the genetic algorithm. On the other hand, each type of operator has parameters which control the algorithm it defines. Finding the appropriate values for these parameters is a difficult task, as well.

IAGA is a general schema of a genetic algorithm which is able to dynamically adapt both the rate and the behavior for each of its operators. The rates of the operators are adapted in a deterministic, reinforcement-based manner. The behavior of each operator (that is, the specific way it operates) is modified by changing its parameter values.

Operators that prove themselves valuable during the optimization process are rewarded; thus, they are applied more often in subsequent generations. The behavior of each operators evolves as well during the run of the genetic algorithm, by modifying the values of the parameters which control its activity.