Genetic operators

What Are Genetic Operators?

Genetic operators are the procedures that a genetic algorithm applies to a population of candidate solutions in order to produce the next generation. Three are standard: selection, which decides which individuals reproduce; crossover, which combines the encoded material of two or more parents; and mutation, which introduces random change into a single individual. Together they define the search dynamics of the algorithm, since the objective function only scores solutions while the operators are what actually move the population through the search space. The terminology comes from the biological analogy that John Holland used when he formalized genetic algorithms at the University of Michigan in the 1970s.

Operator behavior is inseparable from the encoding, or chromosome representation, on which it acts. Binary strings, real-valued vectors, permutations, trees in genetic programming, and variable-length structures each require operators that produce valid offspring, so a crossover that works on bit strings will generally destroy a permutation encoding a traveling salesman tour. Operator choice is also the main lever on the balance between exploration, meaning broad sampling of new regions, and exploitation, meaning refinement of what already scores well.

Selection

Selection assigns reproductive opportunity in proportion to fitness, and its strength is measured by selection pressure. Fitness-proportional selection, also called roulette wheel selection, gives each individual a share of the sampling probability equal to its share of total fitness, which is simple but sensitive to fitness scaling and prone to premature convergence when one individual dominates early. Rank selection replaces raw fitness with position in a sorted list, removing that sensitivity. Tournament selection draws a small random subset and keeps the best, and adjusting the tournament size tunes pressure directly, which is why it is the most common choice in practice. Elitism copies a fixed number of the best individuals unchanged into the next generation, guaranteeing that the best fitness found never decreases. Stochastic universal sampling reduces the variance of proportional selection by using evenly spaced pointers on a single spin.

Crossover

Crossover, also called recombination, is the operator that exchanges information between individuals, and it is applied with a probability typically set between 0.6 and 0.95. One-point and two-point crossover cut both parents at randomly chosen loci and swap the segments, while uniform crossover decides independently at each gene which parent contributes it. Real-coded algorithms use arithmetic blends, simulated binary crossover, and confidence-interval based schemes such as the CIXL2 crossover operator for evolutionary algorithms, which draws offspring from a statistical interval built around the best individuals. Permutation problems need order-preserving designs such as partially mapped, order, and cycle crossover. Comparative studies of new crossover designs for genetic algorithms show that no single operator dominates across problem classes, which motivates adaptive schemes that select among several operators during a run.

Mutation and Diversity

Mutation perturbs a single individual, usually with a low per-gene probability on the order of the reciprocal of chromosome length. Bit flip mutation inverts a binary gene, while real-coded algorithms use Gaussian, uniform, or polynomial perturbation with a step size that often shrinks as the run proceeds. Its role is to restore alleles that selection has removed from the population and to provide an escape route from local optima, which crossover alone cannot supply once diversity collapses. Related mechanisms that preserve diversity include fitness sharing, crowding, and restart strategies, and work on enhancing genetic algorithm performance reports that using several mutation and crossover operators together generally outperforms committing to one.

Applications

Genetic operators are applied wherever genetic algorithms are used, including: