Survivor selection is the policy that determines the composition of the next generation’s population. It decides which individuals—from the combined pool of current parents and newly created offspring—will be kept and which will be discarded. This process has the dual goals of preserving high-quality solutions found so far while continuing to maintain enough diversity to fuel future exploration.

The Role of Elitism

Many Genetic Algorithms employ a strategy known as Elitism. In its simplest form, elitism guarantees that the single fittest individual from the current generation is carried over, unchanged, into the next generation. This ensures that the GA’s best-found solution is never accidentally lost due to the stochastic nature of selection, crossover, or mutation.

Survivor Selection Policies

Two primary survivor selection policies are widely used:

9.1 Age-Based Selection

In this policy, survival is not based on fitness. Instead, it is based on the premise that each individual is allowed to remain in the population for a finite number of generations. After reaching its “age limit,” the individual is removed, regardless of how fit it is. For example, if the oldest members of a population are P4 and P7, they would be removed to make room for new offspring, while the ages of all other individuals would be incremented by one.

9.2 Fitness-Based Selection

This is a more common approach where the newly generated children replace the least fit individuals in the parent population. The selection of which individuals to remove can be done using various methods, such as a reverse tournament selection (where the loser is picked for removal). For instance, if new children are created, they would replace the two individuals with the lowest fitness scores in the current population, such as P1 and P10.

With the mechanisms for creating and shaping a new generation now defined, we turn to the final question in the GA cycle: when to stop this evolutionary process.