Signal, noise, and genetic algorithms
Mike Rudnick · 2018
rganized into competition partitions --- competing partial solutions, or sets of schemata fixing the same bit positions. Thus, every complete solution, such as a member of a GA's population, 1 belongs to exactly one schema within each competition partition. In effect, the partial solutions within a competition partition compete for representation in the GA's population of complete solutions. Each schema has as its schema fitness, f(h), the average fitness of its elements. The partial solutions within each partition will, in general, have a spread in their fitnesses, which can be thought of as the partition's selection pressure, or convergence signal. The greater the fitness spread among the partition's schemata (partial solutions), the greater the partition's signal strength, and the greater will be the convergence occurring within the partition. It is this spread in fitness which enables the GA to select for the better partial solution