A Kernighan-Lin Local Improvement Heuristic That Solves Some Hard Problems in Genetic Algorithms

We present a Kernighan-Lin style local improvement heuristic for genetic algorithms. We analyze the run-time cost of the heuristic. We demonstrate through experiments that the heuristic provides very quick solutions to several problems which have been touted in the literature as especially hard ones for genetic algorithms, such as hierarchical deceptive problems. We suggest why the heuristic works well. In this research, population members (chromosomes) are bit strings, all of the same length, which we denote by N. We will refer to population members as individuals. A local improvement heuristic is a procedure which is applied to an individual, with the intention of modifying it into an related individual of higher fitness. Typically it is applied to a child chromosome, after its manufacture by some crossover operator but before the child is entered into the population. One form of local improvement heuristic is hill-climbing. One step of hill-climbing consists of flipping that bit in ...

• Type: Conference
• Year: 2003
• Where: GECCO
• Authors: William A. Greene
• Keywords: GECCO 2003, Heuristic, Individual, Local Improvement