Formal Memetic Algorithms

Abstract. A formal, representation-independent form of a memetic algorithm-a genetic algorithm incorporating local search--is introduced. A generalisedform of N-point crossover is defined together with representation-independent patching and hill-climbing operators. The resulting formal algorithm is then constructed and tested empirically on the travelling sales-rep problem. Whereas the genetic algorithms tested were unable to make good progress on the problems studied, the memetic algorithms performed very well. 1 Motivation The r^ole of local search in the context of genetic algorithms and the wider field of evolutionary computing has been much discussed. The traditionalview, which can be traced back to Holland (1975), has been that the primary search operator in evolutionary computing should be recombination. In its most extreme form, this view casts mutation and other local operators as mere adjuncts to recombination, playing auxiliary (if important) r^oles such as keeping the gene...

• Type: Conference
• Year: 1994
• Where: AISB
• Authors: Nicholas J. Radcliffe, Patrick D. Surry
• Keywords: AISB 1994, Algorithm-a Genetic Algorithm, Artificial Intelligence, Evolutionary Computing, Genetic Algorithm