The Balance Beam Function: A New Test Function for the Real World Problems

Abstract

A new test function called the Balance Beam Function (BBF) is introduced. This function is designed for understanding the performance of evolutionary algorithms on real world problems in which many “neutral networks” are assumed to be found. The concept of a neutral network originates from genotype-phenotype models of RNA secondary structure and protein structure. It refers to long, narrow, meandering, yet flat, pathways through a fitness landscape which may lead to better points in a search space. The objective of the proposed test function is to provide a measure of the speed at which an evolving population moves along a neutral network; this is an important characteristic of an evolutionary algorithm, especially when solving largescale problems. In empirical experiments, two kinds of GA, the standard GA and an extended GA-called the operon-GA-are applied to BBFs in order to investigate their evolutionary behavior and the ease with which their parameters can be tuned.