An Experimental Investigation of Self-Adaptation in Evolutionary Programming

摘要: Evolutionary programming (EP) has been widely used in numerical optimization recent years. One of EP's key features is its self-adaptation scheme. In EP, mutation typically the only operator to generate new offspring. The often implemented by adding a random number from certain distribution (e.g., Gaussian case classical EP) parent. An important parameter standard deviation (or equivalently variance). scheme this evolved, rather than manually fixed, along with objective variables. This paper investigates empirically how well works on set benchmark functions. Some anomalies have observed empirical studies, which demonstrate that may not work as hoped for some experimental evaluation an existing simple fix problem also carried out paper.