On a property analysis of representations for spanning tree problems

This paper investigates on some properties of encodings of evolutionary algorithms for spanning tree based problems. Although debate continues on how and why evolutionary algorithms work, many researchers have observed that an EA is likely to perform well when its encoding and operators exhibit locality, heritability and diversity. To analyze these properties of various encodings, we use two kinds of analytical methods; static analysis and dynamic analysis and use the Optimum Communication Spanning Tree (OCST) problem as a test problem. We show it through these analysis that the encoding with extremely high locality and heritability may lose the diversity in population. And we show that EA using Edge Window Decoder (EWD) has high locality and high heritability but nevertheless it preserves high diversity for generations.