Parametric Sensitivity and Search-Space Characterization Studies of Genetic Algorithms for Computer-Aided Polymer Design
Anantha Sundaram, Venkat Venkatasubramanian · Journal of Chemical Information and Computer Sciences · 1998
Genetic Algorithms (GAs) and evolutionary methods have been demonstrated to be flexible and efficient optimization techniques with potential for locating global optima under general conditions for computer-aided molecular design (CAMD). However, they often need customization requiring detailed study of parametric sensitivity and search-space character before optimal internal parameters are determined. This paper describes such a parametric sensitivity study for GA-based polymer design. The objective of the study was to study the influence of the internal parameters of the GA on its performance on a large scale polymer design problem. The study yielded qualitative trends that clearly identified the structure of the target sought or the nature of the search-space to be the key factor determining the GA's performance. The study was then focused toward studying the structure to fitness correlation in the underlying search-space. The results of this study indicate that the performance of the GA could be enhanced by diversified sampling schemes, adaptive parameter tuning, and interactive inclusion of additional design knowledge.