We investigate the search properties of pre-evolutionary random catalytic reaction networks, where reactions might be reversible, and replication is not taken for granted. Since it counts only on slow growth rates and weak selective pressure to steer the search process, catalytic search is an inherently slow process. However it presents interesting properties worth exploring, such as the potential to steer the computation flow towards good solutions, and to prevent premature convergence. We have designed a simple catalytic search algorithm, in order to assess its beamed search ability. In this paper we report preliminary results that show that although weak, the search strength achieved with catalytic search is sufficient to solve simple problems, and to find good approximations for more complex problems, while keeping a diversity of solutions and their building blocks in the population. © 2010 Springer-Verlag Berlin Heidelberg.
CITATION STYLE
Yamamoto, L. (2010). Evaluation of a catalytic search algorithm. In Studies in Computational Intelligence (Vol. 284, pp. 75–87). https://doi.org/10.1007/978-3-642-12538-6_7
Mendeley helps you to discover research relevant for your work.