Generator approach to evolutionary optimization of catalysts and its integration with surrogate modeling

This paper presents some unpublished aspects and ongoing developments of the recently elaborated generator approach to the evolutionary optimization of catalytic materials, the purpose of which is to obtain evolutionary algorithms precisely tailored to the problem being solved. It briefly recalls the principles of the approach, and then it describes how the employed evolutionary operations reflect the specificity of the involved mixed constrained optimization tasks, and how the approach tackles checking the feasibility of large polytope systems, frequently resulting from the optimization constraints. Finally, the paper discusses the integration of the approach with surrogate modeling, paying particular attention to surrogate models enhanced with boosting. The usefulness of surrogate modeling in general and of boosted surrogate models in particular is documented on a case study with data from a high-temperature synthesis of hydrocyanic acid.