In this study two catalyst library optimization methods, the Holographic Research Strategy (HRS) and the Genetic Algorithm (GA) were compared based on their ability to find the optimum compositions in a given multi-dimensional experimental space. Results obtained in three different case studies were used to investigate both the rate and the certainty of the optimum search. In these case studies the activity-composition relationships were established using Artificial Neural Networks (ANNs) trained with catalytic data published earlier. The above relationships were used in "virtual optimization experiments" using both HRS and GA for catalyst library optimization. Upon using the stochastic GA its exceedingly divers mode of sampling often resulted in poor catalytic materials in the next catalyst generation. This fact resulted in a decreased rate of convergence to the optimum. In contrast, in HRS, which is a deterministic optimization algorithm, a moderate level of diversity in the catalyst library can easily be achieved. In this way an acceptable rate in optimum search can be accomplished. The visualization ability of HRS allows the illustration of all virtually tested compositions in a two-dimensional form regardless the optimization algorithm used. Upon using HRS a structured arrangement of experimental points in the virtual holograms was observed. However, when GA was applied for virtual optimization "starry sky"-like arrangement of compositions in the virtual holograms was obtained. Therefore based on virtual holograms, upon using HRS the relationship between the composition of catalytic materials and their performance can be qualitatively revealed, while no similar correlation can be obtained using GA. (c) 2006 Elsevier B.V. All rights reserved.