Copyright © 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Cover Picture: ChemPhysChem 11/2002
The cover picture shows how a neural network (center top) simplifies the search for useful heterogeneous catalysts, in this case composed of the chemical elements. The initial data is composed of known combinations of the elements (the ”world” entry). After initial training, the neural network can mine this data and predict worthy combinations of the elements, which can be tested experimentally on a small scale. A generalized evolutionary optimization strategy powerfully selects promising candidates from the experimental results, and this data improves the neural network for the next generation of optimization. These strategies tolerate noisy data and work well within high-dimensional space; conditions appropriate for optimizing catalysts. Find out more in the article by Corma et al. on pages 939–945.