Learning from reaction databases
Johann Gasteiger, Percy Rose, U. Hondelmann, W. Witzenbichler, J. R. Rose · AIP conference proceedings · 1995
Chemists have derived knowledge of chemical reactions largely by inductive learning from individual observations. With the advent of large‐scale reaction databases machine learning techniques can be used to acquire knowledge of chemical reactions. Two model‐driven approaches to learn from data on the regioselectivity of the Diels‐Alder reaction and from data on the reactivity of amide hydrolysis are presented. Furthermore, a system that uses a data‐driven method for the classification and generalization of reactions is presented. This system can be used to derive knowledge for the prediction of reactions and the planning of organic synthesis.