Machine learning of generic reactions: 1. Scope of the project; the GRAMS program

This paper sets the bases of a new project for machine learning of generic reactions, starting from examples described in the literature. This project aims at resolving the difficulty of building large and coherent knowledge bases for Computer-Assisted Design of Organic Synthesis (CADOS) programs. Among the different steps involved in the generic reaction learning process, our present work studies the parametrization step, i.e., the derivation of generic schemes of reactions from specific examples of reactions. The foundations of the project are presented and the use of the GRAMS program for the generalization of reactions is illustrated.