Improvement of the Descriptor

The initial model is altered in several ways to adapt the resulting RDF descriptor to that obtained from the CPG neural network. For this task a program was developed that contains empirical rules for the optimization of an RDF descriptor. The program searches for the molecule having an RDF descriptor most similar to the one retrieved from the CPG neural network using different similarity criteria. Although several similarity criteria may be chosen, some rules must be considered. 

On the other hand, the atomic properties define fhe peak heighfs and are fherefore imporfanf for fhe chemical nafure of fhe atom pairs or fheir chemical environmenfs. 

After an initial model is retrieved, the molecule is manipulated by addition, removal, and changing of the atom type. The decision for the sequence of optimization steps is dynamically adapted to the properties of the RDF descriptor, the chosen atomic properties, and the similarity criterion. The following properties are altered  

---