engleski

Iterative least-squares program for generalized connectivity descriptors.

ITERATIVE LEAST SQUARES PROGRAM FOR GENERALIZED CONNECTIVITY DESCRIPTORS The generalized connectivity ondex for a given pphysicochemical property of molecules associates to different atoms parameters which need to be determined by minimizing the standard error with regard to a learning set ( M.Randić et al., in press). We report a program based on Newton's iterative non linear least square method which starts with experimental values for a number of molecules, their adjacency matrices and an esstimate of ( nonhydrogen) atom parameters. The program calculates the connectivity index, its linear reggression parameters with the physicochemical property and starts an iterative procedure in which the atom parameters are optimized until the best linear regression of the connectivity indeces with the property is obtained. The procedure yields parameters and confidence limits for determination of molecular physicochemical propery values which are either unknown or included in the learning set.

nije evidentirano

nije evidentirano