Abstract. Two approaches to solution of the task of restoring of dependence between a vector of independent variables and a dependent scalar according to training sampling are considered. The first (parametrical) approach is grounded on a hypothesis of existence of piecewise linear dependence, and private linear dependences correspond to some intervals of change of the dependent parameter. The second (nonparametric) approach consists in solution of main task as search of collective solution on set of tasks of recognition

Keywords: dependence restoring, regression, algorithm of recognition, piecewise linear function, feature, dynamic programming

ACM Classification Keywords: A.0 General Literature - Conference proceedings, G.1.2 Approximation: Nonlinear approximation, H.4.2 Types of Systems: Decision support, I.2 Artificial intelligence, I.5 Pattern recognition

Link:

RESTORING OF DEPENDENCES ON SAMPLINGS OF PRECEDENTS WITH USAGE OF MODELS OF RECOGNITION

V.V.Ryazanov, Ju.I.Tkachev

http://foibg.com/ibs_isc/ibs-16/ibs-16-p02.pdf