• Page 50-67

Statistical analysis of characteristics of the recognition method on the training set

B.M. Gavrikov1), M.B. Gavrikov, N.V. Pestryakova1)

1) Federal Research Center “Computer Science and Control” of Russian Academy of Sciences, 44/2 Vavilova str., Moscow, 119333, Russia
Annotation

The results of statistical research of the evaluation function, the vector of primary signs and also the matrix of weight coefficients obtained during the training phase with the use of probabilistic high-precision method of recognition of handprinted characters are described.

Keywords

classification, polynomial regression, printed and hand-printed symbols