One approach to data analysis in recognition problems on precedents

  • Authors:
  • N. N. Katerinochkina

  • Affiliations:
  • Dorodnicyn Computing Centre, Russian Academy of Science, Moscow, Russia 119991

  • Venue:
  • Pattern Recognition and Image Analysis
  • Year:
  • 2013

Quantified Score

Hi-index 0.00

Visualization

Abstract

One approach to data analysis in recognition problems on precedents is investigated. The search task for logical regularities of the classes is considered. The concept of elementary predicate is introduced. This predicate determines the belonging of the object to any half-space in the space of features. The logical regularities, which are the disjunctive forms from elementary predicates, are examined. The search methods for these logical regularities are proposed. These methods are based on the constructing convex hulls of subsets of the training sample.