Simulation of Behavior and Intelligence

Integrality constraints in minimizing the empirical risk of linear decision rules

E. V. Rozhkov

Moscow

Abstract: The paper is concerned with the specifics of minimizing the function of empirical risk for linear decision rules by linear programming methods. The conditions of significance of integrality constraints to be imposed on the variables in such problems are determined for the case of two classes of patterns in a Euclidean feature space.

UDC: 519.82


Received: 11.11.1981

 English version: Automation and Remote Control, 1984, 45:3, 390–395

Bibliographic databases:

• 
•