Multiplicative methods for entropy programming problems and their applications

The problems of entropy functions maximization over the sets specified with the simultaneous linear or quadratic equalities and inequalities (the problems of entropy-linear and entropy quadratic programming) are under consideration. Multiplicative algorithms with p-active dual and primal variables are proposed for such problems solving. Active variables are selected using the feedback about p-maximal errors in the equations of complement non-rigidity. A convergence of the algorithms proposed is studied. The results of such methods application for the problems of investment portfolio formation and traffic modeling are considered.