Використання методу гілок і границь для розв’язання задачі дискретної оптимізації з метою вибору оптимальної регресійної моделі

Abstract

Article suggests a stochastic method of leaps and bounds to solve a discrete optimization task for the optimum regression model choice with minimax function of model quality. For partition of a current set of task solutions into parts of branching subsets, the dichotomy principle is used. For a branching subset, the lower estimation is calculated for the goal function of the optimum model. The choice of a current subset of the branching process is carried out by a stochastic procedure.