作者: Robert Joseph Vanderbei
DOI:
关键词: Mathematical optimization 、 Linear system 、 System of linear equations 、 Function (mathematics) 、 Convex polytope 、 Mathematics 、 Descent direction 、 Vertex enumeration problem 、 Linear programming 、 Gradient descent
摘要: A method for optimizing resource allocations in commercial enterprises that are characterized by a linear set of constraint relationships the controllable parameters. The is particularly useful where cost function to be minimized piece-wise and convex, or such enterprise can conveniently decomposed into plurality relatively independent subentities correlating portion expresses interrelationships among subentities. More specifically, accordance with disclosed method, an system equations master problem subproblems leads need solve (corresponding problem) convex. solved modification algorithm invented Karmarkar. moves iteratively within polytope feasible solutions (as Karmarkar's invention) move direction improves value function. ascertained scaling determining descent through sensitivity vector evaluation, step size determined line search along chosen best allocation values direction. Whenever possible, corresponds steepest iterations repeated until optimal solution found, whereupon last evaluated (controllable parameters system) communicated system.