Computational development of a lagrangian dual approach for quadratic networks

A Lagrangian dual approach for large-scale quadratic programs due to Lin and Pang is adapted to the quadratic cost network flow problem. The unconstrained dual problem is solved iteratively by combining conjugate gradient directions with finite line-search methods. By exploiting the graph structure, matrix and vector operations are streamlined so that the technique can be used for large problems. A computational study of randomly generated networks with up to 500 nodes and 20,000 arcs gives a comparison between combinations of different conjugate direction formulas with three line-search techniques. The results demonstrate the clear superiority of the Polak-Ribiere direction formula combined with a variant of a line-search technique originally developed by Bitran and Hax for convex knapsack problems.

[1]  M. Hestenes,et al.  Methods of conjugate gradients for solving linear systems , 1952 .

[2]  S. Zenios,et al.  Simplicial Decomposition for Convex Generalized Networks , 1990 .

[3]  Leon S. Lasdon,et al.  A Reduced Gradient Algorithm for Nonlinear Network Problems , 1983, TOMS.

[4]  C. Roger Glassey,et al.  A quadratic network optimization model for equilibrium single commodity trade flows , 1978, Math. Program..

[5]  J. Pang,et al.  Iterative methods for large convex quadratic programs: a survey , 1987 .

[6]  Azuma Ohuchi,et al.  Lagrangian dual coordinatewise maximization algorithm for network transportation problems with quadratic costs , 1984, Networks.

[7]  J. G. Klincewicz,et al.  Implementing an "exact" Newton method for separable convex transportation problems , 1989, Networks.

[8]  R. Dembo.  A primal truncated newton algorithm with application to large-scale nonlinear network optimization , 1987 .

[9]  B. Korte,et al.  Minimum norm problems over transportation polytopes , 1980 .

[10]  Darwin Klingman,et al.  NETGEN: A Program for Generating Large Scale Capacitated Assignment, Transportation, and Minimum Cost Flow Network Problems , 1974 .

[11]  D. Bertsekas,et al.  Distributed asynchronous relaxation methods for convex network flow problems , 1987 .

[12]  David E. Boyce,et al.  Improved Efficiency of the Frank-Wolfe Algorithm for Convex Network Programs , 1985, Transp. Sci..

[13]  Jeffery L. Kennington,et al.  Solving the Pipe Network Analysis Problem Using Optimization Techniques , 1978 .

[14]  Jose A. Ventura.  Algorithms for quadratic transportation networks , 1989, Proceedings of the 28th IEEE Conference on Decision and Control,.

[15]  Jeffery L. Kennington,et al.  A polynomially bounded algorithm for a singly constrained quadratic program , 1980, Math. Program..

[16]  R. Cottle,et al.  On the convergence of a block successive over-relaxation method for a class of linear complementarity problems , 1982 .

[17]  R. Cottle,et al.  A Lagrangean relaxation algorithm for the constrained matrix problem , 1986 .

[18]  Pasumarti V. Kamesam,et al.  Multipoint methods for separable nonlinear networks , 1984 .

[19]  François V. Louveaux,et al.  Piecewise convex programs , 1978, Math. Program..

[20]  Guy Matignon,et al.  Comparison of different methods for the calculation of traffic matrices , 1980 .

[21]  C. M. Reeves,et al.  Function minimization by conjugate gradients , 1964, Comput. J..

[22]  S. Zenios,et al.  Relaxation techniques for strictly convex network problems , 1986 .

[23]  L. LeBlanc.  The conjugate gradient technique for certain quadratic network problems , 1976 .

[24]  Arnoldo C. Hax,et al.  Disaggregation and Resource Allocation Using Convex Knapsack Problems with Bounded Variables , 1981 .

[25]  John G. Klincewicz,et al.  A Newton method for convex separable network flow problems , 1983, Networks.

[26]  D. Bertsekas,et al.  Relaxation methods for network flow problems with convex arc costs , 1987 .

[27]  M. Minoux.  A polynomial algorithm for minimum quadratic cost flow problems , 1984 .

[28]  J. A. Ventura,et al.  Restricted simplicial decomposition: computation and extensions , 1987 .