Computational Experience With the Dantzig-Wolfe and Kornai-Liptak Decomposition Algorithm
Presents the modified Kornai-Liptak (KL, 1965) decomposition algorithm which is proven to show a fast and monotonic convergence, and compares its computational efficiency to that of the Dantzig-Wolfe (DW, 1960) algorithm, as these are the two basic decomposition algorithms. It is shown that the modified KL algorithm is quite efficient, due to its primal feasibility, when compared to the DW algorithm for the purpose of getting a `well-defined approximation'. This fact is especially important for large-scale problems where the solution process may need to be terminated before completion due to time constraints. Thus, the modified KL algorithm has a great potential for real world applications.
||Computational Experience With the Dantzig-Wolfe and Kornai-Liptak Decomposition Algorithm
International Journal of Policy and Information (1986)
Vol. 10, No. 1
||Lee, Sang M; Rho, B.