Tatonnement Procedures for Linearly Constrained Convex Optimization
针对经济模型求解精度要求不高的特点，提出一种类似单纯形法的迭代算法，能高效求解大规模线性约束部分均衡模型，并扩展到线性约束凸优化问题。在68个NETLIB线性规划测试中，与单纯形法和Karmarkar算法相比，中等精度下结果令人鼓舞。
The emphasis in this article is to exploit the fact that precision requirements for solutions of most economic models in practice are moderate only. A simple approach is introduced for solving linearly constrained partial equilibrium models based on an iterative scheme similar to the simplex method. It allows large-scale models to be solved, within a practical tolerance, efficiently even in a micro computer environment. Extensions to linearly constrained convex optimization problems are presented. Finally, a set of computational tests on 68 linear programs from the NETLIB library is reported. Comparison of our approach with the simplex method (using MINOS 5.1) and with Karmarkar's algorithm is reported. For moderate precision requirements these preliminary results are highly encouraging.