大型线性模型参数估计的目标补偿法

An Objective Compensated Approach for Parameter Estimation in Large Scale Linear Models

提出了一种关于线性模型参数估计问题的目标补偿算法，它的主要特点是不用Ｌａｇｒａｎｇｅ乘子作协调变量，而通过对子问题的目标引入补偿项来协调，从而简化了算法结构．理论分析和实例计算表明，它的收敛性较好，适宜在多处理机系统上实现。

In order to overcome the so called “dimensional disasters” caused by the parameter estimation problem of large scale linear models, an objective compensated approach is proposed. Its main feature is that the coordination is realised by introducing compensation terms into the subproblems instead of using Lagrangian multiplier as the coordination variable, so the approach structure is simplified. Theoretical analysis and practical calculation show that its convergence is better and calculating time is shorter than the available parallel decomposition approaches and the whole algorithm of least squares.