论文部分内容阅读
本文提出一种对“相关——最小二乘法”的改进算法——分离系数法。理论分析表明,经过一种特殊非奇异变换后所得到的分离系数法,计算更加有效,数值稳定性得到提高,降低了对计算机字长的要求,并保留原算法的若干优点。实验证实,新算法与原算法相比,节省50％～75％计算量,50％～75％的存贮量,80％的计算时间,并且参数估计量的偏差减为原来的1/10。这些优点使分离系数法尤其适合于微型机对工业对象进行建模工作。
In this paper, we propose an improved algorithm of “correlation-least squares” -the separation factor method. The theoretical analysis shows that after a special non-singular transformation, the separation factor method is more effective and the numerical stability is improved. The requirement of computer word length is reduced, and some advantages of the original algorithm are retained. Experiments show that compared with the original algorithm, the new algorithm can save 50% -75% of the computation, 50% -75% of the storage, 80% of the computation time, and the deviation of the parameter estimation is reduced to 1/10. These advantages make the separation factor method particularly suitable for modeling microcomputers on industrial objects.