摘要
算法的计算量可用其乘法运算次数和加法运算次数表示(除法作为乘法对待,减法作为加法对待).一次乘法运算或一次加法运算称为一个flop,即一次浮点运算.作为"辨识方法的计算效率"系列3篇连载论文的第1篇,主要了讨论递推辨识算法的计算量,包括向量和矩阵基本运算的flop数,以及线性回归系统、多元线性回归系统、多变量系统的随机梯度辨识算法、最小二乘辨识算法、递推最小二乘辨识算法的最经济计算量,即实现算法的最少flop数.
The amount of the calculation of an algorithm may be expressed by the number of multiplication and ad- dition operations （one division is treated as a multiplication, one subtraction treated as an addition）. A multiplica- tion or an addition operation is called a flop, i. e. , a floating-point operation. This is the first of three serial papers ＂ Computational efficiency of the identification methods＇, which focuses on the computational efficiency of the recursire algorithms, including the flops of the vector and matrix operations, and the minimum flops of the stochastic gradient identification algorithm, the least squares identification algorithm, the recursive least squares identification algorithm for linear regression systems, multivariate linear regression systems and muhivariable systems.
出处
《南京信息工程大学学报（自然科学版）》
CAS
2012年第4期289-300,共12页
Journal of Nanjing University of Information Science & Technology（Natural Science Edition）
基金
国家自然科学基金(60973043)
高等学校学科创新引智计划(B12018)
