论文部分内容阅读
本文提出了一种将共轭梯度法和最小p乘法结合起来的梯度型最优化方法,简称“共轭—p乘”法。本文用该法以及DFP法,最小二乘法和共轭梯度法设计了一些幅度均衡器,时延均衡器,结果表明“共轭—p乘”法是一种高效和收敛迅速的网络最优化设计方法。为了改善收敛过程,在幅度均衡器和时延均衡器设计中多用了一个恒值参数,并在每次迭代之前单独对它进行优化。这种处理一般都能给出较满意的结果。本文的计算是在FELIXC—512机上用FORTRAN—Ⅳ语言进行的。
In this paper, we propose a gradient-type optimization method combining conjugate gradient method and minimum p-multiplication method, referred to as “conjugate-p multiplication” method. In this paper, some amplitude equalizers and delay equalizers are designed by this method and the DFP method, the least square method and the conjugate gradient method. The results show that the conjugate-p multiplication method is an efficient and convergent network optimization design method. To improve the convergence process, a constant parameter is used in the design of the amplitude equalizer and delay equalizer, and it is optimized separately before each iteration. This treatment can generally give more satisfactory results. The calculation in this paper is done on the FELIXC-512 using the FORTRAN-IV language.