广义回归函数genfit模板

zpz77777

广义回归函数genfit模板:
使用广义回归函数的通用性强，自由度高，他之所以没有
专用回归函数那么高的使用频率，那么广为人知，其主要
原因就是使用难度大。不仅花费很多时间，而且还常常出
现错误。
  为此，鄙人编了一个模板，把很麻烦的、容易错的部分
程式化地固定下来，留下必须手动输入部分给使用者填充
使得需要快速进入应用的朋友，或者怕麻烦的懒人弟兄都
可以受益。也许可以使这个万能的回归函数，更多的受到
青睐。

iomega

One suggestion:
For those functions to be fitted that do not have analytical forms, i.e., y = F(a,x) where F(a,x) might be a result from numerical simulation or integrations, you can use the finite difference form to replace the partial derivatives:
For instance,
(f(a+esp*a,x) - f(a-esp*a,x) )/(2*esp), where esp = 0.001.

iomega

Some time the genfit function (or the mathcad itself) stuck or become very slow. To give a example:

I have the same code written in Mathcad and C++ to do multi-variable least square fitting from measurement data. It took Mathcad almost 1 minute to get result using genfit function while C++ code only spent less than 2 seconds to yield the same result using the own-written least square fitting method based on Gauss-Newton iteration method.

zpz77777

2# iomega

The suggestion that thank you, this template from arrive a set since the idea, altogether just could not spend a , can simplify the complicated operation of some users and want to have never thought the problem of raising the speed as well.
I will according to you of suggestion, try to improve it。Again thank.