FNLSQ in IMSL library

[jjgenglergmail_com]

I want to perform nonlinear least squares. In particular, I want to fit data to a double exponential function of the form y(x) = A*exp(B*x)+C*exp(D*x) where A,B,C, and D can be determined. The routine FNLSQ seems the most appropriate, but the example given makes it appear that this routine cannot do what I want. Is this fitting possible with FNLSQ, or is there a better routine to use?

[ArturGuzik]

Quoting - jjgenglergmail.com

I want to perform nonlinear least squares. In particular, I want to fit data to a double exponential function of the form y(x) = A*exp(B*x)+C*exp(D*x) where A,B,C, and D can be determined. The routine FNLSQ seems the most appropriate, but the example given makes it appear that this routine cannot do what I want. Is this fitting possible with FNLSQ, or is there a better routine to use?

Hi,

the routine should be OK. It uses finite difference approximation for obtaining Jacobian entries. For non-linear LS the most reliable and robust approach is to use Levenberg-Marquardt algorithm. The library offers it in BCLSJ routine which requires user-supplied Jacobian, which you obviously have at hand (as you know your function).

A.