Hi! I’m trying to use Julia to make a curve fit with the functional form: model = ( p[1] / x ) * exp( -p[2] / x ) to a data-set I have. I tried LsqFit, however, if I’m not making really good initial guesses for the parameters p[1] and p[2], the algorithm can’t find a good fit (or even remotely c…

[image] PeX: Second, you are all probably right, my data falls at the very initial section of the functional form Ah, I see. If I plot the sum-of-squares error (SSE) that you are minimizing, computed by sse(model, xdata, ydata, p) = sum(((x,y),) -> abs2(y - model(x, p)), zip(xdata,ydata)) t…

Have you tried fitting the log?

Do you mean to transfer x → log10(x) within the same model structure?

What are the ranges of x and y in the data to be fitted?

1e-5 < x < 1e-3 and 0.6 < y < 0.8

[image] PeX: I tried LsqFit, however, if I’m not making really good initial guesses for the parameters p[1] and p[2], the algorithm can’t find a good fit (or even remotely close one), so this kind of misses the point. I can’t reproduce your problem. For this particular model, I’m finding tha…

First of all, thank you for the help! Second, you are all probably right, my data falls at the very initial section of the functional form (the very small values of x where the curve starts to incline in a semi-parabolic fashion). That’s probably why it is difficult to reproduce the fit with the al…

[image] PeX: Are there alternatives to LsqFit that are more robust? so LsqFit.jl then is the go-to package for fitting equations like this in julia? no one who responded to this post so far has addressed this portion of the OP.

[image] bjarthur: so LsqFit.jl then is the go-to package for fitting equations like this in julia? no one who responded to this post so far has addressed this portion of the OP. There are plenty of other optimization packages if you want to treat least-square fitting as a general optimization…

LsqFit is actually mainly an API for quick use of Optim.jl for fitting. I’d also recommend GLM.jl for anything that can be linearized, which is quite common.

Alternatives to LsqFit for nonlinear curve fitting?

General Usage

PeX December 15, 2023, 9:58pm 10

Fantastic response! Very enlightening and insightful! I definitely learned a lot.
Much appreciated!

Least squares with SciML NonlinearSolve.jl - how to?

Topic		Replies	Views
Fitting parameters of the model against data New to Julia curve-fitting	3	1856	July 22, 2022
Custom nonlinear curve fitting package? General Usage curve-fitting	1	475	September 2, 2023
Question: Least square fitting method for non-linear function General Usage question , lsqfit	2	870	June 6, 2023
Curve fitting with NonlinearSolve.jl General Usage curve-fitting	9	730	December 24, 2024
Why is this scenario of function parameter estimation acting strangely? Using LsqFit General Usage package	2	569	August 13, 2020

Alternatives to LsqFit for nonlinear curve fitting?

Related topics