As the title says, is the package abandoned? There’s PRs that go back to 2020. Is LsqFit.jl no longer maintained? If so what’s the preferred way to do curve fitting similar to scipy.optimize.curve_fit?
1 Like
No recent commits != doesn’t work well. See this recent thread including the comment at this link which mentions a “newer” package alternative. Alternatives to LsqFit for nonlinear curve fitting? - #14 by RomeoV
3 Likes
I would not say it’s abandoned, but probably not the most active project either. I did some work in the fall.
The code in LsqFit.jl was originally in Optim.jl but it felt like a misfit. I moved it out. There are some open issues on complex data and also the variance-covariance calculation in there. The open PR on inference (whether MSE should be included in the variance-covariance calculation or not) has active discussion on the correct behavior.
If some of the older PRs can be processed to completion that is fine, but sometimes old PRs end like this callback implemented and current x added to metadata dictionary of the trace by viktmar · Pull Request #230 · JuliaNLSolvers/LsqFit.jl · GitHub Someone did the work and that is much appreciated, but there were no tests and I asked for them. They were never added, so the PR was never merged because it is not complete.
Edit: I know it’s a bad comparison but julialang has open PRs going back to 2014 
7 Likes