Proposal: Add more verbose interface

[MrHenning]

I am a big fan of both LsqFit.jl and the python package LMFIT.

I made a small wrapper-interface for LsqFit.jl to implement the Parameter functionality of LMFIT, to make the definition of fit-parameters more verbose (e.g. setting min/max values, and toggle if the parameter should be varied),
and maybe other people are interested to use something like this.

Is there any interest to add this interface to LsqFit? Or maybe create a dedicated julia-package?

I use it a follows:

using LMFit

f = 2.45
φ = 0.8
model(x, f, φ) = sin.(x .* f .+ φ)

x = 0:0.1:6pi
y = model(x, f, φ) .+ 0.04 .* randn(length(x))

params = [
    Param(name=:φ, value=0.9, min=0, max=255, vary=true),
    Param(name=:f, value=2.5, min=0, max=255, vary=true),
]

out = curve_fit(model, x , y, params)
@info fitreport(out)
┌ Info: [[Variables]]
│       f:  2.449846120404491  ± 0.0007845311171137703 ( 0.03%) (init = 2.5)
└       φ:  0.7936258619873663 ± 0.008547481398733708  ( 1.08%) (init = 0.9)

I'm happy with any decision 🙂

[garrekstemo]

I also miss the reports and models with parameters in LMFIT

[pkofod]

I see. So you would like to be able to define individual parameters, their domain, and whether they should be optimized or not, rather than just a vector?

[MrHenning]

Yes indeed, I find it useful for prototyping or when I am testing several (slightly different) model.

On the other hand, I came across GModelFit which has similar functionality (see example) which uses LsqFit.jl as a base.
Maybe this is the cleaner option...
(I haven't used GModelFit yet though)