GitHub

Release Notes

0.4.1 (03-09-2025)

Additions

  • m = Minuit(f, x) can be set up with x::ComponentArray and the resultant m.values, m.errors will be ComponentArrays as well.
  • Added MigradOptimizer() which can be used as a solver in the Optimization.jl ecosystem.

Fixes

Other changes

  • Upgraded to CxxWrap 0.17

0.4.0 (21-07-2025)

Breaking changes

  • Replaced local distributions (Chebyshev, ArgusBG) with the ones in DistributionsHEP package
  • To instantiate a RealVar requires a initial value to deduce the T or to specify the parameter T with RealVar{Float64} for example.
  • Removed function RooFit.generateBinned. The data is generated binned if nbins > 0 in the observable variable.
  • The function RooFit.generate returns a DataSet

0.3.3 (03-07-2025)

Fixes

  • Issue #15. Minhit API: Added keyword arguments for limits and fixed.
  • Issue #17. use_pdf=:approximate math regarding binwidth
  • Issue #23. Extra CxxWrap.CxxWrapCore.ConstCxxRef{} when calling array-input FCN?
  • Fix the parameter names for vector FCN without names argument

Additions

  • Added error in the definition of RealVar
  • Added roofit_basics tutorial
  • Plotting is done by using macro @recipe of module RecipesBase
  • Support for non-uniform binned cost functions
  • Added protection in global state (callback array) for multi-threading

0.3.0 (04-04-2025)

Additions

  • Extended cost functions ExtendedBinnedNLL and ExtendedUnbinnedNLL.
    • Added examples in costfunctions example
    • Added unit tests
  • Added CostSum to combine cost functions
    • Added unit test
    • Added combined tutorial
  • Extended Distributions with a Chebyshev polynomial distribution
  • Added elapsed time of the minuit printout
  • Added module RooFit with some very simple elements to build models
    • Added tutorial roofit and unit tests

0.2.1 (5-03-2025)

Fixes

  • Keyword argument for gradient in cost functions is grad
  • Fixed convergence for BinnedNLL with gradient
  • Fix for making assignments of properties with : such as m.values[:] = [...]

Additions

  • Two notebooks for execution performance (julia, python)
  • Added docstring for all properties of Minuit type
  • Added ndof function and property to get the number of degrees of freedom
  • Added extension module Minuit2HistExt to accept FHist histograms as input for BinnedNLL
  • Added simplex! and scan! functions to perform minimization with other methods

0.2.0 (18-02-2025)

  • Initial release that includes the wrapper to all major C++ classes available in Minuit2
  • The currently offered API is minimalistic. It allows to run some introductory tutorials:
    • Introduction to Minuit2.jl
    • Cost Functions
  • The currently available cost functions are:
    • UnbinnedNLL - Unbinned negative log-likelihood cost function.
    • BinnedNLL - Binned negative log-likelihood.
    • LeastSquares - Least-squares cost function (aka chisquare function).
  • The package provides a extension module (if Plots and FHist are loaded) for plotting the data and results of the fit.
  • Documentation is generated using Documenter