pyminufit

This package is an interface to iminuit to allow for fast simple fits and mimics the functionality of PyrooFit which is an interface on top of the ROOT.RooFit package.

The package allows for simple fits of standard PDFs and easy setup of custom PDFs in one or more fit dimensions.

Example

Simple fit and plot of a Gaussian Distribution:

from pyminufit.models import Gauss
import numpy as np

data = np.random.normal(0, 1, 1000)

pdf = Gauss(("x", -3, 3), mean=(-1, 0, 1))
pdf.fit(data)
pdf.plot(
    "example_gauss.pdf",
)

pdf.get()

A more complex example on combination of Gauss pdf for signal and Polynomial for background:

import pyminufit as mnf
import numpy as np

data = np.append(
    np.random.random_sample(1000) * 10 + 745, np.random.normal(750, 1, 1000)
)

x = mnf.create_real_var(("mass", 745, 755), unit="GeV")

pdf_sig = mnf.Gauss(x, mean=(745, 755), sigma=(0.1, 1, 2), title="Signal")
pdf_bkg = mnf.Chebyshev(x, order=2, title="Background")

pdf = pdf_sig + pdf_bkg

pdf.fit(data)
pdf.plot(data, "02_add_pdf.png", dpi=150)
pdf.get()

Fitting multiple pdf to data.

Observables can be initialised by a list or tuple with the column name / variable name as first argument, followed by the range and/or with the initial value and range:

x = ('x', -3, 3)
x = ('mass', -3, 0.02, 3)

Parameters are initialised with a tuple: sigma=(0,1) or again including a starting parameter: sigma=(0.01, 0, 1) The order here is not important.

Installation

You can install pyminufit directly from PyPI using pip:

pip install pyminufit

Alternatively, if you want to install the development version directly from GitHub, you can do so with:

pip install git+https://github.com/simonuu/pyminufit.git

For a development installation, download the git reposityro and run:

pip install -e .

Development

If you have any questions or need help with this package, your contributions are greatly appreciated. This project is a side project, so any assistance is welcome. When contributing, please follow design principles that prioritize simplicity of the interface while preserving the complexity of the underlying functionality. You can contribute by submitting a pull request (PR) to the GitHub repository. For larger changes consider discussing ideas in GitHub Issues. Thank you for your support!

Planned Features

  • [ ] ProdPdf: Product pdf for multiple observables

  • [ ] Convolutions

  • [ ] Shared parameters

  • [ ] Backend options for the pdf calculation

This package was created using the scientific python template from https://scientific-python.org/.

Indices and tables