|zfit_logo|
*******************************
zfit: scalable pythonic fitting
*******************************
.. image:: https://scikit-hep.org/assets/images/Scikit--HEP-Affiliated-blue.svg
:target: https://scikit-hep.org
.. image:: https://img.shields.io/badge/DOI-10.1016%2Fj.softx.2020.100508-yellow
:target: https://www.sciencedirect.com/science/article/pii/S2352711019303851
:alt: DOI 10.1016/j.softx.2020.100508
.. image:: https://img.shields.io/pypi/pyversions/zfit
:target: https://pypi.org/project/zfit/
:alt: PyPI - Python Version
.. image:: https://img.shields.io/pypi/v/zfit.svg
:target: https://pypi.python.org/pypi/zfit
.. image:: https://img.shields.io/conda/vn/conda-forge/zfit
:target: https://anaconda.org/conda-forge/zfit
:alt: conda-forge
.. image:: https://img.shields.io/spack/v/py-zfit
:target: https://github.com/spack/spack/blob/develop/var/spack/repos/builtin/packages/py-zfit/package.py
.. image:: https://github.com/zfit/zfit/workflows/build/badge.svg
:target: https://github.com/zfit/zfit/actions
.. |zfit_logo| image:: docs/images/zfit-logo_hires.png
:target: https://github.com/zfit/zfit
:alt: zfit logo
.. |scikit-hep_logo| image:: docs/images/scikit-hep-logo_168x168.png
:target: https://scikit-hep.org/affiliated
:alt: scikit-hep logo
zfit is a highly scalable and customizable model manipulation and likelihood fitting library. It uses the same computational backend as
`TensorFlow <https://www.tensorflow.org/>`_ and is optimised for simple and direct manipulation of probability density functions. The project is affiliated with
and well integrated into `Scikit-HEP <https://scikit-hep.org/>`_, the HEP Python ecosystem.
- **Tutorials**: `Interactive introduction and tutorials <https://zfit-tutorials.readthedocs.io/en/latest/>`_
- **Quick start**: `Example scripts <examples>`_
- **Documentation**: See `stable documentation`_ or `latest documentation`_
- **Questions**: see the `FAQ <https://github.com/zfit/zfit/wiki/FAQ>`_,
`ask on StackOverflow <https://stackoverflow.com/questions/ask?tags=zfit>`_ (with the **zfit** tag) or `contact`_ us directly.
- **Physics, HEP**: `zfit-physics <https://github.com/zfit/zfit-physics>`_ is the place to contribute and find more HEP
related content
- **Statistical inference**: `hepstats <https://github.com/scikit-hep/hepstats>`_ for limits, CI, sWeights and more
If you use zfit in **research**, please
consider `citing <https://www.sciencedirect.com/science/article/pii/S2352711019303851>`_.
*N.B.*: zfit is currently in *beta stage*, so while most core parts are established,
some may still be missing and bugs may be encountered.
It is, however, mostly ready for production, and is being used in analyses projects.
If you want to use it for your project and you are not sure if all the needed functionality is there,
feel free to `contact`_ us.
Installation
=============
zfit is available on pip and conda-forge. To install it (recommended: use a virtual/conda env!) with all the dependencies (minimizers, uproot, ...), use
.. code-block:: bash
pip install -U zfit[all]
(the ``-U`` just indicates to upgrade zfit, in case you have it already installed)
or for minimal dependencies
.. code-block:: bash
pip install zfit
For conda/mamba, use
.. code-block:: bash
conda install -c conda-forge zfit
How to use
==========
While the zfit library provides a model fitting and sampling framework for a broad list of applications,
we will illustrate its main features with a simple example by fitting a Gaussian distribution with an unbinned
likelihood fit and a parameter uncertainty estimation.
Example in short
----------------
.. code-block:: python
obs = zfit.Space('x', -10, 10)
# create the model
mu = zfit.Parameter("mu" , 2.4, -1, 5)
sigma = zfit.Parameter("sigma", 1.3, 0, 5)
gauss = zfit.pdf.Gauss(obs=obs, mu=mu, sigma=sigma)
# load the data
data_np = np.random.normal(size=10000)
data = zfit.Data(obs=obs, data=data_np)
# or sample from model
data = gauss.sample()
# build the loss
nll = zfit.loss.UnbinnedNLL(model=gauss, data=data)
# minimize (20+ interchangeable minimizers available!)
minimizer = zfit.minimize.Minuit()
result = minimizer.minimize(nll).update_params()
# calculate errors
sym_errors = result.hesse()
asym_errors = result.errors()
This follows the zfit workflow
.. image:: docs/images/zfit_workflow_v2.png
:alt: zfit workflow
Full explanation
----------------
The default space (e.g. normalization range) of a PDF is defined by an *observable space*, which is created using the ``zfit.Space`` class:
.. code-block:: python
obs = zfit.Space('x', -10, 10)
To create a simple Gaussian PDF, we define its parameters and their limits using the ``zfit.Parameter`` class.
.. code-block:: python
# syntax: zfit.Parameter("any_name", value, lower, upper)
mu = zfit.Parameter("mu" , 2.4, -1, 5)
sigma = zfit.Parameter("sigma", 1.3, 0, 5)
gauss = zfit.pdf.Gauss(obs=obs, mu=mu, sigma=sigma)
For simplicity, we create the dataset to be fitted starting from a numpy array, but zfit allows for the use of other sources such as ROOT files:
.. code-block:: python
mu_true = 0
sigma_true = 1
data_np = np.random.normal(mu_true, sigma_true, size=10000)
data = zfit.Data(obs=obs, data=data_np)
Fits are performed in three steps:
1. Creation of a loss function, in our case a negative log-likelihood.
2. Instantiation of our minimiser of choice, in the example the ``Minuit``.
3. Minimisation of the loss function.
.. code-block:: python
# Stage 1: create an unbinned likelihood with the given PDF and dataset
nll = zfit.loss.UnbinnedNLL(model=gauss, data=data)
# Stage 2: instantiate a minimiser (in this case a basic minuit)
minimizer = zfit.minimize.Minuit()
# Stage 3: minimise the given negative log-likelihood
result = minimizer.minimize(nll).update_params()
The ``.update_params()`` changes the default values of the parameters
(*this is currently happen by default but won't anymore in the future*)
Symmetric errors are calculated with a further function call to avoid running potentially expensive operations if not needed. Asymmetric errors using a
profiling method can also be obtained:
.. code-block:: python
sym_errors = result.hesse()
asym_errors = result.errors()
Once we've performed the fit and obtained the corresponding uncertainties, we can examine the fit results by printing it or looking at individual parts
.. code-block:: python
print(result) # nice representation of a whole result
print("Function minimum:", result.fmin)
print("Converged:", result.converged)
# Information on all the parameters in the fit
params = result.params
print(params)
# Printing information on specific parameters, e.g. mu
print("mu={}".format(params[mu]['value']))
And that's it!
For more details and information of what you can do with zfit, checkout the `latest documentation`_.
Why?
====
The basic idea behind zfit is to offer a Python oriented alternative to the very successful RooFit library
from the `ROOT <https://root.cern.ch/>`_ data analysis package that can integrate with the other packages
that are part if the scientific Python ecosystem.
Contrary to the monolithic approach of ROOT/RooFit, the aim of zfit is to be light and flexible enough t
o integrate with any state-of-art tools and to allow scalability going to larger datasets.
These core ideas are supported by two basic pillars:
- The skeleton and extension of the code is minimalist, simple and finite:
the zfit library is exclusively designed for the purpose of model fitting and sampling with no attempt to extend its
functionalities to features such as statistical methods or plotting.
- zfit is designed for optimal parallelisation and scalability by making use of TensorFlow as its backend.
The use of TensorFlow provides crucial features in the context of model fitting like taking care of the
parallelisation and analytic derivatives.
Prerequisites
=============
``zfit`` works with Python versions 3.9 and above.
The main dependency is `tensorflow <https://www.tensorflow.org/>`_: zfit follows a close version compatibility with TensorFlow.
For a full list of all dependencies, check the `requirements <requirements.txt>`_.
Contributing
============
Any idea of how to improve the library? Or interested to write some code?
Contributions are always welcome, please have a look at the `Contributing guide`_.
.. _Contributing guide: CONTRIBUTING.rst
Contact
========
You can contact us directly:
- via e-mail: zfit@physik.uzh.ch
- join our `Gitter channel <https://gitter.im/zfit/zfit>`_
Original Authors
================
| Jonas Eschle <jonas.eschle@cern.ch>
| Albert Puig <albert.puig@cern.ch>
| Rafael Silva Coutinho <rsilvaco@cern.ch>
See here for `all authors and contributors`_
.. _all authors and contributors: AUTHORS.rst
Acknowledgements
================
zfit has been developed with support from the University of Zurich and the Swiss National Science Foundation (SNSF) under contracts 168169 and 174182.
The idea of zfit is inspired by the `TensorFlowAnalysis <https://gitlab.cern.ch/poluekt/TensorFlowAnalysis>`_ framework
developed by Anton Poluektov and `TensorProb <https://github.com/tensorprob/tensorprob>`_ by Chris Burr and Igor Babuschkin
using the TensorFlow open source library and more libraries.
.. _`latest documentation`: https://zfit.readthedocs.io/en/latest/
.. _`stable documentation`: https://zfit.readthedocs.io/en/stable/
Raw data
{
"_id": null,
"home_page": "https://github.com/zfit/zfit",
"name": "zfit",
"maintainer": "zfit",
"docs_url": null,
"requires_python": "<3.13,>=3.9",
"maintainer_email": "zfit@physik.uzh.ch",
"keywords": "TensorFlow, model, fitting, scalable, HEP, likelihood",
"author": "Jonas Eschle <Jonas.Eschle@cern.ch>, Albert Puig <apuignav@gmail.com>, Rafael Silva Coutinho <rsilvaco@cern.ch>, Matthieu Marinangeli <matthieu.marinangeli@cern.ch>",
"author_email": null,
"download_url": "https://files.pythonhosted.org/packages/ea/00/4d0c8b247999a892140bd841351502797dbc4b95a8a3ecc65c8fb0e1c900/zfit-0.23.0.tar.gz",
"platform": null,
"description": "|zfit_logo|\n\n*******************************\nzfit: scalable pythonic fitting\n*******************************\n\n.. image:: https://scikit-hep.org/assets/images/Scikit--HEP-Affiliated-blue.svg\n :target: https://scikit-hep.org\n\n.. image:: https://img.shields.io/badge/DOI-10.1016%2Fj.softx.2020.100508-yellow\n :target: https://www.sciencedirect.com/science/article/pii/S2352711019303851\n :alt: DOI 10.1016/j.softx.2020.100508\n\n.. image:: https://img.shields.io/pypi/pyversions/zfit\n :target: https://pypi.org/project/zfit/\n :alt: PyPI - Python Version\n\n.. image:: https://img.shields.io/pypi/v/zfit.svg\n :target: https://pypi.python.org/pypi/zfit\n\n.. image:: https://img.shields.io/conda/vn/conda-forge/zfit\n :target: https://anaconda.org/conda-forge/zfit\n :alt: conda-forge\n\n.. image:: https://img.shields.io/spack/v/py-zfit\n :target: https://github.com/spack/spack/blob/develop/var/spack/repos/builtin/packages/py-zfit/package.py\n\n.. image:: https://github.com/zfit/zfit/workflows/build/badge.svg\n :target: https://github.com/zfit/zfit/actions\n\n\n\n\n\n.. |zfit_logo| image:: docs/images/zfit-logo_hires.png\n :target: https://github.com/zfit/zfit\n :alt: zfit logo\n\n.. |scikit-hep_logo| image:: docs/images/scikit-hep-logo_168x168.png\n :target: https://scikit-hep.org/affiliated\n :alt: scikit-hep logo\n\nzfit is a highly scalable and customizable model manipulation and likelihood fitting library. It uses the same computational backend as\n`TensorFlow <https://www.tensorflow.org/>`_ and is optimised for simple and direct manipulation of probability density functions. The project is affiliated with\nand well integrated into `Scikit-HEP <https://scikit-hep.org/>`_, the HEP Python ecosystem.\n\n- **Tutorials**: `Interactive introduction and tutorials <https://zfit-tutorials.readthedocs.io/en/latest/>`_\n- **Quick start**: `Example scripts <examples>`_\n- **Documentation**: See `stable documentation`_ or `latest documentation`_\n- **Questions**: see the `FAQ <https://github.com/zfit/zfit/wiki/FAQ>`_,\n `ask on StackOverflow <https://stackoverflow.com/questions/ask?tags=zfit>`_ (with the **zfit** tag) or `contact`_ us directly.\n- **Physics, HEP**: `zfit-physics <https://github.com/zfit/zfit-physics>`_ is the place to contribute and find more HEP\n related content\n- **Statistical inference**: `hepstats <https://github.com/scikit-hep/hepstats>`_ for limits, CI, sWeights and more\n\n\nIf you use zfit in **research**, please\nconsider `citing <https://www.sciencedirect.com/science/article/pii/S2352711019303851>`_.\n\n*N.B.*: zfit is currently in *beta stage*, so while most core parts are established,\nsome may still be missing and bugs may be encountered.\nIt is, however, mostly ready for production, and is being used in analyses projects.\nIf you want to use it for your project and you are not sure if all the needed functionality is there,\nfeel free to `contact`_ us.\n\nInstallation\n=============\n\nzfit is available on pip and conda-forge. To install it (recommended: use a virtual/conda env!) with all the dependencies (minimizers, uproot, ...), use\n\n.. code-block:: bash\n\n pip install -U zfit[all]\n\n(the ``-U`` just indicates to upgrade zfit, in case you have it already installed)\nor for minimal dependencies\n\n.. code-block:: bash\n\n pip install zfit\n\nFor conda/mamba, use\n\n.. code-block:: bash\n\n conda install -c conda-forge zfit\n\n\n\n\nHow to use\n==========\n\nWhile the zfit library provides a model fitting and sampling framework for a broad list of applications,\nwe will illustrate its main features with a simple example by fitting a Gaussian distribution with an unbinned\nlikelihood fit and a parameter uncertainty estimation.\n\n\nExample in short\n----------------\n.. code-block:: python\n\n obs = zfit.Space('x', -10, 10)\n\n # create the model\n mu = zfit.Parameter(\"mu\" , 2.4, -1, 5)\n sigma = zfit.Parameter(\"sigma\", 1.3, 0, 5)\n gauss = zfit.pdf.Gauss(obs=obs, mu=mu, sigma=sigma)\n\n # load the data\n data_np = np.random.normal(size=10000)\n data = zfit.Data(obs=obs, data=data_np)\n # or sample from model\n data = gauss.sample()\n\n # build the loss\n nll = zfit.loss.UnbinnedNLL(model=gauss, data=data)\n\n # minimize (20+ interchangeable minimizers available!)\n minimizer = zfit.minimize.Minuit()\n result = minimizer.minimize(nll).update_params()\n\n # calculate errors\n sym_errors = result.hesse()\n asym_errors = result.errors()\n\nThis follows the zfit workflow\n\n.. image:: docs/images/zfit_workflow_v2.png\n :alt: zfit workflow\n\n\n\n\nFull explanation\n----------------\n\nThe default space (e.g. normalization range) of a PDF is defined by an *observable space*, which is created using the ``zfit.Space`` class:\n\n\n.. code-block:: python\n\n obs = zfit.Space('x', -10, 10)\n\n\nTo create a simple Gaussian PDF, we define its parameters and their limits using the ``zfit.Parameter`` class.\n\n.. code-block:: python\n\n # syntax: zfit.Parameter(\"any_name\", value, lower, upper)\n mu = zfit.Parameter(\"mu\" , 2.4, -1, 5)\n sigma = zfit.Parameter(\"sigma\", 1.3, 0, 5)\n gauss = zfit.pdf.Gauss(obs=obs, mu=mu, sigma=sigma)\n\nFor simplicity, we create the dataset to be fitted starting from a numpy array, but zfit allows for the use of other sources such as ROOT files:\n\n.. code-block:: python\n\n mu_true = 0\n sigma_true = 1\n data_np = np.random.normal(mu_true, sigma_true, size=10000)\n data = zfit.Data(obs=obs, data=data_np)\n\nFits are performed in three steps:\n\n1. Creation of a loss function, in our case a negative log-likelihood.\n2. Instantiation of our minimiser of choice, in the example the ``Minuit``.\n3. Minimisation of the loss function.\n\n.. code-block:: python\n\n # Stage 1: create an unbinned likelihood with the given PDF and dataset\n nll = zfit.loss.UnbinnedNLL(model=gauss, data=data)\n\n # Stage 2: instantiate a minimiser (in this case a basic minuit)\n minimizer = zfit.minimize.Minuit()\n\n # Stage 3: minimise the given negative log-likelihood\n result = minimizer.minimize(nll).update_params()\n\nThe ``.update_params()`` changes the default values of the parameters\n(*this is currently happen by default but won't anymore in the future*)\n\nSymmetric errors are calculated with a further function call to avoid running potentially expensive operations if not needed. Asymmetric errors using a\nprofiling method can also be obtained:\n\n.. code-block:: python\n\n sym_errors = result.hesse()\n asym_errors = result.errors()\n\nOnce we've performed the fit and obtained the corresponding uncertainties, we can examine the fit results by printing it or looking at individual parts\n\n.. code-block:: python\n\n print(result) # nice representation of a whole result\n\n print(\"Function minimum:\", result.fmin)\n print(\"Converged:\", result.converged)\n\n # Information on all the parameters in the fit\n params = result.params\n print(params)\n\n # Printing information on specific parameters, e.g. mu\n print(\"mu={}\".format(params[mu]['value']))\n\nAnd that's it!\nFor more details and information of what you can do with zfit, checkout the `latest documentation`_.\n\nWhy?\n====\n\nThe basic idea behind zfit is to offer a Python oriented alternative to the very successful RooFit library\nfrom the `ROOT <https://root.cern.ch/>`_ data analysis package that can integrate with the other packages\nthat are part if the scientific Python ecosystem.\nContrary to the monolithic approach of ROOT/RooFit, the aim of zfit is to be light and flexible enough t\no integrate with any state-of-art tools and to allow scalability going to larger datasets.\n\nThese core ideas are supported by two basic pillars:\n\n- The skeleton and extension of the code is minimalist, simple and finite:\n the zfit library is exclusively designed for the purpose of model fitting and sampling with no attempt to extend its\n functionalities to features such as statistical methods or plotting.\n\n- zfit is designed for optimal parallelisation and scalability by making use of TensorFlow as its backend.\n The use of TensorFlow provides crucial features in the context of model fitting like taking care of the\n parallelisation and analytic derivatives.\n\nPrerequisites\n=============\n\n``zfit`` works with Python versions 3.9 and above.\nThe main dependency is `tensorflow <https://www.tensorflow.org/>`_: zfit follows a close version compatibility with TensorFlow.\n\nFor a full list of all dependencies, check the `requirements <requirements.txt>`_.\n\n\nContributing\n============\n\nAny idea of how to improve the library? Or interested to write some code?\nContributions are always welcome, please have a look at the `Contributing guide`_.\n\n.. _Contributing guide: CONTRIBUTING.rst\n\n\nContact\n========\n\nYou can contact us directly:\n - via e-mail: zfit@physik.uzh.ch\n - join our `Gitter channel <https://gitter.im/zfit/zfit>`_\n\n\nOriginal Authors\n================\n\n| Jonas Eschle <jonas.eschle@cern.ch>\n| Albert Puig <albert.puig@cern.ch>\n| Rafael Silva Coutinho <rsilvaco@cern.ch>\n\n\nSee here for `all authors and contributors`_\n\n.. _all authors and contributors: AUTHORS.rst\n\n\nAcknowledgements\n================\n\nzfit has been developed with support from the University of Zurich and the Swiss National Science Foundation (SNSF) under contracts 168169 and 174182.\n\nThe idea of zfit is inspired by the `TensorFlowAnalysis <https://gitlab.cern.ch/poluekt/TensorFlowAnalysis>`_ framework\ndeveloped by Anton Poluektov and `TensorProb <https://github.com/tensorprob/tensorprob>`_ by Chris Burr and Igor Babuschkin\nusing the TensorFlow open source library and more libraries.\n\n.. _`latest documentation`: https://zfit.readthedocs.io/en/latest/\n.. _`stable documentation`: https://zfit.readthedocs.io/en/stable/\n",
"bugtrack_url": null,
"license": "BSD-3-Clause",
"summary": "scalable pythonic model fitting for high energy physics",
"version": "0.23.0",
"project_urls": {
"Homepage": "https://github.com/zfit/zfit"
},
"split_keywords": [
"tensorflow",
" model",
" fitting",
" scalable",
" hep",
" likelihood"
],
"urls": [
{
"comment_text": "",
"digests": {
"blake2b_256": "1ef56b561d4d53b421e65a32220dfe6f61cdf74903280b56de00cba644d85c71",
"md5": "30f1bad8c3cded08a8363b0c0282e23d",
"sha256": "da627cbe7f4af04a97ed2c6190f5a683b5586b6815d83a8786ccfb0ed450b0a3"
},
"downloads": -1,
"filename": "zfit-0.23.0-py2.py3-none-any.whl",
"has_sig": false,
"md5_digest": "30f1bad8c3cded08a8363b0c0282e23d",
"packagetype": "bdist_wheel",
"python_version": "py2.py3",
"requires_python": "<3.13,>=3.9",
"size": 2822069,
"upload_time": "2024-11-07T03:41:51",
"upload_time_iso_8601": "2024-11-07T03:41:51.849343Z",
"url": "https://files.pythonhosted.org/packages/1e/f5/6b561d4d53b421e65a32220dfe6f61cdf74903280b56de00cba644d85c71/zfit-0.23.0-py2.py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": "",
"digests": {
"blake2b_256": "ea004d0c8b247999a892140bd841351502797dbc4b95a8a3ecc65c8fb0e1c900",
"md5": "51599e35f19720f8413a5d300dc90482",
"sha256": "19ec469e1703bd38f8b8957871851ee22fa2e68f0a57b7867cc40ea77df98cc5"
},
"downloads": -1,
"filename": "zfit-0.23.0.tar.gz",
"has_sig": false,
"md5_digest": "51599e35f19720f8413a5d300dc90482",
"packagetype": "sdist",
"python_version": "source",
"requires_python": "<3.13,>=3.9",
"size": 2921126,
"upload_time": "2024-11-07T03:41:55",
"upload_time_iso_8601": "2024-11-07T03:41:55.903620Z",
"url": "https://files.pythonhosted.org/packages/ea/00/4d0c8b247999a892140bd841351502797dbc4b95a8a3ecc65c8fb0e1c900/zfit-0.23.0.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2024-11-07 03:41:55",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "zfit",
"github_project": "zfit",
"travis_ci": false,
"coveralls": false,
"github_actions": true,
"requirements": [],
"lcname": "zfit"
}
JSON
