Script 'mail_helper' called by obssrc
Hello community,
here is the log from the commit of package python-iminuit for openSUSE:Factory
checked in at 2023-01-07 17:19:58
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Comparing /work/SRC/openSUSE:Factory/python-iminuit (Old)
and /work/SRC/openSUSE:Factory/.python-iminuit.new.1563 (New)
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Package is "python-iminuit"
Sat Jan 7 17:19:58 2023 rev:24 rq:1056767 version:2.18.0
Changes:
--------
--- /work/SRC/openSUSE:Factory/python-iminuit/python-iminuit.changes
2022-08-18 16:49:39.361514200 +0200
+++ /work/SRC/openSUSE:Factory/.python-iminuit.new.1563/python-iminuit.changes
2023-01-07 17:23:19.879450638 +0100
@@ -1,0 +2,12 @@
+Sat Jan 7 12:22:22 UTC 2023 - Dirk Müller <dmuel...@suse.com>
+
+- update to v2.18.0:
+ * Bump actions/checkout from 2 to 3
+ * ci: update to Python 3.11 final release
+ * move tutorials
+ * added visualize function to Minuit
+ * Bump pypa/cibuildwheel from 2.10.2 to 2.11.2
+ * add more checks for gradients
+ * Add Python 3.11, drop 3.6
+
+-------------------------------------------------------------------
Old:
----
iminuit-2.16.0.tar.gz
New:
----
iminuit-2.18.0.tar.gz
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Other differences:
------------------
++++++ python-iminuit.spec ++++++
--- /var/tmp/diff_new_pack.OhamsZ/_old 2023-01-07 17:23:20.403453764 +0100
+++ /var/tmp/diff_new_pack.OhamsZ/_new 2023-01-07 17:23:20.407453788 +0100
@@ -1,7 +1,7 @@
#
-# spec file for package python-iminuit
+# spec file
#
-# Copyright (c) 2022 SUSE LLC
+# Copyright (c) 2023 SUSE LLC
#
# All modifications and additions to the file contributed by third parties
# remain the property of their copyright owners, unless otherwise agreed
@@ -22,14 +22,14 @@
%define skip_python36 1
%define modname iminuit
Name: python-%{modname}
-Version: 2.16.0
+Version: 2.18.0
Release: 0
Summary: Python bindings for MINUIT2
License: MIT
URL: https://github.com/scikit-hep/iminuit
Source0:
https://files.pythonhosted.org/packages/source/i/iminuit/%{modname}-%{version}.tar.gz
BuildRequires: %{python_module Cython}
-BuildRequires: %{python_module devel}
+BuildRequires: %{python_module devel >= 3.7}
BuildRequires: %{python_module numpy >= 1.11.3}
BuildRequires: %{python_module numpy-devel}
BuildRequires: %{python_module pybind11 >= 2.9.0}
++++++ iminuit-2.16.0.tar.gz -> iminuit-2.18.0.tar.gz ++++++
diff -urN '--exclude=CVS' '--exclude=.cvsignore' '--exclude=.svn'
'--exclude=.svnignore' old/iminuit-2.16.0/CONTRIBUTING.md
new/iminuit-2.18.0/CONTRIBUTING.md
--- old/iminuit-2.16.0/CONTRIBUTING.md 2022-08-16 18:02:13.000000000 +0200
+++ new/iminuit-2.18.0/CONTRIBUTING.md 2022-12-14 11:10:58.000000000 +0100
@@ -1,3 +1 @@
See doc/contribute.rst in this repository or its [html
version](https://iminuit.readthedocs.io/en/latest/contribute.html).
-
-The wheel building system comes from
[scikit-hep/azure-wheel-helpers](https://github.com/scikit-hep/azure-wheel-helpers),
see that first if you need to update the wheel build system (for a new Python
version, for example).
diff -urN '--exclude=CVS' '--exclude=.cvsignore' '--exclude=.svn'
'--exclude=.svnignore' old/iminuit-2.16.0/PKG-INFO new/iminuit-2.18.0/PKG-INFO
--- old/iminuit-2.16.0/PKG-INFO 2022-08-16 18:02:43.544662700 +0200
+++ new/iminuit-2.18.0/PKG-INFO 2022-12-14 11:11:30.625680000 +0100
@@ -1,6 +1,6 @@
Metadata-Version: 2.1
Name: iminuit
-Version: 2.16.0
+Version: 2.18.0
Summary: Jupyter-friendly Python frontend for MINUIT2 in C++
Home-page: http://github.com/scikit-hep/iminuit
Author: Piti Ongmongkolkul and the iminuit team
@@ -19,11 +19,11 @@
Classifier: Programming Language :: C++
Classifier: Programming Language :: Python
Classifier: Programming Language :: Python :: 3
-Classifier: Programming Language :: Python :: 3.6
Classifier: Programming Language :: Python :: 3.7
Classifier: Programming Language :: Python :: 3.8
Classifier: Programming Language :: Python :: 3.9
Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
Classifier: Programming Language :: Python :: 3 :: Only
Classifier: Programming Language :: Python :: Implementation :: CPython
Classifier: Programming Language :: Python :: Implementation :: PyPy
@@ -34,7 +34,7 @@
Classifier: Operating System :: POSIX
Classifier: Operating System :: Unix
Classifier: Operating System :: MacOS
-Requires-Python: >=3.6
+Requires-Python: >=3.7
Description-Content-Type: text/x-rst
Provides-Extra: test
Provides-Extra: doc
diff -urN '--exclude=CVS' '--exclude=.cvsignore' '--exclude=.svn'
'--exclude=.svnignore' old/iminuit-2.16.0/pyproject.toml
new/iminuit-2.18.0/pyproject.toml
--- old/iminuit-2.16.0/pyproject.toml 2022-08-16 18:02:13.000000000 +0200
+++ new/iminuit-2.18.0/pyproject.toml 2022-12-14 11:10:58.000000000 +0100
@@ -16,7 +16,7 @@
[tool.cibuildwheel]
# update skip when numpy wheels become available
-skip = ["*-musllinux_*", "cp310-win32", "cp310-manylinux_i686"]
+skip = ["*-musllinux_*", "cp31?-win32", "cp31?-manylinux_i686"]
test-requires = "pytest"
test-command = "python -m pytest {package}/tests"
test-skip = ["*universal2:arm64"]
diff -urN '--exclude=CVS' '--exclude=.cvsignore' '--exclude=.svn'
'--exclude=.svnignore' old/iminuit-2.16.0/setup.cfg new/iminuit-2.18.0/setup.cfg
--- old/iminuit-2.16.0/setup.cfg 2022-08-16 18:02:43.548662700 +0200
+++ new/iminuit-2.18.0/setup.cfg 2022-12-14 11:11:30.625680000 +0100
@@ -22,11 +22,11 @@
Programming Language :: C++
Programming Language :: Python
Programming Language :: Python :: 3
- Programming Language :: Python :: 3.6
Programming Language :: Python :: 3.7
Programming Language :: Python :: 3.8
Programming Language :: Python :: 3.9
Programming Language :: Python :: 3.10
+ Programming Language :: Python :: 3.11
Programming Language :: Python :: 3 :: Only
Programming Language :: Python :: Implementation :: CPython
Programming Language :: Python :: Implementation :: PyPy
@@ -42,45 +42,34 @@
package_dir =
= src
packages = iminuit
-python_requires = >=3.6
+python_requires = >=3.7
install_requires = numpy
[options.extras_require]
test =
+ coverage
cython
- flake8
- ipykernel
ipywidgets
- jax
- jaxlib
joblib
jacobi
- jupyter_client
matplotlib
- mypy
- nbconvert
- nbformat
numba
numba-stats
- numpy
- pre-commit
- pydocstyle
- pylint
pytest
- pytest-cov
- pytest-xdist
scipy
- sphinx
- sphinx_rtd_theme
tabulate
- nbsphinx
boost_histogram
- resample
+ resample>=1.5
doc =
- sphinx>=4.1
+ sphinx>=4.1,<5.2
sphinx-rtd-theme
- Jinja2==3.0
nbsphinx
+ nbconvert
+ nbformat
+ jupyter_client
+ ipykernel
+ jax==0.3.2
+ jaxlib==0.3.0
[check-manifest]
ignore =
diff -urN '--exclude=CVS' '--exclude=.cvsignore' '--exclude=.svn'
'--exclude=.svnignore' old/iminuit-2.16.0/src/iminuit/cost.py
new/iminuit-2.18.0/src/iminuit/cost.py
--- old/iminuit-2.16.0/src/iminuit/cost.py 2022-08-16 18:02:13.000000000
+0200
+++ new/iminuit-2.18.0/src/iminuit/cost.py 2022-12-14 11:10:58.000000000
+0100
@@ -19,7 +19,7 @@
histogram of weighted samples
- Fit a template to binned data with bin-wise uncertainties on the template:
- :class:`BarlowBeestonLite`, which also supports weighted data and weighted
templates
+ :class:`Template`, which also supports weighted data and weighted templates
- Fit of a function f(x) to (x, y, yerror) pairs with normal-distributed
fluctuations. x
is one- or multi-dimensional, y is one-dimensional.
@@ -62,6 +62,9 @@
import typing as _tp
import warnings
+CHISQUARE = 1.0
+NEGATIVE_LOG_LIKELIHOOD = 0.5
+
# correct ArrayLike from numpy.typing generates horrible looking signatures
# in python's help(), so we use this as a workaround
_ArrayLike = _tp.Sequence
@@ -100,6 +103,8 @@
Apply Bohm-Zech transform.
See Bohm and Zech, NIMA 748 (2014) 1-6.
+
+ :meta private:
"""
def __init__(self, val: _ArrayLike, var: _ArrayLike):
@@ -114,10 +119,16 @@
Estimated variance of observed values.
"""
val, var = np.atleast_1d(val, var)
- self._scale = val / (var + 1e-323)
+
+ self._scale = np.ones_like(val)
+ np.divide(val, var, out=self._scale, where=var > 0)
self._obs = val * self._scale
- def __call__(self, val: _ArrayLike, var: _tp.Optional[_ArrayLike] = None):
+ def __call__(
+ self, val: _ArrayLike, var: _ArrayLike = None
+ ) -> _tp.Union[
+ _tp.Tuple[np.ndarray, np.ndarray], _tp.Tuple[np.ndarray, np.ndarray,
np.ndarray]
+ ]:
"""
Return precomputed scaled data and scaled prediction.
@@ -218,7 +229,7 @@
return 2 * np.sum(mu - n + n * (_safe_log(n) - _safe_log(mu)))
-def barlow_beeston_lite_chi2_jsc(
+def template_chi2_jsc(
n: _ArrayLike[float], mu: _ArrayLike[float], mu_var: _ArrayLike[float]
) -> float:
"""
@@ -239,7 +250,7 @@
Returns
-------
float
- Cost function value.
+ Asymptotically chi-square-distributed test statistic.
Notes
-----
@@ -251,23 +262,22 @@
beta_var = mu_var / mu**2
- # need to solve quadratic equation b^2 + (mu beta_var - 1) b - n beta_var
= 0
- p = mu * beta_var - 1
- q = -n * beta_var
- beta = 0.5 * (-p + np.sqrt(p**2 - 4 * q))
+ # Eq. 15 from https://doi.org/10.48550/arXiv.2206.12346
+ p = 0.5 - 0.5 * mu * beta_var
+ beta = p + np.sqrt(p**2 + n * beta_var)
return poisson_chi2(n, mu * beta) + np.sum( # type:ignore
(beta - 1) ** 2 / beta_var
)
-def barlow_beeston_lite_chi2_hpd(
+def template_chi2_da(
n: _ArrayLike[float], mu: _ArrayLike[float], mu_var: _ArrayLike[float]
) -> float:
"""
Compute asymptotically chi2-distributed cost for a template fit.
- H.P. Dembinski, https://doi.org/10.48550/arXiv.2206.12346
+ H.P. Dembinski, A. Abdelmotteleb, https://doi.org/10.48550/arXiv.2206.12346
Parameters
----------
@@ -282,7 +292,7 @@
Returns
-------
float
- Cost function value.
+ Asymptotically chi-square-distributed test statistic.
"""
n, mu, mu_var = np.atleast_1d(n, mu, mu_var)
k = mu**2 / mu_var
@@ -290,6 +300,50 @@
return poisson_chi2(n, mu * beta) + poisson_chi2(k, k * beta) #
type:ignore
+def template_nll_asy(
+ n: _ArrayLike[float], mu: _ArrayLike[float], mu_var: _ArrayLike[float]
+) -> float:
+ """
+ Compute marginalized negative log-likelikihood for a template fit.
+
+ This is the negative logarithm of equation 3.15 of the paper by
+ C.A. Argüelles, A. Schneider, T. Yuan,
+ https://doi.org/10.1007/JHEP06(2019)030.
+
+ The authors use a Bayesian approach and integrate over the nuisance
+ parameters. Like the other Barlow-Beeston-lite methods, this is an
+ approximation. The resulting likelihood cannot be turned into an
+ asymptotically chi-square distributed test statistic as detailed
+ in Baker & Cousins, NIM 221 (1984) 437-442.
+
+ Parameters
+ ----------
+ n : array-like
+ Observed counts.
+ mu : array-like
+ Expected counts. This is the sum of the normalised templates scaled
+ with the component yields.
+ mu_var : array-like
+ Expected variance of mu. Must be positive everywhere.
+
+ Returns
+ -------
+ float
+ Negative log-likelihood function value.
+ """
+ from scipy.special import loggamma as lg
+
+ n, mu, mu_var = np.atleast_1d(n, mu, mu_var)
+
+ alpha = mu**2 / mu_var + 1
+ beta = mu / mu_var
+ return -np.sum(
+ alpha * np.log(beta)
+ + lg(n + alpha)
+ - (lg(n + 1) + (n + alpha) * np.log(1 + beta) + lg(alpha))
+ )
+
+
# If numba is available, use it to accelerate computations in float32 and
float64
# precision. Fall back to plain numpy for float128 which is not currently
supported
# by numba.
@@ -299,11 +353,11 @@
@_overload(_safe_log, inline="always")
def _ol_safe_log(x):
- return _safe_log
+ return _safe_log # pragma: no cover
@_overload(_z_squared, inline="always")
def _ol_z_squared(y, ye, ym):
- return _z_squared
+ return _z_squared # pragma: no cover
_unbinned_nll_np = _unbinned_nll
_unbinned_nll_nb = _njit(
@@ -383,7 +437,7 @@
@_overload(_soft_l1_loss, inline="always")
def _ol_soft_l1_loss(z_sqr):
- return _soft_l1_loss_np
+ return _soft_l1_loss_np # pragma: no cover
_soft_l1_cost_np = _soft_l1_cost
_soft_l1_cost_nb = _njit(
@@ -403,22 +457,36 @@
class Cost(abc.ABC):
- """Base class for all cost functions."""
+ """
+ Base class for all cost functions.
+
+ :meta private:
+ """
__slots__ = ("_func_code", "_verbose")
@property
def errordef(self):
- """For internal use."""
- return 1.0
+ """
+ For internal use.
+
+ :meta private:
+ """
+ return self._errordef()
+
+ def _errordef(self):
+ return CHISQUARE
@property
def func_code(self):
- """For internal use."""
+ """
+ For internal use.
+
+ :meta private:
+ """
return self._func_code
@property
- @abc.abstractmethod
def ndata(self):
"""
Return number of points in least-squares fits or bins in a binned fit.
@@ -426,6 +494,10 @@
Infinity is returned if the cost function is unbinned. This is used by
Minuit to
compute the reduced chi2, a goodness-of-fit estimate.
"""
+ return self._ndata()
+
+ @abc.abstractmethod
+ def _ndata(self):
NotImplemented # pragma: no cover
@property
@@ -502,9 +574,7 @@
self.value = value
super().__init__((), False)
- @Cost.ndata.getter # type:ignore
- def ndata(self):
- """See Cost.ndata."""
+ def _ndata(self):
return 0
def _call(self, args):
@@ -566,12 +636,10 @@
def _call(self, args):
r = 0.0
for comp, cargs in self._split(args):
- r += comp._call(cargs)
+ r += comp._call(cargs) / comp.errordef
return r
- @Cost.ndata.getter # type:ignore
- def ndata(self):
- """See Cost.ndata."""
+ def _ndata(self):
return sum(c.ndata for c in self._items)
def __len__(self):
@@ -609,23 +677,29 @@
args = np.atleast_1d(args)
n = sum(hasattr(comp, "visualize") for comp in self)
- fig = plt.gcf()
- if n > 1:
- fig.set_figheight(n * fig.get_figheight())
+
+ w, h = plt.rcParams["figure.figsize"]
+ fig, ax = plt.subplots(1, n, figsize=(w, h * n))
if component_kwargs is None:
component_kwargs = {}
+
i = 0
for k, (comp, cargs) in enumerate(self._split(args)):
- if hasattr(comp, "visualize"):
- i += 1
- plt.subplot(n, 1, i)
- kwargs = component_kwargs.get(k, {})
- comp.visualize(cargs, **kwargs)
+ if not hasattr(comp, "visualize"):
+ continue
+ kwargs = component_kwargs.get(k, {})
+ plt.sca(ax[i])
+ comp.visualize(cargs, **kwargs)
+ i += 1
class MaskedCost(Cost):
- """Base class for cost functions that support data masking."""
+ """
+ Base class for cost functions that support data masking.
+
+ :meta private:
+ """
__slots__ = "_data", "_mask", "_masked"
@@ -665,15 +739,13 @@
class UnbinnedCost(MaskedCost):
- """Base class for unbinned cost functions."""
+ """
+ Base class for unbinned cost functions.
- __slots__ = "_model", "_log"
+ :meta private:
+ """
- @Cost.ndata.getter # type:ignore
- def ndata(self):
- """See Cost.ndata."""
- # unbinned likelihoods have infinite degrees of freedom
- return np.inf
+ __slots__ = "_model", "_log"
def __init__(self, data, model: _tp.Callable, verbose: int, log: bool):
"""For internal use."""
@@ -691,6 +763,10 @@
"""Get number density model."""
... # pragma: no cover
+ def _ndata(self):
+ # unbinned likelihoods have infinite degrees of freedom
+ return np.inf
+
def visualize(self, args: _ArrayLike, model_points: int = 0):
"""
Visualize data and model agreement (requires matplotlib).
@@ -874,7 +950,11 @@
class BinnedCost(MaskedCost):
- """Base class for binned cost functions."""
+ """
+ Base class for binned cost functions.
+
+ :meta private:
+ """
__slots__ = "_xe", "_ndim", "_bztrafo"
@@ -885,11 +965,6 @@
"""Access bin edges."""
return self._xe
- @Cost.ndata.getter # type:ignore
- def ndata(self):
- """See Cost.ndata."""
- return np.prod(self._masked.shape[: self._ndim])
-
def __init__(self, args, n, xe, verbose, *updater):
"""For internal use."""
if not isinstance(xe, _tp.Iterable):
@@ -921,6 +996,9 @@
super().__init__(args, n, verbose)
+ def _ndata(self):
+ return np.prod(self._masked.shape[: self._ndim])
+
def _update_cache(self):
super()._update_cache()
if self._bztrafo:
@@ -959,7 +1037,11 @@
class BinnedCostWithModel(BinnedCost):
- """Base class for binned cost functions."""
+ """
+ Base class for binned cost functions with parametric model.
+
+ :meta private:
+ """
__slots__ = "_xe_shape", "_model", "_model_arg"
@@ -991,23 +1073,49 @@
return d
-class BarlowBeestonLite(BinnedCost):
+class Template(BinnedCost):
"""
Binned cost function for a template fit with uncertainties on the template.
- Compared to the original Beeston-Barlow method, the lite methods uses one
nuisance
- parameter per bin instead of one nuisance parameter per component per bin,
which
- is an approximation. This class offers two different lite methods. The
default
- method used is the one which performs better on average.
-
- The cost function works for both weighted data and weighted templates. The
cost
- function assumes that the weights are independent of the data. This is not
the
- case for sWeights, and the uncertaintes for results obtained with sWeights
will
- only be approximately correct, see C. Langenbruch, Eur.Phys.J.C 82 (2022)
5, 393.
-
- Barlow and Beeston, Comput.Phys.Commun. 77 (1993) 219-228,
- https://doi.org/10.1016/0010-4655(93)90005-W)
- J.S. Conway, PHYSTAT 2011, https://doi.org/10.48550/arXiv.1103.0354
+ This cost function is for a mixture model. Samples originate from two or
more
+ components and we are interested in estimating the yield that originates
from each
+ component. In high-energy physics, one component is often a peaking signal
over a
+ smooth background component. Templates are shape estimates for these
components which
+ are obtained from Monte-Carlo simulation. Even if the Monte-Carlo
simulation is exact,
+ the templates introduce some uncertainty since the Monte-Carlo simulation
produces
+ only a finite sample of events that contribute to each template. This cost
function
+ takes that additional uncertainty into account.
+
+ There are several ways to approach this problem. Barlow and Beeston [1]_
found an
+ exact likelihood for this problem, with one nuisance parameter per
component per bin.
+ Solving this likelihood is somewhat challenging though. The Barlow-Beeston
likelihood
+ also does not handle the additional uncertainty in weighted templates
unless the
+ weights per bin are all equal.
+
+ Other works [2]_ [3]_ [4]_ describe likelihoods that use only one nuisance
parameter
+ per bin, which is an approximation. Some marginalize over the nuisance
parameters with
+ some prior, while others profile over the nuisance parameter. This class
implements
+ several of these methods. The default method is the one which performs
best under most
+ conditions, according to current knowledge. The default may change if this
assessment
+ changes.
+
+ The cost function returns an asymptotically chi-square distributed test
statistic,
+ except for the method "asy", where it is the negative logarithm of the
marginalised
+ likelihood instead. The standard transform [5]_ which we use convert
likelihoods into
+ test statistics only works for (profiled) likelihoods, not for likelihoods
+ marginalized over a prior.
+
+ All methods implemented here have been generalized to work with both
weighted data and
+ weighted templates, under the assumption that the weights are independent
of the data.
+ This is not the case for sWeights, and the uncertaintes for results
obtained with
+ sWeights will only be approximately correct [6]_.
+
+ .. [1] Barlow and Beeston, Comput.Phys.Commun. 77 (1993) 219-228
+ .. [2] Conway, PHYSTAT 2011 proceeding,
https://doi.org/10.48550/arXiv.1103.0354
+ .. [3] Argüelles, Schneider, Yuan, JHEP 06 (2019) 030
+ .. [4] Dembinski and Abdelmotteleb,
https://doi.org/10.48550/arXiv.2206.12346
+ .. [5] Baker and Cousins, NIM 221 (1984) 437-442
+ .. [6] Langenbruch, Eur.Phys.J.C 82 (2022) 5, 393
"""
__slots__ = "_bbl_data", "_impl"
@@ -1019,7 +1127,7 @@
templates: _tp.Sequence[_tp.Sequence],
name: _tp.Collection[str] = None,
verbose: int = 0,
- method: str = "hpd",
+ method: str = "da",
):
"""
Initialize cost function with data and model.
@@ -1028,31 +1136,29 @@
----------
n : array-like
Histogram counts. If this is an array with dimension D+1, where D
is the
- number of histogram axes, then the last dimension must have two
elements
- and is interpreted as pairs of sum of weights and sum of weights
squared.
+ number of histogram axes, then the last dimension must have two
elements and
+ is interpreted as pairs of sum of weights and sum of weights
squared.
xe : array-like or collection of array-like
- Bin edge locations, must be len(n) + 1, where n is the number of
bins.
- If the histogram has more than one axis, xe must be a collection
of the
- bin edge locations along each axis.
+ Bin edge locations, must be len(n) + 1, where n is the number of
bins. If the
+ histogram has more than one axis, xe must be a collection of the
bin edge
+ locations along each axis.
templates : collection of array-like
- Collection of arrays, which contain the histogram counts of each
template.
- The template histograms must use the same axes as the data
histogram. If
- the counts are represented by an array with dimension D+1, where D
is the
- number of histogram axes, then the last dimension must have two
elements
- and is interpreted as pairs of sum of weights and sum of weights
squared.
+ Collection of arrays, which contain the histogram counts of each
template. The
+ template histograms must use the same axes as the data histogram.
If the
+ counts are represented by an array with dimension D+1, where D is
the number
+ of histogram axes, then the last dimension must have two elements
and is
+ interpreted as pairs of sum of weights and sum of weights squared.
name : collection of str, optional
Optional name for the yield of each template. Must have length K.
verbose : int, optional
- Verbosity level. 0: is no output (default).
- 1: print current args and negative log-likelihood value.
- method : {"jsc", "hpd"}, optional
- Which version of the lite method to use. jsc: Method developed by
- J.S. Conway, PHYSTAT 2011,
https://doi.org/10.48550/arXiv.1103.0354.
- hpd: Method developed by H.P. Dembinski. Default is "hpd", which
seems to
- perform slightly better on average. The default may change in the
future
- when more practical experience with both method is gained. Set this
- parameter explicitly to ensure that a particular method is used
now and
- in the future.
+ Verbosity level. 0: is no output (default). 1: print current args
and negative
+ log-likelihood value.
+ method : {"jsc", "asy", "da"}, optional
+ Which method to use. "jsc": Conway's method [2]_. "asy": ASY
method [3]_.
+ "da": DA method [4]_. Default is "da", which to current knowledge
offers the
+ best overall performance. The default may change in the future, so
please set
+ this parameter explicitly in code that has to be stable. For all
methods
+ except the "asy" method, the minimum value is chi-square
distributed.
"""
M = len(templates)
if M < 1:
@@ -1089,13 +1195,24 @@
nt_var.append(tv * f**2)
self._bbl_data = (nt, nt_var)
- if method == "jsc":
- self._impl = barlow_beeston_lite_chi2_jsc
- elif method == "hpd":
- self._impl = barlow_beeston_lite_chi2_hpd
- else:
+ known_methods = {
+ "jsc": template_chi2_jsc,
+ "asy": template_nll_asy,
+ "hpd": template_chi2_da,
+ "da": template_chi2_da,
+ }
+ try:
+ self._impl = known_methods[method]
+ except KeyError:
raise ValueError(
- f"method {method} is not understood, allowed values: {{'jsc',
'hpd'}}"
+ f"method {method} is not understood, allowed values:
{known_methods}"
+ )
+
+ if method == "hpd":
+ warnings.warn(
+ "key 'hpd' is deprecated, please use 'da' instead",
+ category=np.VisibleDeprecationWarning,
+ stacklevel=2,
)
super().__init__(name, n, xe, verbose)
@@ -1125,6 +1242,9 @@
ma = mu > 0
return self._impl(n[ma], mu[ma], mu_var[ma])
+ def _errordef(self):
+ return NEGATIVE_LOG_LIKELIHOOD if self._impl is template_nll_asy else
CHISQUARE
+
def visualize(self, args: _ArrayLike):
"""
Visualize data and model agreement (requires matplotlib).
@@ -1365,11 +1485,6 @@
loss(_z_squared(y, ye, ym)) # type:ignore
)
- @Cost.ndata.getter # type:ignore
- def ndata(self):
- """See Cost.ndata."""
- return len(self._masked)
-
def __init__(
self,
x: _ArrayLike,
@@ -1439,6 +1554,9 @@
ym = _normalize_model_output(ym)
return self._cost(y, yerror, ym)
+ def _ndata(self):
+ return len(self._masked)
+
def visualize(self, args: _ArrayLike, model_points: int = 0):
"""
Visualize data and model agreement (requires matplotlib).
@@ -1557,9 +1675,7 @@
return np.sum(delta**2 * self._covinv)
return np.einsum("i,ij,j", delta, self._covinv, delta)
- @Cost.ndata.getter # type:ignore
- def ndata(self):
- """See Cost.ndata."""
+ def _ndata(self):
return len(self._value)
def visualize(self, args: _ArrayLike):
@@ -1629,3 +1745,23 @@
if isinstance(xe[0], _tp.Iterable):
return tuple(len(xei) - 1 for xei in xe)
return (len(xe) - 1,)
+
+
+_deprecated_content = {
+ "BarlowBeestonLite": ("Template", Template),
+ "barlow_beeston_lite_chi2_jsc": ("template_chi2_jsc", template_chi2_jsc),
+ "barlow_beeston_lite_chi2_hpd": ("template_chi2_da", template_chi2_da),
+}
+
+
+def __getattr__(name: str) -> _tp.Any:
+ if name in _deprecated_content:
+ new_name, obj = _deprecated_content[name]
+ warnings.warn(
+ f"{name} was renamed to {new_name}, please import {new_name}
instead",
+ np.VisibleDeprecationWarning,
+ stacklevel=2,
+ )
+ return obj
+
+ raise AttributeError
diff -urN '--exclude=CVS' '--exclude=.cvsignore' '--exclude=.svn'
'--exclude=.svnignore' old/iminuit-2.16.0/src/iminuit/minuit.py
new/iminuit-2.18.0/src/iminuit/minuit.py
--- old/iminuit-2.16.0/src/iminuit/minuit.py 2022-08-16 18:02:13.000000000
+0200
+++ new/iminuit-2.18.0/src/iminuit/minuit.py 2022-12-14 11:10:58.000000000
+0100
@@ -1244,6 +1244,14 @@
return self
+ def visualize(self, plot=None):
+ """Visualize agreement of current model with data.
+
+ This raises an AttributeError if the cost function has no visualize
+ method.
+ """
+ return self._visualize(plot)(self.values)
+
def hesse(self, ncall: int = None) -> "Minuit":
"""
Run Hesse algorithm to compute asymptotic errors.
@@ -1672,9 +1680,7 @@
y: str,
*,
size: int = 50,
- bound: _tp.Union[
- float, _tp.Tuple[_tp.Tuple[float, float], _tp.Tuple[float, float]]
- ] = 2,
+ bound: _tp.Union[float, _tp.Iterable[_tp.Tuple[float, float]]] = 2,
grid: _tp.Tuple[_ArrayLike, _ArrayLike] = None,
subtract_min: bool = False,
) -> _tp.Tuple[np.ndarray, np.ndarray, np.ndarray]:
@@ -1731,7 +1737,7 @@
if xv.ndim != 1 or yv.ndim != 1:
raise ValueError("grid per parameter must be 1D array-like")
else:
- if isinstance(bound, tuple):
+ if isinstance(bound, _tp.Iterable):
xb, yb = bound
xrange = self._normalize_bound(x, xb)
yrange = self._normalize_bound(y, yb)
@@ -2119,6 +2125,7 @@
Layout,
Dropdown,
)
+ from ipywidgets.widgets.interaction import
show_inline_matplotlib_plots
from IPython.display import clear_output
from matplotlib import pyplot as plt
except ModuleNotFoundError as e:
@@ -2128,16 +2135,7 @@
)
raise
- pyfcn = self.fcn._fcn
-
- if plot is None:
- if hasattr(pyfcn, "visualize"):
- plot = pyfcn.visualize
- else:
- raise ValueError(
- f"class {pyfcn.__class__.__name__} has no visualize
method, "
- "please use the plot argument to pass a visualization
function"
- )
+ plot = self._visualize(plot)
def plot_with_frame(args, from_fit, report_success):
trans = plt.gca().transAxes
@@ -2239,11 +2237,12 @@
out.block = False
self.fixed = save
from_fit = True
- # mutil._show_inline_matplotlib_plots()
with out:
clear_output(wait=True)
plot_with_frame(args, from_fit, report_success)
- mutil._show_inline_matplotlib_plots()
+ with warnings.catch_warnings():
+ warnings.simplefilter("ignore")
+ show_inline_matplotlib_plots()
def on_fit_button_clicked(change):
for x in parameters:
@@ -2262,7 +2261,9 @@
with out:
clear_output(wait=True)
plot_with_frame(self.values, True, report_success)
- mutil._show_inline_matplotlib_plots()
+ with warnings.catch_warnings():
+ warnings.simplefilter("ignore")
+ show_inline_matplotlib_plots()
def on_update_button_clicked(change):
for x in parameters:
@@ -2309,7 +2310,10 @@
for x in parameters:
x.slider.observe(on_slider_change, "value")
- # mutil._show_inline_matplotlib_plots()
+
+ with warnings.catch_warnings():
+ warnings.simplefilter("ignore")
+ show_inline_matplotlib_plots()
on_slider_change(None)
return HBox([out, ui])
@@ -2436,6 +2440,18 @@
else:
p.text(str(self))
+ def _visualize(self, plot):
+ pyfcn = self.fcn._fcn
+ if plot is None:
+ if hasattr(pyfcn, "visualize"):
+ plot = pyfcn.visualize
+ else:
+ raise AttributeError(
+ f"class {pyfcn.__class__.__name__} has no visualize
method, "
+ "please use the plot argument to pass a visualization
function"
+ )
+ return plot
+
def _make_init_state(
pos2var: _tp.Tuple[str, ...], args: np.ndarray, kwds: _tp.Dict[str, float]
diff -urN '--exclude=CVS' '--exclude=.cvsignore' '--exclude=.svn'
'--exclude=.svnignore' old/iminuit-2.16.0/src/iminuit/util.py
new/iminuit-2.18.0/src/iminuit/util.py
--- old/iminuit-2.16.0/src/iminuit/util.py 2022-08-16 18:02:13.000000000
+0200
+++ new/iminuit-2.18.0/src/iminuit/util.py 2022-12-14 11:10:58.000000000
+0100
@@ -36,6 +36,8 @@
Derived classes need to implement methods _set and _get to access
specific properties of the parameter state.
+
+ :meta private:
"""
__slots__ = ("_minuit", "_ndim")
@@ -470,11 +472,11 @@
"""
Get chi2/ndof of the fit.
- This returns NaN if the cost function is unbinned or does not support
- reporting the degrees of freedom.
+ This returns NaN if the cost function is unbinned, errordef is not 1,
+ or if the cost function does not report the degrees of freedom.
"""
- if np.isfinite(self._ndof) and self._ndof > 0:
- return self.fval / self.errordef / self._ndof
+ if np.isfinite(self._ndof) and self._ndof > 0 and self.errordef == 1:
+ return self.fval / self._ndof
return np.nan
@property
@@ -1434,28 +1436,6 @@
return segments
-def _show_inline_matplotlib_plots():
- # Code taken from ipywidgets/interactive.py
- #
- # See comments in the original why this is needed.
- #
- # Copyright (c) Jupyter Development Team.
- # Distributed under the terms of the Modified BSD License.
- #
- # This version was stripped down and modified to remove a deprecation
warning.
- try:
- import matplotlib as mpl
- from matplotlib_inline.backend_inline import flush_figures
- except ImportError: # pragma: no cover
- return # pragma: no cover
-
- if (
- mpl.get_backend() == "module://ipykernel.pylab.backend_inline"
- or mpl.get_backend() == "module://matplotlib_inline.backend_inline"
- ):
- flush_figures() # pragma: no cover
-
-
def _smart_sampling(f, xmin, xmax, start=5, tol=5e-3):
x = np.linspace(xmin, xmax, start)
ynew = f(x)
diff -urN '--exclude=CVS' '--exclude=.cvsignore' '--exclude=.svn'
'--exclude=.svnignore' old/iminuit-2.16.0/src/iminuit/version.py
new/iminuit-2.18.0/src/iminuit/version.py
--- old/iminuit-2.16.0/src/iminuit/version.py 2022-08-16 18:02:13.000000000
+0200
+++ new/iminuit-2.18.0/src/iminuit/version.py 2022-12-14 11:10:58.000000000
+0100
@@ -8,7 +8,7 @@
# - During development, add suffix .devN with N >= 0
# - For release candidates, add suffix .rcN with N >= 0
# - For beta releases, add suffix .betaN with N >= 0
-version = "2.16.0"
+version = "2.18.0"
# We list the corresponding ROOT version of the C++ Minuit2 library here
root_version = "v6-25-02-2017-gd0b406db5e"
diff -urN '--exclude=CVS' '--exclude=.cvsignore' '--exclude=.svn'
'--exclude=.svnignore' old/iminuit-2.16.0/src/iminuit.egg-info/PKG-INFO
new/iminuit-2.18.0/src/iminuit.egg-info/PKG-INFO
--- old/iminuit-2.16.0/src/iminuit.egg-info/PKG-INFO 2022-08-16
18:02:43.000000000 +0200
+++ new/iminuit-2.18.0/src/iminuit.egg-info/PKG-INFO 2022-12-14
11:11:30.000000000 +0100
@@ -1,6 +1,6 @@
Metadata-Version: 2.1
Name: iminuit
-Version: 2.16.0
+Version: 2.18.0
Summary: Jupyter-friendly Python frontend for MINUIT2 in C++
Home-page: http://github.com/scikit-hep/iminuit
Author: Piti Ongmongkolkul and the iminuit team
@@ -19,11 +19,11 @@
Classifier: Programming Language :: C++
Classifier: Programming Language :: Python
Classifier: Programming Language :: Python :: 3
-Classifier: Programming Language :: Python :: 3.6
Classifier: Programming Language :: Python :: 3.7
Classifier: Programming Language :: Python :: 3.8
Classifier: Programming Language :: Python :: 3.9
Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
Classifier: Programming Language :: Python :: 3 :: Only
Classifier: Programming Language :: Python :: Implementation :: CPython
Classifier: Programming Language :: Python :: Implementation :: PyPy
@@ -34,7 +34,7 @@
Classifier: Operating System :: POSIX
Classifier: Operating System :: Unix
Classifier: Operating System :: MacOS
-Requires-Python: >=3.6
+Requires-Python: >=3.7
Description-Content-Type: text/x-rst
Provides-Extra: test
Provides-Extra: doc
diff -urN '--exclude=CVS' '--exclude=.cvsignore' '--exclude=.svn'
'--exclude=.svnignore' old/iminuit-2.16.0/src/iminuit.egg-info/requires.txt
new/iminuit-2.18.0/src/iminuit.egg-info/requires.txt
--- old/iminuit-2.16.0/src/iminuit.egg-info/requires.txt 2022-08-16
18:02:43.000000000 +0200
+++ new/iminuit-2.18.0/src/iminuit.egg-info/requires.txt 2022-12-14
11:11:30.000000000 +0100
@@ -1,38 +1,27 @@
numpy
[doc]
-sphinx>=4.1
+sphinx<5.2,>=4.1
sphinx-rtd-theme
-Jinja2==3.0
nbsphinx
+nbconvert
+nbformat
+jupyter_client
+ipykernel
+jax==0.3.2
+jaxlib==0.3.0
[test]
+coverage
cython
-flake8
-ipykernel
ipywidgets
-jax
-jaxlib
joblib
jacobi
-jupyter_client
matplotlib
-mypy
-nbconvert
-nbformat
numba
numba-stats
-numpy
-pre-commit
-pydocstyle
-pylint
pytest
-pytest-cov
-pytest-xdist
scipy
-sphinx
-sphinx_rtd_theme
tabulate
-nbsphinx
boost_histogram
-resample
+resample>=1.5
diff -urN '--exclude=CVS' '--exclude=.cvsignore' '--exclude=.svn'
'--exclude=.svnignore' old/iminuit-2.16.0/tests/test_cost.py
new/iminuit-2.18.0/tests/test_cost.py
--- old/iminuit-2.16.0/tests/test_cost.py 2022-08-16 18:02:13.000000000
+0200
+++ new/iminuit-2.18.0/tests/test_cost.py 2022-12-14 11:10:58.000000000
+0100
@@ -11,13 +11,14 @@
LeastSquares,
Constant,
NormalConstraint,
- BarlowBeestonLite,
+ Template,
multinominal_chi2,
_soft_l1_loss,
PerformanceWarning,
)
from typing import Sequence
import pickle
+from sys import version_info as pyver
try:
# pytest.importorskip does not work for scipy.stats;
@@ -25,9 +26,9 @@
# even if scipy is not installed
from scipy.stats import norm, truncexpon, multivariate_normal
- scipy_stats_available = True
+ scipy_available = True
except ImportError:
- scipy_stats_available = False
+ scipy_available = False
try:
@@ -57,7 +58,8 @@
def expon_cdf(x, a):
- return 1 - np.exp(-x / a)
+ with np.errstate(over="ignore"):
+ return 1 - np.exp(-x / a)
@pytest.fixture
@@ -96,13 +98,13 @@
return a + b * x
-@pytest.mark.skipif(not scipy_stats_available, reason="scipy.stats is needed")
+@pytest.mark.skipif(not scipy_available, reason="scipy.stats is needed")
def test_norm_logpdf():
x = np.linspace(-3, 3)
assert_allclose(norm_logpdf(x, 3, 2), norm.logpdf(x, 3, 2))
-@pytest.mark.skipif(not scipy_stats_available, reason="scipy.stats is needed")
+@pytest.mark.skipif(not scipy_available, reason="scipy.stats is needed")
def test_norm_pdf():
x = np.linspace(-3, 3)
assert_allclose(norm_pdf(x, 3, 2), norm.pdf(x, 3, 2))
@@ -131,7 +133,7 @@
assert_equal(m.fmin.reduced_chi2, np.nan)
-@pytest.mark.skipif(not scipy_stats_available, reason="scipy.stats is needed")
+@pytest.mark.skipif(not scipy_available, reason="scipy.stats is needed")
def test_UnbinnedNLL_2D():
def model(x_y, mux, muy, sx, sy, rho):
return mvnorm(mux, muy, sx, sy, rho).pdf(x_y.T)
@@ -198,7 +200,7 @@
c.visualize((1, 2), model_points=10)
-@pytest.mark.skipif(not scipy_stats_available, reason="scipy.stats is needed")
+@pytest.mark.skipif(not scipy_available, reason="scipy.stats is needed")
@pytest.mark.skipif(not matplotlib_available, reason="matplotlib is needed")
def test_UnbinnedNLL_visualize_2D():
def model(x_y, mux, muy, sx, sy, rho):
@@ -245,7 +247,7 @@
assert_equal(m.fmin.reduced_chi2, np.nan)
-@pytest.mark.skipif(not scipy_stats_available, reason="scipy.stats is needed")
+@pytest.mark.skipif(not scipy_available, reason="scipy.stats is needed")
def test_ExtendedUnbinnedNLL_2D():
def model(x_y, n, mux, muy, sx, sy, rho):
return n * 1000, n * 1000 * mvnorm(mux, muy, sx, sy, rho).pdf(x_y.T)
@@ -318,7 +320,7 @@
c.visualize((1, 2, 3))
-@pytest.mark.skipif(not scipy_stats_available, reason="scipy.stats is needed")
+@pytest.mark.skipif(not scipy_available, reason="scipy.stats is needed")
@pytest.mark.skipif(not matplotlib_available, reason="matplotlib is needed")
def test_ExtendedUnbinnedNLL_visualize_2D():
def model(x_y, n, mux, muy, sx, sy, rho):
@@ -340,7 +342,7 @@
assert_equal(c.data, c2.data)
-@pytest.mark.skipif(not scipy_stats_available, reason="scipy.stats is needed")
+@pytest.mark.skipif(not scipy_available, reason="scipy.stats is needed")
@pytest.mark.parametrize("verbose", (0, 1))
def test_BinnedNLL(binned, verbose):
mle, nx, xe = binned
@@ -417,7 +419,7 @@
BinnedNLL(1, 2, lambda x, a: 0)
-@pytest.mark.skipif(not scipy_stats_available, reason="scipy.stats is needed")
+@pytest.mark.skipif(not scipy_available, reason="scipy.stats is needed")
def test_BinnedNLL_2D():
truth = (0.1, 0.2, 0.3, 0.4, 0.5)
x, y = mvnorm(*truth).rvs(size=1000, random_state=1).T
@@ -443,7 +445,7 @@
assert cost(*m.values) > m.fval
-@pytest.mark.skipif(not scipy_stats_available, reason="scipy.stats is needed")
+@pytest.mark.skipif(not scipy_available, reason="scipy.stats is needed")
def test_BinnedNLL_2D_with_zero_bins():
truth = (0.1, 0.2, 0.3, 0.4, 0.5)
x, y = mvnorm(*truth).rvs(size=1000, random_state=1).T
@@ -500,7 +502,7 @@
c.visualize((1,))
-@pytest.mark.skipif(not scipy_stats_available, reason="scipy.stats is needed")
+@pytest.mark.skipif(not scipy_available, reason="scipy.stats is needed")
@pytest.mark.skipif(not matplotlib_available, reason="matplotlib is needed")
def test_BinnedNLL_visualize_2D():
truth = (0.1, 0.2, 0.3, 0.4, 0.5)
@@ -523,7 +525,7 @@
assert_equal(c.data, c2.data)
-@pytest.mark.skipif(not scipy_stats_available, reason="scipy.stats is needed")
+@pytest.mark.skipif(not scipy_available, reason="scipy.stats is needed")
@pytest.mark.parametrize("verbose", (0, 1))
def test_ExtendedBinnedNLL(binned, verbose):
mle, nx, xe = binned
@@ -563,7 +565,7 @@
ExtendedBinnedNLL([1], [1], lambda x, a: 0)
-@pytest.mark.skipif(not scipy_stats_available, reason="scipy.stats is needed")
+@pytest.mark.skipif(not scipy_available, reason="scipy.stats is needed")
def test_ExtendedBinnedNLL_2D():
truth = (1.0, 0.1, 0.2, 0.3, 0.4, 0.5)
x, y = mvnorm(*truth[1:]).rvs(size=int(truth[0] * 1000), random_state=1).T
@@ -583,7 +585,7 @@
assert_allclose(m.values, truth, atol=0.1)
-@pytest.mark.skipif(not scipy_stats_available, reason="scipy.stats is needed")
+@pytest.mark.skipif(not scipy_available, reason="scipy.stats is needed")
def test_ExtendedBinnedNLL_3D():
truth = (1.0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7)
n = int(truth[0] * 10000)
@@ -638,7 +640,7 @@
c.visualize((1, 2))
-@pytest.mark.skipif(not scipy_stats_available, reason="scipy.stats is needed")
+@pytest.mark.skipif(not scipy_available, reason="scipy.stats is needed")
@pytest.mark.skipif(not matplotlib_available, reason="matplotlib is needed")
def test_ExtendedBinnedNLL_visualize_2D():
truth = (1.0, 0.1, 0.2, 0.3, 0.4, 0.5)
@@ -892,11 +894,29 @@
assert cs((1, 1)) == lsq((1, 1)) + con((1, 1)) + 1.5
+@pytest.mark.skipif(not scipy_available, reason="scipy is needed")
+def test_CostSum_4():
+
+ t = Template([1, 2], [1, 2, 3], [[1, 1], [0, 1]], method="asy")
+ assert t.errordef == Minuit.LIKELIHOOD
+
+ m1 = Minuit(t, 1, 1)
+ m1.migrad()
+
+ cs = CostSum(t)
+ assert cs.errordef == Minuit.LEAST_SQUARES
+
+ m2 = Minuit(cs, 1, 1)
+ m2.migrad()
+
+ assert_allclose(m1.errors, m2.errors)
+
+
@pytest.mark.skipif(not matplotlib_available, reason="matplotlib is needed")
def test_CostSum_visualize():
lsq = LeastSquares([1, 2, 3], [3, 4, 5], 1, line)
con = NormalConstraint(("a", "b"), (1, 1), (1, 1))
- c = lsq + con
+ c = lsq + con + 1
c.visualize((1, 2))
@@ -1069,18 +1089,29 @@
assert c(1) == 0
-@pytest.mark.parametrize("method", ("jsc", "hpd"))
-def test_BarlowBeestonLite(method):
- n = np.array([1, 2, 3])
+@pytest.mark.parametrize("method", ("jsc", "asy", "da"))
+def test_Template(method):
+ if method == "asy" and not scipy_available:
+ pytest.skip(reason="scipy needed")
xe = np.array([0, 1, 2, 3])
t = np.array([[1, 1, 0], [0, 1, 3]])
+ n = t[0] + t[1]
- c = BarlowBeestonLite(n, xe, t, method=method)
+ c = Template(n, xe, t, method=method)
m = Minuit(c, 1, 1)
m.migrad()
assert m.valid
- assert_allclose(m.fval, 0, atol=1e-4)
- assert_allclose(m.values, [2, 4], atol=1e-2)
+ assert m.ndof == 1
+ if method == "asy":
+ assert c.errordef == 0.5
+ assert_equal(m.fmin.reduced_chi2, np.nan)
+ # asy produces values far away from truth in this case
+ assert_allclose(m.values, [1, 3], atol=0.2)
+ else:
+ assert c.errordef == 1.0
+ assert_allclose(m.fval, 0, atol=1e-4)
+ assert_allclose(m.fmin.reduced_chi2, 0, atol=1e-5)
+ assert_allclose(m.values, [2, 4], atol=1e-2)
def generate(rng, nmc, truth, bins, tf=1, df=1):
@@ -1099,18 +1130,20 @@
return n, xe, np.array(t)
-@pytest.mark.skipif(not scipy_stats_available, reason="scipy.stats is needed")
-@pytest.mark.parametrize("method", ("jsc", "hpd"))
+@pytest.mark.skipif(not scipy_available, reason="scipy.stats is needed")
+@pytest.mark.parametrize("method", ("jsc", "asy", "da"))
@pytest.mark.parametrize("with_mask", (False, True))
@pytest.mark.parametrize("weighted_data", (False, True))
-def test_BarlowBeestonLite_weighted(method, with_mask, weighted_data):
+def test_Template_weighted(method, with_mask, weighted_data):
+ if method == "asy" and not scipy_available:
+ pytest.skip(reason="scipy needed")
rng = np.random.default_rng(1)
truth = 750, 250
z = []
rng = np.random.default_rng(1)
for itoy in range(100):
ni, xe, ti = generate(rng, 400, truth, 15, 1.5, 1.5 if weighted_data
else 1)
- c = BarlowBeestonLite(ni, xe, ti, method=method)
+ c = Template(ni, xe, ti, method=method)
if with_mask:
cx = 0.5 * (xe[1:] + xe[:-1])
c.mask = cx != 1.5
@@ -1124,40 +1157,38 @@
break
assert m.valid
z.append((m.values[1] - truth[1]) / m.errors[1])
- assert_allclose(np.mean(z), 0, atol=0.15)
+ assert_allclose(np.mean(z), 0, atol=0.3)
assert_allclose(np.std(z), 1, rtol=0.1)
-@pytest.mark.parametrize("method", ("jsc", "hpd"))
-def test_BarlowBeestonLite_bad_input(method):
+def test_Template_bad_input():
with pytest.raises(ValueError):
- BarlowBeestonLite([1, 2], [1, 2, 3], [], method=method)
+ Template([1, 2], [1, 2, 3], [])
with pytest.raises(ValueError, match="do not match"):
- BarlowBeestonLite([1, 2], [1, 2, 3], [[1, 2, 3], [1, 2, 3]],
method=method)
+ Template([1, 2], [1, 2, 3], [[1, 2, 3], [1, 2, 3]])
with pytest.raises(ValueError, match="do not match"):
- BarlowBeestonLite(
+ Template(
[1, 2],
[1, 2, 3],
[[[1, 2], [3, 4]], [[1, 2], [3, 4], [5, 6]]],
- method=method,
)
with pytest.raises(ValueError, match="not understood"):
- BarlowBeestonLite([1], [1, 2], [[1]], method="foo")
+ Template([1], [1, 2], [[1]], method="foo")
with pytest.raises(ValueError, match="number of names"):
- BarlowBeestonLite([1], [1, 2], [[1]], name=("b", "s"))
+ Template([1], [1, 2], [[1]], name=("b", "s"))
@pytest.mark.skipif(not matplotlib_available, reason="matplotlib is needed")
-def test_BarlowBeestonLite_visualize():
+def test_Template_visualize():
xe = [0, 1, 2]
n = [1, 2]
t = [[1, 2], [5, 4]]
- c = BarlowBeestonLite(n, xe, t)
+ c = Template(n, xe, t)
c.visualize((1, 2))
@@ -1166,24 +1197,49 @@
@pytest.mark.skipif(not matplotlib_available, reason="matplotlib is needed")
-def test_BarlowBeestonLite_visualize_2D():
+def test_Template_visualize_2D():
xe = ([0, 1, 2], [0, 1, 2])
n = [[1, 2], [3, 4]]
t = [[[1, 2], [1, 2]], [[5, 4], [5, 4]]]
- c = BarlowBeestonLite(n, xe, t)
+ c = Template(n, xe, t)
with pytest.raises(ValueError, match="not implemented for
multi-dimensional"):
c.visualize((1, 2))
-def test_BarlowBeestonLite_pickle():
+def test_Template_pickle():
n = np.array([1, 2, 3])
xe = np.array([0, 1, 2, 3])
t = np.array([[1, 1, 0], [0, 1, 3]])
- c = BarlowBeestonLite(n, xe, t)
+ c = Template(n, xe, t)
b = pickle.dumps(c)
c2 = pickle.loads(b)
assert_equal(c.data, c2.data)
+
+
+@pytest.mark.skipif(pyver < (3, 7), reason="module getattr requires
Python-3.7+")
+def test_deprecated():
+ from iminuit import cost
+
+ with pytest.warns(np.VisibleDeprecationWarning):
+ from iminuit.cost import BarlowBeestonLite
+ assert BarlowBeestonLite is cost.Template
+
+ with pytest.warns(np.VisibleDeprecationWarning):
+ from iminuit.cost import barlow_beeston_lite_chi2_jsc
+ assert barlow_beeston_lite_chi2_jsc is cost.template_chi2_jsc
+
+ with pytest.warns(np.VisibleDeprecationWarning):
+ from iminuit.cost import barlow_beeston_lite_chi2_hpd
+ assert barlow_beeston_lite_chi2_hpd is cost.template_chi2_da
+
+
+def test_deprecated_Template_method():
+ from iminuit import cost
+
+ with pytest.warns(np.VisibleDeprecationWarning):
+ t = Template([1], [2, 3], [[1], [2]], method="hpd")
+ t._impl is cost.template_chi2_da
diff -urN '--exclude=CVS' '--exclude=.cvsignore' '--exclude=.svn'
'--exclude=.svnignore' old/iminuit-2.16.0/tests/test_draw.py
new/iminuit-2.18.0/tests/test_draw.py
--- old/iminuit-2.16.0/tests/test_draw.py 2022-08-16 18:02:13.000000000
+0200
+++ new/iminuit-2.18.0/tests/test_draw.py 2022-12-14 11:10:58.000000000
+0100
@@ -177,7 +177,7 @@
ipywidgets_available = True
m = Minuit(cost, 1, 1)
- with pytest.raises(ValueError, match="no visualize method"):
+ with pytest.raises(AttributeError, match="no visualize method"):
m.interactive(raise_on_exception=True)
with plot.assert_call():
diff -urN '--exclude=CVS' '--exclude=.cvsignore' '--exclude=.svn'
'--exclude=.svnignore' old/iminuit-2.16.0/tests/test_minuit.py
new/iminuit-2.18.0/tests/test_minuit.py
--- old/iminuit-2.16.0/tests/test_minuit.py 2022-08-16 18:02:13.000000000
+0200
+++ new/iminuit-2.18.0/tests/test_minuit.py 2022-12-14 11:10:58.000000000
+0100
@@ -211,11 +211,13 @@
m.scipy()
-def test_func0(): # check that providing gradient improves convergence
+def test_func0():
m1 = func_test_helper(func0)
m2 = func_test_helper(func0, grad=func0_grad)
assert m1.ngrad == 0
assert m2.ngrad > 0
+ # check that providing gradient improves convergence
+ assert m2.nfcn < m1.nfcn
def test_lambda():
@@ -330,6 +332,7 @@
assert m.fixed == (False, True)
assert m.limits["a"] == (0, 2)
m.migrad()
+ assert m.fmin.ngrad > 0
assert_allclose(m.values, (1, 1), rtol=1e-2)
c = m.covariance
assert_allclose(c, ((1, 0), (0, 0)), rtol=1e-2)
@@ -1034,20 +1037,23 @@
assert m.nfcn < ncalls_without_limit
-def test_ngrad():
+@pytest.mark.parametrize("arg", (1, np.array([1.0, 2.0])))
+def test_ngrad(arg):
class Func:
ngrad = 0
def __call__(self, x):
- return x**2
+ return np.sum(x**2)
def grad(self, x):
self.ngrad += 1
+ if np.ndim(x) == 1:
+ return 2 * x
return [2 * x]
# check that counting is accurate
fcn = Func()
- m = Minuit(fcn, 1)
+ m = Minuit(fcn, arg)
m.migrad()
assert m.ngrad > 0
assert m.ngrad == fcn.ngrad
@@ -1061,6 +1067,15 @@
m.hesse()
assert m.ngrad == before
+ m.reset()
+ m.migrad()
+ m2 = Minuit(lambda x: fcn(x), arg)
+ m2.migrad()
+ assert m.ngrad > 0
+ assert m2.ngrad == 0
+ # apparently this is not always the case:
+ # assert m2.nfcn > m.nfcn
+
def test_errordef():
m = Minuit(lambda x: x**2, 0)
@@ -1619,3 +1634,14 @@
assert_allclose(m.errors, 10)
m.errors = (1, 2)
assert_allclose(m.errors, (1, 2))
+
+
+def test_visualize():
+ m = Minuit(func0, 1, 1)
+ m.migrad()
+ with pytest.raises(AttributeError):
+ m.visualize()
+
+ func0.visualize = lambda args: None
+ m.visualize()
+ del func0.visualize
diff -urN '--exclude=CVS' '--exclude=.cvsignore' '--exclude=.svn'
'--exclude=.svnignore' old/iminuit-2.16.0/tests/test_util.py
new/iminuit-2.18.0/tests/test_util.py
--- old/iminuit-2.16.0/tests/test_util.py 2022-08-16 18:02:13.000000000
+0200
+++ new/iminuit-2.18.0/tests/test_util.py 2022-12-14 11:10:58.000000000
+0100
@@ -321,11 +321,12 @@
assert repr(mes) == f"<MErrors\n {mes['x']!r}\n>"
-def test_FMin():
+@pytest.mark.parametrize("errordef", (0.5, 1.0))
+def test_FMin(errordef):
fm = Namespace(
fval=1.23456e-10,
edm=1.23456e-10,
- errordef=0.5,
+ errordef=errordef,
is_valid=True,
has_valid_parameters=True,
has_accurate_covar=True,
@@ -371,13 +372,19 @@
assert fmin != util.FMin(fm, "bar", 1, 2, 1, 0.1, 1.2)
assert fmin != util.FMin(fm, "foo", 1, 2, 1, 0.1, 1.5)
+ if errordef == 1:
+ reduced_chi2 = fmin.fval
+ else:
+ reduced_chi2 = np.nan
+
assert repr(fmin) == (
- "<FMin algorithm='foo' edm=1.23456e-10 edm_goal=0.1 errordef=0.5
fval=1.23456e-10"
+ f"<FMin algorithm='foo' edm=1.23456e-10 edm_goal=0.1
errordef={errordef}"
+ " fval=1.23456e-10"
" has_accurate_covar=True has_covariance=True
has_made_posdef_covar=False"
" has_parameters_at_limit=False has_posdef_covar=True"
" has_reached_call_limit=False has_valid_parameters=True"
" hesse_failed=False is_above_max_edm=False is_valid=True"
- " nfcn=1 ngrad=2 reduced_chi2=2.46912e-10 time=1.2>"
+ f" nfcn=1 ngrad=2 reduced_chi2={reduced_chi2} time=1.2>"
)
