SciPy v1.3.0.rc1 Release Notes
Release Date: 2019-04-27 // almost 5 years ago-
π SciPy 1.3.0 Release Notes
Note : Scipy 1.3.0 is not released yet!
SciPy 1.3.0 is the culmination of 5 months of hard work. It contains
β many new features, numerous bug-fixes, improved test coverage and better
π documentation. There have been some API changes
π in this release, which are documented below. All users are encouraged to
π upgrade to this release, as there are a large number of bug-fixes and
β¬οΈ optimizations. Before upgrading, we recommend that users check that
π their own code does not use deprecated SciPy functionality (to do so,
π run your code withpython -Wd
and check forDeprecationWarning
s).
π Our development attention will now shift to bug-fix releases on the
1.3.x branch, and on adding new features on the master branch.π This release requires Python 3.5+ and NumPy 1.13.3 or greater.
For running on PyPy, PyPy3 6.0+ and NumPy 1.15.0 are required.
π Highlights of this release
- Three new
stats
functions, a rewrite ofpearsonr
, and an exact
β computation of the Kolmogorov-Smirnov two-sample test - β‘οΈ A new Cython API for bounded scalar-function root-finders in
scipy.optimize
- π Substantial
CSR
andCSC
sparse matrix indexing performance
π improvements - β Added support for interpolation of rotations with continuous angular
rate and acceleration inRotationSpline
π New features
scipy.interpolate
improvementsA new class
CubicHermiteSpline
is introduced. It is a piecewise-cubic
interpolator which matches observed values and first derivatives. Existing
cubic interpolatorsCubicSpline
,PchipInterpolator
and
Akima1DInterpolator
were made subclasses ofCubicHermiteSpline
.scipy.io
improvementsFor the Attribute-Relation File Format (ARFF)
scipy.io.arff.loadarff
π now supports relational attributes.π
scipy.io.mmread
can now parse Matrix Market format files with empty lines.scipy.linalg
improvementsβ Added wrappers for
?syconv
routines, which convert a symmetric matrix
given by a triangular matrix factorization into two matrices and vice versa.scipy.linalg.clarkson_woodruff_transform
now uses an algorithm that leverages
sparsity. This may provide a 60-90 percent speedup for dense input matrices.
π Truly sparse input matrices should also benefit from the improved sketch
algorithm, which now correctly runs inO(nnz(A))
time.β Added new functions to calculate symmetric Fiedler matrices and
Fiedler companion matrices, namedscipy.linalg.fiedler
and
scipy.linalg.fiedler_companion
, respectively. These may be used
for root finding.scipy.ndimage
improvementsπ Gaussian filter performances may improve by an order of magnitude in
some cases, thanks to removal of a dependence onnp.polynomial
. This
may impactscipy.ndimage.gaussian_filter
for example.β‘οΈ
scipy.optimize
improvementsβ‘οΈ The
scipy.optimize.brute
minimizer obtained a new keywordworkers
, which
can be used to parallelize computation.β‘οΈ A Cython API for bounded scalar-function root-finders in
scipy.optimize
β‘οΈ is available in a new modulescipy.optimize.cython_optimize
viacimport
.
This API may be used withnogil
andprange
to loop
over an array of function arguments to solve for an array of roots more
quickly than with pure Python.0οΈβ£
'interior-point'
is now the default method forlinprog
, and
π'interior-point'
now uses SuiteSparse for sparse problems when the
π required scikits (scikit-umfpack and scikit-sparse) are available.
On benchmark problems (gh-10026), execution time reductions by factors of 2-3
were typical. Also, a newmethod='revised simplex'
has been added.
It is not as fast or robust asmethod='interior-point'
, but it is a faster,
more robust, and equally accurate substitute for the legacy
method='simplex'
.differential_evolution
can now use aBounds
class to specify the
β‘οΈ bounds for the optimizing argument of a function.π
scipy.optimize.dual_annealing
performance improvements related to
vectorisation of some internal code.π¦
scipy.signal
improvementsπ Two additional methods of discretization are now supported by
π¦scipy.signal.cont2discrete
:impulse
andfoh
.π¦
scipy.signal.firls
now uses faster solversπ¦
scipy.signal.detrend
now has a lower physical memory footprint in some
cases, which may be leveraged using the newoverwrite_data
keyword argumentπ¦
scipy.signal.firwin
pass_zero
argument now accepts new string arguments
that allow specification of the desired filter type:'bandpass'
,
'lowpass'
,'highpass'
, and'bandstop'
π
scipy.signal.sosfilt
may have improved performance due to lower retention
π of the global interpreter lock (GIL) in algorithmπ
scipy.sparse
improvementsA new keyword was added to
csgraph.dijsktra
that
π allows users to query the shortest path to ANY of the passed in indices,
β as opposed to the shortest path to EVERY passed index.π
scipy.sparse.linalg.lsmr
performance has been improved by roughly 10 percent
on large problemsπ Improved performance and reduced physical memory footprint of the algorithm
π used byscipy.sparse.linalg.lobpcg
π
CSR
andCSC
sparse matrix fancy indexing performance has been
π improved substantiallyscipy.spatial
improvementsscipy.spatial.ConvexHull
now has agood
attribute that can be used
alongsize theQGn
Qhull options to determine which external facets of a
convex hull are visible from an external query point.scipy.spatial.cKDTree.query_ball_point
has been modernized to use some newer
π Cython features, including GIL handling and exception translation. An issue
π withreturn_sorted=True
and scalar queries was fixed, and a new mode named
return_length
was added.return_length
only computes the length of the
returned indices list instead of allocating the array every time.scipy.spatial.transform.RotationSpline
has been added to enable interpolation
of rotations with continuous angular rates and accelerationscipy.stats
improvementsβ Added a new function to compute the Epps-Singleton test statistic,
scipy.stats.epps_singleton_2samp
, which can be applied to continuous and
discrete distributions.New functions
scipy.stats.median_absolute_deviation
andscipy.stats.gstd
(geometric standard deviation) were added. Thescipy.stats.combine_pvalues
π method now supportspearson
,tippett
andmudholkar_george
pvalue
combination methods.The
scipy.stats.ortho_group
andscipy.stats.special_ortho_group
β‘οΈrvs(dim)
functions' algorithms were updated from aO(dim^4)
implementation to aO(dim^3)
which gives large speed improvements
fordim>100
.A rewrite of
scipy.stats.pearsonr
to use a more robust algorithm,
β provide meaningful exceptions and warnings on potentially pathological input,
and fix at least five separate reported issues in the original implementation.π Improved the precision of
hypergeom.logcdf
andhypergeom.logsf
.β Added exact computation for Kolmogorov-Smirnov (KS) two-sample test, replacing
β the previously approximate computation for the two-sided teststats.ks_2samp
.
β Also added a one-sided, two-sample KS test, and a keywordalternative
to
stats.ks_2samp
.Backwards incompatible changes
scipy.interpolate
changesFunctions from
scipy.interpolate
(spleval
,spline
,splmake
,
andspltopp
) and functions fromscipy.misc
(bytescale
,
fromimage
,imfilter
,imread
,imresize
,imrotate
,
πimsave
,imshow
,toimage
) have been removed. The former set has
π been deprecated since v0.19.0 and the latter has been deprecated since v1.0.0.
Similarly, aliases fromscipy.misc
(comb
,factorial
,
πfactorial2
,factorialk
,logsumexp
,pade
,info
,source
,
πwho
) which have been deprecated since v1.0.0 are removed.
SciPy documentation for v1.1.0 <https://docs.scipy.org/doc/scipy-1.1.0/reference/misc.html>
__
can be used to track the new import locations for the relocated functions.scipy.linalg
changes0οΈβ£ For
pinv
,pinv2
, andpinvh
, the default cutoff values are changed
π for consistency (see the docs for the actual values).scipy.stats
changesβ Previously,
ks_2samp(data1, data2)
would run a two-sided test and return
β the approximated p-value. The new signature,ks_2samp(data1, data2, alternative="two-sided", method="auto")
, still runs the two-sided test by
0οΈβ£ default but returns the exact p-value for small samples and the approximated
value for large samples.method="asymp"
would be equivalent to the
π old version butauto
is the better choice.Other changes
β‘οΈ Our tutorial has been expanded with a new section on global optimizers
There has been a rework of the
stats.distributions
tutorials.β‘οΈ
scipy.optimize
now correctly sets the convergence flag of the result to
CONVERR
, a convergence error, for bounded scalar-function root-finders
if the maximum iterations has been exceeded,disp
is false, and
full_output
is true.β‘οΈ
scipy.optimize.curve_fit
no longer fails ifxdata
andydata
dtypes
differ; they are both now automatically cast tofloat64
.scipy.ndimage
functions includingbinary_erosion
,binary_closing
, and
binary_dilation
now require an integer value for the number of iterations,
which alleviates a number of reported issues.π Fixed normal approximation in case
zero_method == "pratt"
in
scipy.stats.wilcoxon
.π Fixes for incorrect probabilities, broadcasting issues and thread-safety
related to stats distributions setting member variables inside_argcheck()
.β‘οΈ
scipy.optimize.newton
now correctly raises aRuntimeError
, when default
arguments are used, in the case that a derivative of value zero is obtained,
which is a special case of failing to converge.A draft toolchain roadmap is now available, laying out a compatibility plan
including Python versions, C standards, and NumPy versions.Authors
- ananyashreyjain +
- ApamNapat +
- Scott Calabrese Barton +
- Christoph Baumgarten
- Peter Bell +
- Jacob Blomgren +
- Doctor Bob +
- Mana Borwornpadungkitti +
- Matthew Brett
- Evgeni Burovski
- CJ Carey
- Vega Theil Carstensen +
- Robert Cimrman
- Forrest Collman +
- Pietro Cottone +
- David +
- Idan David +
- Christoph Deil
- Dieter WerthmΓΌller
- Conner DiPaolo +
- Dowon
- Michael Dunphy +
- Peter Andreas Entschev +
- Gâkçen Eraslan +
- Johann Faouzi +
- Yu Feng
- Piotr Figiel +
- Matthew H Flamm
- Franz Forstmayr +
- Christoph Gohlke
- Richard Janis Goldschmidt +
- Ralf Gommers
- Lars Grueter
- Sylvain Gubian
- Matt Haberland
- Yaroslav Halchenko
- Charles Harris
- Lindsey Hiltner
- JakobStruye +
- He Jia +
- Jwink3101 +
- Greg Kiar +
- Julius Bier Kirkegaard
- John Kirkham +
- Thomas Kluyver
- Vladimir Korolev +
- Joseph Kuo +
- Michael Lamparski +
- Eric Larson
- Denis Laxalde
- Katrin Leinweber
- Jesse Livezey
- ludcila +
- Dhruv Madeka +
- Magnus +
- Nikolay Mayorov
- Mark Mikofski
- Jarrod Millman
- Markus Mohrhard +
- Eric Moore
- Andrew Nelson
- Aki Nishimura +
- OGordon100 +
- Petar MlinariΔ +
- Stefan Peterson
- Matti Picus +
- Ilhan Polat
- Aaron Pries +
- Matteo Ravasi +
- Tyler Reddy
- Ashton Reimer +
- Joscha Reimer
- rfezzani +
- Riadh +
- Lucas Roberts
- Heshy Roskes +
- Mirko Scholz +
- Taylor D. Scott +
- Srikrishna Sekhar +
- Kevin Sheppard +
- Sourav Singh
- skjerns +
- Kai Striega
- SyedSaifAliAlvi +
- Gopi Manohar T +
- Albert Thomas +
- Timon +
- Paul van Mulbregt
- Jacob Vanderplas
- Daniel Vargas +
- Pauli Virtanen
- VNMabus +
- Stefan van der Walt
- Warren Weckesser
- Josh Wilson
- Nate Yoder +
- Roman Yurchak
π A total of 97 people contributed to this release.
People with a "+" by their names contributed a patch for the first time.
This list of names is automatically generated, and may not be fully complete. - Three new