SciPy 1.16.0 Release Notes

Note

SciPy 1.16.0 is not released yet!

SciPy 1.16.0 is the culmination of 6 months of hard work. It contains many new features, numerous bug-fixes, improved test coverage and better documentation. There have been a number of deprecations and API changes in this release, which are documented below. All users are encouraged to upgrade to this release, as there are a large number of bug-fixes and optimizations. Before upgrading, we recommend that users check that their own code does not use deprecated SciPy functionality (to do so, run your code with python -Wd and check for DeprecationWarning s). Our development attention will now shift to bug-fix releases on the 1.16.x branch, and on adding new features on the main branch.

This release requires Python 3.11-3.13 and NumPy 1.25.2 or greater.

Highlights of this release

• Improved experimental support for the Python array API standard, including new support in scipy.signal, and additional support in scipy.stats and scipy.special. Improved support for JAX and Dask backends has been added, with notable support in scipy.cluster.hierarchy, many functions in scipy.special, and many of the trimmed stats functions.

• scipy.optimize now uses the new Python implementation from the PRIMA package for COBYLA. The PRIMA implementation fixes many bugs in the old Fortran 77 implementation with a better performance on average.

• scipy.sparse.coo_array now supports nD arrays with reshaping, arithmetic and reduction operations like sum/mean/min/max. No nD indexing or random_array support yet.

• Updated guide and tools for migration from sparse matrices to sparse arrays.

• All functions in the scipy.linalg namespace that accept array arguments now support N-dimensional arrays to be processed as a batch.

• Two new scipy.signal functions, firwin_2d and closest_STFT_dual_window, for creation of a 2-D FIR filter and ShortTimeFFT dual window calculation, respectivly.

• A new class, RigidTransform, is available in the transform submodule. It provides functionality to convert between different representations of rigid transforms in 3D space.

New features

scipy.io improvements

• savemat now provides informative warnings for invalid field names.

• scipy.io.mmread now provides a clearer error message when provided with a source file path that does not exist.

• scipy.io.wavfile.read can now read non-seekable files.

scipy.integrate improvements

• Improved the error estimate of scipy.integrate.tanhsinh

scipy.interpolate improvements

• Added batch support to scipy.interpolate.make_smoothing_spline

scipy.linalg improvements

• All functions in the scipy.linalg namespace that accept array arguments now support N-dimensional arrays to be processed as a batch. See Batched Linear Operations for details.

• linalg.sqrtm is rewritten in C and its performance is improved. It also tries harder to return real-valued results for real-valued inputs if possible. See the function docstring for more details. In this version the input argument disp and the optional output argument errest are deprecated and will be removed four versions later. Similarly, after changing the underlying algorithm to recursion, the blocksize keyword argument has no effect and will be removed two versions later.

• Added wrappers for ?stevd, ?langb, ?sytri, ?hetri and ?gbcon to scipy.linalg.lapack.

• Improved default driver of scipy.linalg.eigh_tridiagonal.

• scipy.linalg.solve can now estimate the reciprocal condition number and the matrix norm calculation is more efficient.

scipy.ndimage improvements

• Added scipy.ndimage.vectorized_filter for generic filters that take advantage of a vectorized Python callable.

• scipy.ndimage.rotate has improved performance, especially on ARM platforms.

scipy.optimize improvements

• COBYLA was updated to use the new Python implementation from the PRIMA package. The PRIMA implementation fixes many bugs in the old Fortran 77 implementation. In addition, it results in fewer function evaluations on average, but it depends on the problem and for some problems it can result in more function evaluations or a less optimal result. For those cases the user can try modifying the initial and final trust region radii given by rhobeg and tol respectively. A larger rhobeg can help the algorithm take bigger steps initially, while a smaller tol can help it continue and find a better solution. For more information, see the PRIMA documentation.

• Several of the minimize methods, and the least_squares function, have been given a workers keyword. This allows parallelization of some calculations via a map-like callable, such as multiprocessing.Pool. These parallelization opportunities typically occur during numerical differentiation. This can greatly speed-up minimization when the objective function is expensive to calculate.

• The lm method of least_squares can now accept 3-point and cs for the jac keyword.

• SLSQP Fortran77 code ported to C. When this method is used now the constraint multipliers are exposed to the user through the multiplier keyword of the returned OptimizeResult object.

• NNLS code has been corrected and rewritten in C to address the performance regression introduced in 1.15.x

• optimize.root now warns for invalid inner parameters when using the newton_krylov method

• The return value of minimization with method='L-BFGS-B' now has a faster hess_inv.todense() implementation. Time complexity has improved from cubic to quadratic.

• optimize.least_squares has a new callback argument that is applicable to trf and dogbox methods. callback may be used to track optimization results at each step or to provide custom conditions for stopping.

scipy.signal improvements

• A new function firwin_2d for the creation of a 2-D FIR Filter using the 1-D window method was added.

• signal.cspline1d_eval and signal.qspline1d_eval now provide an informative error on empty input rather than hitting the recursion limit.

• A new function scipy.signal.closest_STFT_dual_window to calculate the ShortTimeFFT dual window of a given window closest to a desired dual window.

• A new classmethod scipy.signal.ShortTimeFFT.from_win_equals_dual to create a ShortTimeFFT instance where the window and its dual are equal up to a scaling factor. It allows to create short-time Fourier transforms which are unitary mappings.

• The performance of signal.convolve2d has improved.

scipy.sparse improvements

• coo_array supports nD arrays using binary and reduction operations.

• Math is faster between two DIA arrays/matrices: add, sub, multiply, matmul.

• csgraph.dijkstra shortest_path is more efficient.

• csgraph.yen has performance improvements.

• Import of _lib.issparse allows checking for scipy.sparse object without full import of scipy.

• add lazy loading of sparse.csgraph and sparse.linalg.

scipy.spatial improvements

• A new class RigidTransform is available in the transform submodule. It provides functionality to convert between different representations of rigid transforms in 3D space, its application to vectors and transform composition. It follows the same design approach as scipy.spatial.transform.Rotation.

• scipy.spatial.transform.Rotation now has an appropriate __repr__ method.

• Rotation.apply has improved performance.

scipy.stats improvements

• Added scipy.stats.quantile, an array API compatible function for quantile estimation.

• Extended scipy.stats.make_distribution to:

  • work with existing discrete distributions and

  • facilitate the creation of custom distributions

  in the new random variable infrastructure.

• Added scipy.stats.Binomial.

• Added equal_var keyword to:

  • scipy.stats.tukey_hsd (enables the Games-Howell test) and

  • scipy.stats.f_oneway (enables Welch ANOVA).

• Improved moment calculation for scipy.stats.gennorm.

• Added native vectorization to scipy.stats.mode for faster batch calculation.

• Added axis, nan_policy, and keepdims to scipy.stats.power_divergence, scipy.stats.chisquare, scipy.stats.pointbiserialr, scipy.stats.kendalltau, scipy.stats.weightedtau, scipy.stats.theilslopes, scipy.stats.siegelslopes, and scipy.stats.boxcox_llf.

• Improved speed of scipy.stats.special_ortho_group.

• Improved speed of scipy.stats.pearsonr.

Array API Standard Support

Experimental support for array libraries other than NumPy has been added to existing sub-packages in recent versions of SciPy. Please consider testing these features by setting an environment variable SCIPY_ARRAY_API=1 and providing PyTorch, JAX, or CuPy arrays as array arguments. Many functions in scipy.stats, scipy.special, scipy.optimize, and scipy.constants now provide tables documenting compatible array and device types as well as support for lazy arrays and JIT compilation. New features with support and old features with support added for SciPy 1.16.0 include:

• Most features of scipy.signal

• scipy.ndimage.vectorized_filter

• scipy.special.stdtrit

• scipy.special.softmax

• scipy.special.log_softmax

• scipy.stats.quantile

• scipy.stats.gstd

• scipy.stats.rankdata

Features with extended array API support (generally, improved support for JAX and Dask) in SciPy 1.16.0 include:

• many of the scipy.cluster.hierarchy functions

• many functions in scipy.special

• many of the trimmed stats functions

Deprecated features

• The unused atol argument of scipy.optimize.nnls is deprecated and will be removed in SciPy 1.18.0.

• The disp argument of scipy.linalg.signm, scipy.linalg.logm, and scipy.linalg.sqrtm will be removed in SciPy 1.18.0.

• scipy.stats.multinomial now emits a FutureWarning if the rows of p do not sum to 1.0. This condition will produce NaNs beginning in SciPy 1.18.0.

Expired Deprecations

• scipy.sparse.conjtransp has been removed. Use .T.conj() instead.

• The quadrature='trapz' option has been removed from scipy.integrate.quad_vec and scipy.stats.trapz has been removed. Use trapezoid in both instances instead.

• scipy.special.comb and scipy.special.perm now raise when exact=True and arguments are non-integral.

• Support for inference of the two sets of measurements from the single argument x has been removed from scipy.stats.linregress. The data must be specified separately as x and y.

• Support for NumPy masked arrays has been removed from scipy.stats.power_divergence and scipy.stats.chisquare.

Backwards incompatible changes

• Several of the scipy.linalg functions for solving a linear system (e.g. scipy.linalg.solve) documented that the RHS argument must be either 1-D or 2-D but did not always raise an error when the RHS argument had more the two dimensions. Now, many-dimensional right hand sides are treated according to the rules specified in Batched Linear Operations.

• scipy.stats.bootstrap now explicitly broadcasts elements of data to the same shape (ignoring axis) before performing the calculation.

Other changes

• The lm method of least_squares function now has a different behavior for the maximum number of function evaluations, max_nfev. The default for the lm method is changed to 100 * n, for both a callable and a numerically estimated jacobian. This limit on function evaluations excludes those used for any numerical estimation of the Jacobian. Previously the default when using an estimated jacobian was 100 * n * (n + 1), because the method included evaluations used in the estimation. In addition, for the lm method the number of function calls used in Jacobian approximation is no longer included in OptimizeResult.nfev. This brings the behavior of lm, trf, and dogbox into line.

• For Cython>=3.1.0b1, SciPy now uses the new cython --generate-shared functionality, which reduces the total size of SciPy’s wheels and on-disk installations significantly.

• The output of scipy.stats.wrapcauchy rvs is now mapped to the unit circle between 0 and 2 * pi.

• The vendored Qhull library was upgraded from version 2019.1 to 2020.2.

• A new build option -Duse-system-libraries has been added. It allows opting in to using system libraries instead of using vendored sources. Currently Boost.Math and Qhull are supported as system build dependencies.

• The namedtuple-like bunch objects returned by some SciPy functions now have improved compatibility with the polars library.

• The testsuite thread safety with free-threaded CPython has improved.

Authors

• Name (commits)

• h-vetinari (4)

• aiudirog (1) +

• Anton Akhmerov (2)

• Thorsten Alteholz (1) +

• Gabriel Augusto (1) +

• Backfisch263 (1) +

• Nickolai Belakovski (5)

• Peter Bell (1)

• Benoît W. (1) +

• Maxwell Bileschi (1) +

• Sam Birch (1) +

• Florian Bourgey (3) +

• Charles Bousseau (2) +

• Richard Strong Bowen (2) +

• Jake Bowhay (126)

• Matthew Brett (1)

• Dietrich Brunn (52)

• Evgeni Burovski (252)

• Christine P. Chai (12) +

• Saransh Chopra (2) +

• Omer Cohen (1) +

• Lucas Colley (91)

• crusaderky (56) +

• Yahya Darman (3) +

• dartvader316 (2) +

• Benjamin Eisele (1) +

• Donnie Erb (1)

• Evandro (1)

• Sagi Ezri (57) +

• Alexander Fabisch (2) +

• Matthew H Flamm (1)

• Gautzilla (1) +

• Neil Girdhar (1)

• Ralf Gommers (149)

• Rohit Goswami (4)

• Saarthak Gupta (4) +

• Matt Haberland (320)

• Sasha Hafner (1) +

• Joren Hammudoglu (9)

• Chengyu Han (1) +

• Charles Harris (1)

• Kim Hsieh (4) +

• Lukas Huber (1) +

• Guido Imperiale (47) +

• Jigyasu (1) +

• karthik-ganti-2025 (1) +

• Robert Kern (2)

• Harin Khakhi (2) +

• Agriya Khetarpal (4)

• Tetsuo Koyama (1)

• David Kun (1) +

• Eric Larson (3)

• lciti (1)

• Antony Lee (1)

• Kieran Leschinski (1) +

• Thomas Li (2) +

• Christian Lorentzen (2)

• Loïc Estève (4)

• Panos Mavrogiorgos (1) +

• Nikolay Mayorov (2)

• Melissa Weber Mendonça (10)

• Miguel Cárdenas (2) +

• MikhailRyazanov (6) +

• Swastik Mishra (1) +

• Sturla Molden (2)

• Andreas Nazlidis (1) +

• Andrew Nelson (209)

• Parth Nobel (1) +

• Nick ODell (9)

• Giacomo Petrillo (1)

• pmav99 (1) +

• Ilhan Polat (72)

• pratham-mcw (3) +

• Tyler Reddy (73)

• redpinecube (1) +

• Érico Nogueira Rolim (1) +

• Pamphile Roy (10)

• sagi-ezri (1) +

• Atsushi Sakai (9)

• Marco Salathe (1) +

• sanvi (1) +

• Neil Schemenauer (2) +

• Daniel Schmitz (20)

• Martin Schuck (1) +

• Dan Schult (32)

• Tomer Sery (19)

• Adrian Seyboldt (1) +

• Scott Shambaugh (4)

• ShannonS00 (1) +

• sildater (3) +

• PARAM SINGH (1) +

• G Sreeja (7) +

• Albert Steppi (133)

• Kai Striega (3)

• Anushka Suyal (2)

• Julia Tatz (1) +

• Tearyt (1) +

• Elia Tomasi (1) +

• Jamie Townsend (2) +

• Edgar Andrés Margffoy Tuay (4)

• Matthias Urlichs (1) +

• Jacob Vanderplas (2)

• David Varela (2) +

• Christian Veenhuis (3)

• vfdev (1)

• vpecanins (10) +

• vrossum (1) +

• Stefan van der Walt (2)

• Warren Weckesser (5)

• Jason N. White (1) +

• windows-server-2003 (5)

• Zhiqing Xiao (1)

• Pavadol Yamsiri (1)

• YongcaiHuang (2) +

• Rory Yorke (3)

• yuzie007 (2) +

• Irwin Zaid (4)

• zaikunzhang (1) +

• Austin Zhang (1) +

• William Zijie Zhang (1) +

• Eric Zitong Zhou (5) +

• zitongzhoueric (6) +

• Case Zumbrum (2) +

• ਗਗਨਦੀਪ ਸਿੰਘ (Gagandeep Singh) (45)

  A total of 124 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.

Issues closed for 1.16.0

• #4800: ENH: ndimage.median_filter: behavior with NaNs

• #4878: ENH: ndimage.median_filter: excessive memory usage

• #5137: ENH: ndimage.generic_filter: function to return higher-dimensional…

• #5435: savemat silently drops entries starting with “_”

• #5451: ENH: linalg.solve: support broadcasting

• #6052: savemat does not save keys starting with underscore

• #6606: BUG: signal.bilinear: can’t handle leading zeros

• #6689: ENH: optimize: consider using NLopt’s version of slsqp

• #6755: ENH: ndimage.percentile_filter: take multiple percentiles

• #7518: DOC: optimize: meaning of accuracy in fmin_slsqp undocumented

• #7818: ENH: ndimage.uniform_filter: expands NaNs all the way to the…

• #8140: sparse LU decomposition does not solve with complex right-hand…

• #8367: ENH: stats.mvndst: make thread-safe

• #8411: nan with betainc for a=0, b=3 and x=0.5

• #8916: ENH: ndimage.generic_filter: slow on large images

• #9077: maximum_filter is not symetrical with nans

• #9841: ENH: linalg: 0-th dimension must be fixed to 1 but got 2 (real…

• #9873: ENH: ndimage: majority voting filter

• #10416: ENH: optimize.minimize: slsqp: give better error when work array…

• #10793: BUG: integrate: solve_ivp and odeint with lsoda have…

• #11312: BUG: signal.cont2discrete not handling lti instances as documented

• #11328: Scipy unable to read piped wav file

• #12133: How to define new distributions?

• #12544: signal.spectral._triage_segments doesn’t support window as tuple…

• #12994: ENH: linalg.sqrtm: efficiently process upper triangular matrices

• #13577: Split scipy.signal.spectral._spectral_helper into two to support…

• #13666: ENH: invgauss.pdf should return correct output when mu=infinity

• #13788: Documentation for scipy.signal.resample should say what to use…

• #13789: Documentation for scipy.signal.decimate doesn’t say what to use…

• #13823: BUG: signal.bilinear: doesn’t work for complex valued arrays

• #13914: DOC: sparse.csgraph.shortest_path: predecessors array contains…

• #13952: fmin_cobyla result violates constraint

• #13982: ENH: linalg.eigh_tridiagonal: divide and conquer option

• #14394: ENH: optmize.slsqp: return Lagrange multipliers

• #14569: BUG: signal.resample: inconsistency across dtypes

• #14915: BUG: optimize.minimize: corruption/segfault with constraints

• #15153: BUG: signal.resample: incorrect with datetime[ns] for t…

• #15527: BUG: optimize: COBYLA hangs on some CPUs

• #16009: BUG: act fails for local GitHub Actions CI run

• #16142: ENH: Fix the random state in scipy.stats.multivariate_normal.cdf()

• #16203: BUG: scipy.io.savemat discards nested names with a leading digit

• #16234: BUG: Memory leak in _superluobject.c when ENUM_CHECK is not…

• #16452: doit based dev interface garbles pdb command history (in some…

• #17546: ENH: Adding ‘valid’ mode to ndimage.generic_filter

• #17787: BUG: Erratic results from RectBivariateSpline when smoothing…

• #17891: BUG: inconsistent checks for integrality in several distributions

• #17968: ENH: creation of a 2-D FIR Filter using 1-D window method

• #18046: BUG: dev.py does not work in a Windows CI environment on GHA…

• #18105: ENH: optimize LbfgsInvHessProduct.todense(), 10x speed…

• #18118: ENH: The Fortran 77 implementation of COBYLA is buggy and challenging…

• #18214: DOC: inconsistent definitions of “OP” and “OPinv” in eigsh

• #18346: DOC: optimize: l_bfgs_b: sets maxiter and maxfun to the…

• #18437: ENH: ndimage.generic_filter: support complex input

• #18740: BUG: scipy.optimize.bisect gives incorrect results for very small…

• #18866: MAINT: follow-up actions for array API support in cluster

• #18951: ENH: improve python dev.py test experience caused by imp…

• #18998: BUG: dev.py has issues with site-packages and Python installed…

• #19254: ENH: spatial.transform: cover proper rigid transformations with…

• #19362: BUG: optimize: warning generated by SLSQP is useless

• #19415: BUG: linalg.sqrtm results different between version 1.11.1 and…

• #19459: BUG: optimize.least_squares giving poor result compared to optimize.leastsq…

• #20219: BUG: failing sqrtm regression test

• #20366: ENH: Yens algorithm improvements and enhancements

• #20608: BUG: refguide-check incorrectly flags references to equations…

• #20622: DOC: signal: add an example cross-spectrogram application

• #20806: Failures for new pytest-fail-slow check in Windows CI jobs

• #20972: BUG: special.chdtrc: returns 1.0 when both degrees of freedom…

• #20999: BUG: ndimage.zoom: wrong output with zoom factor of 1

• #21020: DOC: signal: Use where='post' when plotting discrete response

• #21095: DOC: RegularGridInterpolator uses half down rounding instead…

• #21102: RFC/ENH?: optimize.curve_fit: option to use global optimization…

• #21293: DOC: stats.qmc.discrepancy: clarify deviation from reference

• #21317: BUG: special.gammainc: returns finite results with NaN…

• #21323: DOC: build fails with Sphinx 8

• #21341: DOC: signal.correlate: formula doesn’t match behavior when x…

• #21484: DEP: optimize.nnls: deprecate atol parameter which does nothing

• #21531: MAINT: stats.dirichlet_multinomial: relax n to >=0

• #21547: STY/DEV: fix and enable lint rule UP038

• #21606: ENH: stats: generic power law with negative index

• #21649: RFC: Splitting off special function scalar kernels into separate…

• #21692: BUG: optimize.shgo: not working with jac=True

• #21717: DOC: assert_allclose instead of xp_assert_close is recommended…

• #21740: CI: adding a GPU-enabled CI job

• #21764: ENH: linalg.lapack: add symmetric solvers

• #21844: ENH: linalg: wrap ?gbcon/?langb and use in linalg.solve

• #21879: BUG: scipy.datasets failing with Error 403 for readthedocs…

• #21971: ENH: ndimage.median_filter: extended dtype support?

• #21972: STY: fix and enable lint rule UP031

• #21986: ENH: optimize.root: warn when inner parameters are ignored with…

• #21995: BUG: optimize.curve_fit with method='lm' fails to determine…

• #21999: ENH: io.mmread: Provide better error message when loading…

• #22000: DOC: ndimage.median_filter: document behaviour with nan...

• #22011: BUG: interpolate.Akima1DInterpolator: different values on subsequent…

• #22044: TST: optimize.elementwise.bracket_minimum: CuPy failure

• #22045: DOC: stats: clarify the support of a distribution is unaffected…

• #22051: BUG: AttributeError: module ‘numpy’ has no attribute ‘AxisError’…

• #22054: BUG: ndimage, array types: minimum_position and extrema…

• #22055: DOC: ndimage.minimum and maximum: incorrect return type

• #22057: DOC: stats.order_statistic: docstring missing the “Returns”…

• #22065: DOC: sparse: Several functions are missing the ‘Returns’ section…

• #22072: DOC: PchipInterpolator: missing integrate function

• #22086: MAINT: signal: build warning (sprintf) on macOS

• #22093: DOC: integrate.quad: uses Gauss-Kronrod not Curtis-Clenshaw?

• #22136: DOC: linalg.matrix_balance: equation does not render

• #22144: Query: optimize.minimize: trust_constr does not avoid Nonlinear…

• #22163: DOC: update scipy module docstring for lazy loading

• #22164: MAINT: undo ignored errors in mypy

• #22195: Query: optimize.basinhopping: lowest minimum not accepted if…

• #22224: MAINT: remove end year from copyright

• #22252: MAINT: Fix a dtype check in scipy.signal._waveforms.py

• #22258: BUG: Constructing sparse matrix with big-endian float32/64 raises…

• #22263: BUG: linalg.solve doesn’t raise an error when A is a singular…

• #22265: BUG: linalg: hecon returns NaN incorrectly with some lower…

• #22271: Query: empty Rotation is not allowed in scipy=1.15

• #22282: QUERY/DEV: test failure in IDE with SCIPY_ARRAY_API

• #22288: QUERY: Pyright raises error/warning in IDE

• #22303: ENH: stats.special_ortho_group: improve and simplify

• #22309: DOC: optimize.elementwise.find_minimum: harmonize documented/implemented…

• #22328: QUERY: stats.beta.fit: FitError on reasonable data

• #22338: QUERY: Intellisense Autocomplete Not Working for spatial.transform.Rotation

• #22361: BUG: interpolation test TestSmoothingSpline.test_compare_with_GCVSPL…

• #22363: BUG: special test TestHyp2f1.test_region3[hyp2f1_test_case23]…

• #22367: QUERY/TYP: sparse: Pylance reports unreachable after toarray()

• #22378: DOC/TST: interpolate, signal: smoke-docs failures

• #22382: ENH: sparse.spmatrix: allow fast import

• #22395: BUG: special: failure of TestSystematic.test_besselj_complex…

• #22403: DOC: gaussian_kde‘s bw_method='silverman' deviates…

• #22415: Two TestBatch failures in macOS x86-64 Accelerate wheel build…

• #22429: DOC: integrate: missing bold font for a vector in tutorial

• #22437: DOC: The code of conduct link is dead

• #22449: BUG: sparse.csgraph.construct_dist_matrix: buffer dtype mismatch

• #22450: QUERY: difference between namedtuples and objects produced…

• #22461: DOC: freqz_sos: claims that it was introduced in 0.19; no mention…

• #22470: BUG: lfiltic‘s handling of a[0] != 1 differs from lfilter...

• #22485: DOC: remove links to the reference guide in the tutorials page

• #22488: DOC: interpolate.lagrange: the Lagrange function is using the…

• #22495: BUG: special test TestHyp2f1.test_region4[hyp2f1_test_case42]…

• #22501: BUG: min_weight_full_bipartite_matching fails for coo_matrix…

• #22508: DOC: Inconsistent notation in Linear algebra (scipy.linalg) page

• #22534: CI: failures */tests/test_extending due to a regression in…

• #22559: BUG: ndimage: Numerical regressions in Dask 2025.2.0

• #22565: BUG: stats.multinomial.pmf: inconsistent results?

• #22581: DOC: stats.gaussian_kde: clarify the meaning of factor

• #22591: BUG: sparse.coo: ImportError for upcast

• #22601: BUG: special.logsumexp: inconsistency in phase when one element…

• #22626: BUG: scipy.stats: tmin/tmax: loss of precision for large integers

• #22646: CI/DOC: CloughTocher2DInterpolator: UserWarning in docs build

• #22659: BUG: spatial: RigidTransform does not support zero-length…

• #22692: DOC: interpolate.make_smoothing_spline: example plot uses the…

• #22700: CI: new failures: segfault in free-threaded, linprog invalid…

• #22703: DOC: integrate: quad_vec info return type is _Bunch not…

• #22767: BUG: test_cython Failing on Windows on ARM64 with clang-cl

• #22768: DOC/DEV: outdated references to Cirrus CI

• #22769: ENH: optimize: Return bound multiplier for SLSQP

• #22775: ENH: Use cython shared utility module

• #22791: BUG: optimize.nnls: unstable on i686 (32-bit) machine

• #22800: BUG: signal.windows.kaiser_bessel_derived uses array…

• #22881: DOC: Update minimum NumPy and Python in toolchain roadmap

• #22904: BUG: Wrong use of __builtin_prefetch()

• #22912: BUG: optimize: SyntaxWarning: 'break' in a 'finally' block…

• #22920: BUG: check_test_name fails with UnicodeDecodeError?

• #22921: DOC: clarify the status of Apple’s Accelerate Framework support

• #22931: BUG: interpolate._dierckx: check_array() can crash if the…

• #22942: TST: special: test_compiles_in_cupy is broken

• #22945: TST: Nested arrays failing in array-api-strict git tip

• #22951: BUG: stats.wrapcauchy: output isn’t wrapped around the unit circle

• #22956: BUG: special._ufuncs._ncx2_pdf: interpreter crash with extreme…

• #22965: BUG: The attribute “nit” is not found when using the callback…

• #22981: Bug with freqz when specifying worN after #22886

Pull requests for 1.16.0

• #18375: ENH: signal: Add firwin_2d filter

• #20610: ENH: signal.ShortTimeFFT: determine arbitrary dual windows

• #20639: ENH: stats.rankdata: add array API standard support

• #20717: ENH: Speed up sparse.csgraph.dijkstra 2.0

• #20772: ENH: array types, signal: delegate to CuPy and JAX for correlations…

• #20950: ENH: spatial: speed up Rotation.apply by replacing np.einsum…

• #21180: ENH: sparse: efficient arithmetic operations for DIA format

• #21233: ENH: stats.boxcox_llf: vectorize for n-D arrays

• #21270: MAINT: make boost_math a subproject

• #21462: ENH: linalg.eig: support batched input

• #21482: MAINT/DEV: use Sphinx 8 for documentation builds

• #21557: ENH: stats._continued_fraction: elementwise, Array API…

• #21628: BUG:signal: Fix passing lti as system to cont2discrete

• #21674: DEV: use spin

• #21684: MAINT: stats.dirichlet_multinomial relax n to >= 0

• #21713: ENH: signal: add array API support / delegation to lfilter et…

• #21783: ENH: signal.windows: add array API support (take 2)

• #21863: CI: use macos-15 for a macOS run

• #21987: STY: fix lint rule UP031

• #22008: ENH: signal.vectorstrength: add array API standard support

• #22010: REL: set version to 1.16.0.dev0

• #22012: MAINT: bump min NumPy to 1.25.2, min Python to 3.11

• #22013: DEV: gh_lists: fix asterisk sanitisation

• #22015: DEV: lint: add option to lint all files

• #22019: MAINT: signal: remove tempita templating

• #22042: DOC, MAINT: Add a "jupyterlite_sphinx_strip" tag to the scipy.stats…

• #22046: TST: optimize: fix CuPy failure for bracket_minimum

• #22052: DOC: sparse.linalg: add note about complex matrices to splu…

• #22056: MAINT: stats.wilcoxon: fix attempt to access np.AxisError

• #22061: BUG: ndimage: convert array scalars on return

• #22062: MAINT: _lib: co-vendor array-api-extra and array-api-compat

• #22064: MAINT: sparse.linalg._isolve: Remove postprocess function

• #22068: ENH: optimize: migrate to use sparray

• #22070: ENH: _lib: JAX support (non-jitted)

• #22071: MAINT: Use ENUM_CHECK_NAME for avoiding memory leaks in _superluobject.c

• #22073: DEP: sparse: remove conjtransp

• #22074: DEP: remove remaining trapz references

• #22075: DEP: stats.linregress: remove one arg use

• #22076: BUG: datasets: add headers to fetchers to avoid 403 errors

• #22079: DEP: stats: remove support for masked arrays from power_divergence…

• #22087: DEP: special: raise error for non-integer types with exact=True…

• #22088: TST: optimize.elementwise.find_root: refactor tests to use find_root…

• #22089: TST: optimize: suppress incorrect sparray warning from scikit-sparse

• #22090: ENH: optimize: migrate to sparray (docs)

• #22092: MAINT: signal: fixed build warning (sprintf) on MacOS

• #22100: DEP: signal.spline: use standard submodule deprecation machinery

• #22101: DOC: update stats, integrate, optimize, and…

• #22108: CI: Run ‘Checkout scipy’ and ‘Check for skips’ only on Github…

• #22110: TST: linalg: use infinity norm of matrix when norm=’I’

• #22115: DOC: release notes: ensure TOC links to headings below

• #22116: DOC: update the interpolate roadmap

• #22122: MAINT: signal.oaconvolve: avoid xp <-> numpy conversions

• #22125: TST: stats: ensure tests are thread-safe

• #22127: ENH: linalg: add batch support for matrix -> scalar funcs

• #22130: TST: ndimage: array API-related cosmetic tweaks in tests

• #22131: TST: skip|xfail_xp_backends disregards reason=

• #22132: TST: array types: enforce namespace in tests

• #22133: ENH: linalg: add batch support for functions that accept a single…

• #22140: DOC: linalg.matrix_balance: move math to notes; ensure that it…

• #22142: ENH: signal: add CuPy/JAX delegation to scipy.signal

• #22148: TST: ndimage: fix test skip typo

• #22152: ENH: stats.f_oneway: add equal_var for Welch ANOVA

• #22154: ENH: linalg.clarkson_woodruff_transform: add batch support

• #22155: ENH: stats: add axis/nan_policy/keepdims/etc. support to correlation…

• #22157: ENH: linalg: add batch support for remaining cholesky functions

• #22160: DEP: interpolate: remove incidental imports from private modules

• #22161: DOC, MAINT: Add updates for interactive notebooks via jupyterlite-sphinx…

• #22165: ENH: linalg: add batch support to remaining eigenvalue functions

• #22166: ENH: linalg.block_diag: add batch support

• #22169: MAINT: sparse: refactor CSC to use CSR sparsetools

• #22170: ENH: signal: convert symiirorder and related filters to work…

• #22172: MAINT: improve overflow handling in factorial functions

• #22173: DOC: interpolate: add missing method integrate for PchipInterpolator

• #22174: MAINT: optimize: switch suppress_warnings to catch_warnings

• #22176: MAINT: special: Move Faddeeva into xsf

• #22179: DOC/DEV: mention scipy-stubs in building from source guide

• #22182: TST: ndimage: cupy tweaks for inplace out=

• #22185: ENH: stats.tukey_hsd: equal_var=False option to perform Games-Howell…

• #22186: DOC: interpolate: add a note about rounding rule of the nearest…

• #22190: MAINT: special: Migrate remaining exp and log functions to xsf

• #22192: ENH: linalg: add batch support to linear system solvers

• #22196: DOC: update scipy module docstring for lazy loading

• #22197: ENH: linalg.cossin: add batch support

• #22198: DOC: basinhopping, clarify when lowest_optimization_result is…

• #22201: DOC: Clarify support behavior in rv_continuous documentation

• #22208: ENH: io.wavfile: read unseekable files

• #22211: DOC: interpolate: add missed integrate doc link for Akima1DInterpolator

• #22212: ENH: linalg: wrap ?gbcon

• #22213: BUG: zpk2tf works correctly with complex k, real p, z

• #22214: TST: make torch default dtype configurable

• #22215: ENH: io: throw FileNotFoundError exception when the source…

• #22216: TST: TestBracketMinimum MPS shims

• #22217: ENH: linalg: wrap ?langb

• #22219: ENH: _lib: deobfuscate jax.jit crash in _asarray

• #22220: MAINT: stats: replace nonstandard calls in (mostly) array API…

• #22221: MAINT: linalg.leslie: use _apply_over_batch

• #22222: ENH: special/stats: implement xp-compatible stdtrit…

• #22226: ENH: signal.upfirdn: array API standard support

• #22227: TST: linalg: add missing lower arguments in test_sy_hetrs

• #22228: ENH: linalg.lapack: wrap ?sytri and ?hetri

• #22229: MAINT: cluster: remove unnecessary namespace changes

• #22231: ENH: add callback to optimize.least_squares

• #22234: MAINT: forward port 1.15.0 relnotes

• #22237: BENCH: sparse.csgraph.dijkstra: add benchmark

• #22240: ENH: array types: add dask.array support

• #22242: MAINT: integrate.cubature: fix undefined asarray use

• #22243: DOC: sparse: docstring example of random_array with uint32 data_sampler

• #22251: ENH: linalg.solve: use langb

• #22255: EHN: cluster: JAX support (non-jitted)

• #22256: ENH: special: JAX support (non-jitted)

• #22259: TST: signal: fix symiir tests

• #22260: TST: Make @pytest.mark.usefixtures("skip_xp_backends") redundant

• #22261: TST: dev.py quietly ignores user markers

• #22262: TST: Mark with xp all tests in Array API-compatible modules

• #22264: MAINT: interpolate: make BSpline allocate out arrays in C

• #22266: MAINT: linalg.solve: raise when diagonal matrix is exactly singular

• #22267: ENH: spatial.transform: baseline implementation of RigidTransform

• #22268: TST: clean up obsolete Array API fixtures

• #22269: DOC: optimize.curve_fit: add note about more advanced curve fitting

• #22273: ENH: linalg.solve: use gbcon

• #22274: ENH: _contains_nan for lazy arrays

• #22275: CI: add a GPU CI job

• #22278: BUG: Fix Akima1DInterpolator by returning linear interpolant…

• #22279: TST: Add skips for GPU CI failures

• #22280: TST: _lib: more idiomatic conditional skips

• #22281: TST: special: better skip message for stdtrit on JAX

• #22283: BUG: Fix banded Jacobian for lsoda: ode and solve_ivp

• #22284: BUG: sparse: better error message for unsupported dtypes

• #22289: CI: fix skip/trigger condition of GPU CI job

• #22293: ENH: Add __repr__ method to scipy.spatial.transform.Rotation

• #22295: DOC: signal.ShortTimeFFT.nearest_k_p: fix typo

• #22298: MAINT: stats: remove mvn fortran calls from multivariate_normal.cdf

• #22300: MAINT: remove end year from copyright

• #22302: MAINT: remove unused library import

• #22304: ENH: stats.special_ortho_group: speed up, allow 1x1 and 0x0 ortho…

• #22305: MAINT, DOC: forward port 1.15.1 relnotes

• #22308: TST: _lib: run tests with @jax.jit

• #22311: TST: replace pytest.xfail with skip/xfail_xp_backends

• #22312: ENH: stats.Binomial: add binomial distribution with new infrastructure

• #22313: BUG: signal.bilinear handles complex input, and strips leading…

• #22320: TST: array types: wrap namespaces centrally

• #22324: ENH: io: add invalid field name warning for savemat

• #22330: ENH: sparse.csgraph.yen: performance improvements

• #22340: MAINT: linalg: reorganize tridiagonal eigenvalue routines

• #22342: ENH: cluster: linkage support for jax.jit and dask

• #22343: ENH: signal.{envelope,resample,resample_poly}: array API…

• #22344: BUG: Fix bug with dpss degenerate case

• #22348: DOC: Harmonize summary line of docstrings of iterative sparse…

• #22350: ENH: Replace Fortran COBYLA with Python version from PRIMA

• #22351: DOC: sparse.linalg.eigsh: fix inconsistent definitions of OP…

• #22352: ENH: stats.quantile: add array API compatible quantile function

• #22358: MAINT: special.nctdtrit: migrate to boost

• #22359: MAINT: remove temporary # type: ignore‘s from #22162

• #22364: TST: bump tolerance on TestHyp2f1.test_region3[hyp2f1_test_case23]

• #22366: DOC: integrate: fix quad documentation to correctly describe…

• #22371: ENH: stats.make_distribution: allow definition of custom distributions

• #22375: DOC: sparse.linalg: fix doctest in scipy.sparse.linalg._norm.py

• #22376: DOC: sparse.linalg: sparray updates in doc_strings and Sakurai…

• #22379: DOC: interpolate.AAA: add may vary to example

• #22380: DOC: Replace link to X in header with link to scientific python…

• #22381: MAINT: special: A bit of clean up in stirling2.h

• #22386: DEP: optimize.nnls: deprecate unused atol parameter

• #22387: DOC: Add example to show usage of predecessors matrix returned…

• #22388: DOC: Fix documentation for predecessors matrix in shortest_path...

• #22389: DOC: Add “Assert function selection guideline” doc in the new…

• #22393: TST: stats: test support for array API compatible masked arrays

• #22396: DOC: signal: Use where=’post’ when plotting discrete response…

• #22397: DOC: spatial: Added mention of Davenport Angles to Rotation class…

• #22398: MAINT: special: clean up os/warnings modules exposed in special…

• #22399: TST: remove thread-unsafe skips for a now fixed Cython fused…

• #22401: TYP: Runtime-subscriptable sparray and spmatrix types

• #22406: ENH: linalg: Rewrite sqrtm in C with low-level nD support

• #22407: MAINT: remove _lib->``sparse`` dependency

• #22411: DOC: stats.gaussian_kde: clarify Silverman method

• #22413: DOC: stats: Edited the NIST Handbook reference

• #22416: TST: linalg: bump tolerances in two TestBatch tests

• #22419: MAINT: special: Remove libsf_error_state shared library in…

• #22420: TST: use singular reason= in skip_xp_backends

• #22421: BUG: ndimage: binary_erosion vs. broadcasted input

• #22422: MAINT: _lib: adapt array_namespace to accept scalars…

• #22425: MAINT: special: Update handling of betainc and betaincc…

• #22426: ENH: linalg: wrap ?stevd

• #22427: DEP: linalg: deprecate disp argument for signm, logm, sqrtm

• #22428: DOC: add note on getting the version switcher to behave to release…

• #22430: MAINT: cluster: vectorize tests in is_valid_linkage

• #22431: DOC: integrate: correct tutorial formatting

• #22433: BUG: interpolate.RectBivariateSpline: fix NaN output when…

• #22434: DOC: integrate.tanhsinh: remove incorrect reference to _differentiate

• #22435: MAINT: bump to array-api-extra git tip

• #22439: MAINT: special: Add log1mexp for log(1 - exp(x))

• #22440: DOC: Fix year of publication in _dual_annealing.py

• #22441: BUG: special: Fix incorrect handling of nan input in gammainc…

• #22442: DOC: Modified Link for code of conduct documentation

• #22443: DOC: Corrected Path

• #22445: CI: avoid mpmath pre-release version that’s failing in CI

• #22448: DOC: optimize.elementwise.find_minimum: fix documented termination…

• #22452: ENH: linalg.eigh_tridiagonal: add stevd as a driver and make…

• #22453: DOC: Improve docstrs of dlsim, dimpulse, dstep...

• #22454: BUG: signal.ShortTimeFFT: make attributes win and dual_win…

• #22455: ENH: stats.gstd: add array API support

• #22456: ENH: stats: add nan_policy support to power_divergence, chisquare

• #22457: TST: sparse: add tests for subscriptable types

• #22459: DOC: ndimage: fix wrong return type doc for ndimage.minimum…

• #22460: MAINT: signal.csd: port away from using _spectral_helper

• #22462: ENH: stats.pearsonr: two simple (but substantial) efficiency…

• #22463: DOC: update Halton docs

• #22464: DOC: Prevent A@x=b from becoming a URL

• #22467: MAINT/TST: address nits from Dask PR

• #22469: TST: stats: improve JAX test coverage

• #22475: BUG: optimize.shgo: delegate options['jac'] to minimizer_kwargs['jac']

• #22478: ENH: optimize: add workers kwarg to BFGS, SLSQP, trust-constr

• #22480: CI: use mpmath pre-release again

• #22481: BUG: fix make_lsq_spline with a non-default axis

• #22483: MAINT: spatial: missing Cython type in build

• #22484: ENH: allow batching in make_smoothing_spline

• #22489: MAINT: simplifications related to NumPy bounds

• #22490: ENH: stats: add marray support to most remaining array API…

• #22491: DOC: stats: resampling tutorial fixups

• #22493: DOC: Add a docstring to OptimizeWarning

• #22494: ENH: _lib._make_tuple_bunch: pretend to be namedtuple even more

• #22496: MAINT: stats.invgauss: return correct result when mu=inf

• #22498: TST: bump tolerance in TestHyp2f1.test_region4[hyp2f1_test_case42]

• #22499: DOC: remove links to the reference guide in the tutorials page

• #22504: BLD: bump min version of Clang to 15.0, and macOS min version…

• #22505: ENH: stats.quantile: add discontinuous (HF 1-3) and Harrell-Davis…

• #22507: BENCH: make Benchmark.change_dimensionality a class variable

• #22509: DOC: sparse.linalg: add explanation for MatrixRankWarning

• #22511: BUG: sparse.csgraph: Added support for casting coo array to csc/csr…

• #22514: TST: special: Add edgecase tests for gammainc and friends

• #22516: STY: enable lint rule UP038 and fix instances in violation of…

• #22518: DOC: interpolate.FloaterHormannInterpolator: fix typos

• #22519: ENH: add workers to least_squares

• #22520: MAINT: Remove an extraneous dtype check in scipy/signal/_waveforms.py

• #22524: ENH:MAINT:optimize: Rewrite SLSQP and NNLS in C

• #22526: DOC: interpolate: reorganize the API listing

• #22527: DOC: sparse: add returns sections to some _construct.py functions

• #22528: DOC: interpolate: improve visibility of univariate interpolator…

• #22529: DOC: Update a link in SciPy Core Developer Guide

• #22530: DOC: interpolate: improve one-line descriptions

• #22531: DOC: batching in 1D/ND interpolation/smoothing routines

• #22535: DOC: update roadmap sparse

• #22536: DOC: io: link to netcdf4-python

• #22537: DOC: linalg: fix inconsistent notation

• #22541: Interpolate tutorial: discuss the bases and interconversions

• #22542: MAINT, DOC: forward port 1.15.2 release notes

• #22546: DOC: Add docstring for QhullError in _qhull.pyx [docs only]

• #22548: DOC: interpolate.lagrange: add notes / references; recommend…

• #22549: ENH: use workers keyword in optimize._differentiable_functions.VectorFunct…

• #22552: MAINT: sparse.csgraph: Raise error if predecessors.dtype !=…

• #22554: BUG: lfiltic‘s handling of a[0] != 1 differs from lfilter...

• #22556: ENH: optimize: speed up LbfgsInvHessProduct.todense on large…

• #22557: ENH: Replace _lazywhere with xpx.apply_where

• #22560: ENH: Allow endpoints of custom distributions created with stats.make_distribut…

• #22562: DOC: Correct a typo: MATLAB(R) -> MATLAB®

• #22564: TST: add missing custom markers to pytest.ini

• #22566: TST: skip_xp_backends(eager_only=True)

• #22569: CI: fix dev-deps job by not testing Meson master

• #22572: TST: skip two ndimage tests that are failing for Dask

• #22573: DOC: sparse: Add docstrings to warnings in scipy.sparse

• #22575: ENH: ndimage.vectorized_filter: generic_filter with…

• #22579: DOC: signal.correlate: improve notes section

• #22584: TST: ndimage: tidy skip_xp_backends

• #22585: MAINT: stats.multinomial: FutureWarning about normalization…

• #22593: TST: add one more missing custom marker (fail_slow) to…

• #22597: ENH: stats.make_distribution: improve interface for overriding…

• #22598: MAINT: stats.bootstrap: broadcast like other stats functions

• #22602: DOC: stats.pearsonr: add tutorial

• #22603: MAINT: _lib: bump version array_api_compat to 1.11

• #22605: MAINT: signal: clean up unnecessary shims

• #22606: DOC: Ignore dict subclass docstring warning

• #22607: MAINT: special.logsumexp: improve behavior with complex infinities

• #22609: ENH: stats: shared array api support information to generate…

• #22610: ENH: _lib.doccer: Simplify and optimize indentation loop

• #22611: MAINT: stats: rewrite gaussian_kde.integrate_box, remove…

• #22614: MAINT: linalg: fix cython lint failures in build output

• #22616: ENH: stats: use vecdot and nonzero where appropriate

• #22618: BUG: Fix dual quaternion normalization procedure

• #22619: DOC: stats.gaussian_kde: clarify the meaning of factor

• #22621: MAINT: sparse: remove incidental imports from private modules

• #22623: ENH: signal.convolve2d: Performance Enhancement on WoA

• #22624: BUG: stats: kde.integrate_box was missing an rng parameter

• #22625: MAINT: Bump array-api-compat and array-api-strict

• #22628: MAINT: stats.tmin/tmax: ensure exact results with unreasonably…

• #22630: MAINT: stats: tmin/tmax tweaks

• #22631: DOC: interpolate.BarycentricInterpolator: documentation improvements

• #22632: MAINT: stats.multinomial: use dtype-dependent tolerance

• #22633: ENH: special: softmax / log_softmax Array API support

• #22634: TST: special: cosmetic nits

• #22636: MAINT: fix domain check for ncfdtri

• #22639: ENH: special: support_alternative_backends on Dask and jax.jit

• #22641: ENH: special: add Dask support to rel_entr

• #22645: DOC: stats.special_ortho_group: update algorithm description

• #22647: MAINT: sparse: rewrite sparse._sputils.validateaxis to centralize…

• #22648: MAINT: stats.quantile: fixup quantile for p < minimum plotting…

• #22649: DOC, CI: Fix legend warning for CloughTocher2DInterpolator docstring

• #22650: TST: stats: mark nct fit xslow

• #22651: MAINT: ndimage.zoom: eliminate noise when zoom=1

• #22653: DOC: add COBYQA to local optimizer comparison table

• #22658: CI: clean up free-threading job, add new job using pytest-run-parallel

• #22661: TST: fix some test failures and excessive memory use on Guix

• #22666: MAINT: interpolate: move NdBSpline evaluations to C

• #22667: DEV: cap Sphinx version in environment.yml

• #22668: DOC: document Array API support for the constants module and…

• #22669: TST: constants: tidy up tests

• #22671: MAINT: enforce modularity with tach

• #22675: ENH: stats: Improvements to support/domain endpoints in custom…

• #22676: ENH: stats.mode: vectorize implementation

• #22677: MAINT: use function handles rather than custom strings in xp_capabilities_tabl…

• #22683: MAINT: remove outdated xp_ functions, xp.asarray on elementwise…

• #22686: TST/DOC: lazy_xp_backends in xp_capabilities

• #22687: MAINT: Bump Array API to 2024.12

• #22691: DOC: signal: fix freqz_sos and sosfreqz docstrings

• #22694: DOC: interpolate.make_smoothing_spline: improve example visibility

• #22695: MAINT: improve dtype handling now that xp.result_type accepts…

• #22696: MAINT: spatial: support empty case in RigidTransform

• #22698: MAINT/DOC: Update incomplete examples of expectile()

• #22701: TST: optimize: add more tests

• #22710: DOC: integrate.quad_vec: returned object is not a dictionary

• #22711: DOC: stats: Extend documentation of random_correlation matrix

• #22712: MAINT: bump array-api-extra to 0.7.0

• #22713: DOC: linalg.solve: clarify symmetry requirement

• #22714: MAINT: ndimage.maximum_filter: recommend vectorized_filter…

• #22715: ENH: ndimage.vectorized_filter: make CuPy-compatible

• #22716: DOC: optimize: Clarify use of xtol in 1D rootfinder docstrings

• #22718: TST: special: overhaul test_support_alternative_backends

• #22719: TST: add tests for ncfdtri

• #22722: DOC: ndimage.affine_transformation: add examples to docstring

• #22723: DOC: fft.dst: add example to docstring

• #22725: MAINT: ndimage.affine_transform: remove outdated and unhelpful…

• #22729: DOC: datasets.download_all: add examples to docstring

• #22735: ENH: stats: lazy trimmed stats for Dask and JAX

• #22738: DOC: PRIMA licence and reference fix

• #22740: TST: special: remove test skips due to array-api-strict#131

• #22741: CI: fix crash of free-threading job in sparse, bump GHA…

• #22742: CI/MAINT: make special.errstate thread-safe and run pytest-run-parallel…

• #22745: DOC: fft.rfft2: add example to docstring

• #22749: ENH: stats: add support for multiple parameterizations for custom…

• #22750: DOC: fft.hfft2: added example

• #22751: TST: linalg.test_batch: minor tolerance bumps

• #22755: MAINT: special: refine logsumexp writeback behaviour

• #22756: BUG/TST: special.logsumexp on non-default device

• #22759: TST: weightedtau rng thread safety

• #22760: BUG: optimize: VectorFunction.f_updated wasn’t being set…

• #22761: DOC: optimize: l-bfgs-b: clarify what is meant by maxfun...

• #22764: MAINT: optimize: VectorFunction: remove reference cycle

• #22766: DOC: improve docstrings of boxcox and yeojohnson

• #22770: TST: stats: add marray tests for _length_nonmasked directly

• #22771: TST: stats: don’t encapsulate pytest.warns

• #22778: MAINT: switch to vendoring libprima/prima

• #22779: MAINT: optimize: VectorFunction: fix array copy for sparse

• #22782: MAINT: fix failures in free-threading(parallel=1) job

• #22783: TST/MAINT: signal.symiirorder2: r, omega, precision are floats;…

• #22785: DOC/DEV: remove references to CirrusCI in skipping CI doc

• #22787: DOC: optimize: Add the multiplier details to SLSQP funcs

• #22788: TST: stats.quantile: add edge test case for axis=None && keepdims=True

• #22790: MAINT: optimize.least_squares: change x_scale default

• #22796: ENH/BLD: cython: share memoryview utility between extension modules

• #22798: TST: stats: mark some tests as slow

• #22802: BUG: optimize: Fix instability with NNLS on 32bit systems

• #22803: MAINT: use xp.asarray instead of xp.array

• #22805: CI: start using the CIBW_ENABLE env var

• #22807: TST: fix issue with cython_special test which was missing…

• #22808: BUG: special.logsumexp device propagation on PyTorch

• #22809: ENH: optimize.root: add warning for invalid inner parameters…

• #22811: ENH: ndimage.rotate: performance enhancement on WoA

• #22814: BUG: signal.resample: Fix bug for parameter num=2 (including…

• #22815: MAINT: sparse: add lazy loading for csgraph and linalg

• #22818: DEV: add .editorconfig

• #22820: MAINT: signal: consolidate order_filter tests

• #22821: ENH: signal.lp2{lp,hp,bp,bs}: add array API standard support

• #22823: MAINT: integrate.tanhsinh: simplify error estimate

• #22829: DOC: stats.qmc.discrepancy: clarify definitions

• #22832: DOC: interpolate: remove outdated deprecation notices

• #22833: DOC: special.comb: remove missed deprecation notice

• #22835: MAINT: stats.boxcox_llf: refactor for simplicity

• #22842: MAINT: bump boost_math to 1.88.0

• #22843: DOC: special: add xp_capabilities to logsumexp

• #22844: TST: stats: minor nits to test_stats.py

• #22845: TST: stats: reorder tests to match xp_capabilities

• #22846: MAINT: _lib/differentiate: update EIM with at.set

• #22848: MAINT: _lib: eliminate try/excepts in EIM

• #22850: TST: optimize VectorFunction add test for J0=None branch…

• #22852: TST: fix boxcox_llf test failure on main

• #22854: MAINT: special: Add xsf as a submodule of SciPy

• #22855: MAINT: spatial.pdist: make dimensionality error more descriptive

• #22858: DOC: Fix typo in ndimage.generic_gradient_magnitude()

• #22859: DOC: rewording of “ties” into “tied pairs” for clearer meaning

• #22862: TST: integrate/spatial: make fail_slow allowances

• #22863: TST: reintroduce eager_warns and fix free-threading test…

• #22864: MAINT: linalg.svd: raise correct error message for GESDD when…

• #22873: ENH: sparse: Support nD sum/mean/min/max/argmin for sparse arrays

• #22875: CI: limit pytest-fail-slow usage to a single CI job

• #22886: ENH: signal: filter design functions array API standard support

• #22891: DOC: Document allowed NumPy / Python versions

• #22893: MAINT: vendor qhull as subproject and add -Duse-system-libraries

• #22895: MAINT: signal: correct the get_window delegator

• #22896: ENH: signal: tf2zpk et al Array API

• #22897: ENH: sparse: ND binary operations support

• #22898: DEV: add editable install support for spin

• #22899: MAINT: bump array-api submodules

• #22900: MAINT: fix np.copyto warnings on Dask

• #22908: MAINT: bump qhull to 2020.2

• #22909: TST: Use jax_autojit

• #22913: BUG: fix syntax warning break in finally block under 3.14

• #22915: BLD: optimize sdist contents through a dist script

• #22916: DOC: integrate.solve_bvp: add missing reference details

• #22917: DEV: fix invocation of linter on Windows

• #22918: TST: linalg add test coverage to exception handling for invalid…

• #22926: MAINT: spatial.cKDTree: remove software prefetching and software…

• #22927: MAINT: tools/check_test_name: specify encoding

• #22930: DOC: linalg: update roadmap entry for BLAS/LAPACK bindings

• #22932: BUG: interpolate: do not call PyArray macros on non-arrays

• #22934: MAINT: optimize.zeros: fix error message

• #22939: TST: spatial.transform: Add array API standard support for testing

• #22941: MAINT: stats.qmc.Sobol: fix stacklevel of warning

• #22944: MAINT: fix regressions in array-api-strict after disabling np.float64

• #22946: ENH: special: add xp_capabilities

• #22947: MAINT: avoid nested asarray calls

• #22949: MAINT: mass rename make_skip_xp_backends to make_xp_test_case

• #22950: MAINT: refresh gpu-ci pixi.lock

• #22952: MAINT, DOC: forward port 1.15.3 release notes

• #22955: MAINT: wheel downloader

• #22959: ENH: cluster: more lazy functions

• #22960: DOC/TST: cluster.hierarchy: use xp_capabilities

• #22961: TST: cluster: reduce test reliance from linkage

• #22963: MAINT: wrap wrapcauchy samples around the circle

• #22967: CI: address some potential vulnerabilities

• #22968: DOC: outline that not all attributes of OptimizeResult may be…

• #22969: MAINT: stats.make_distribution: fix most remaining discrete distribution…

• #22970: MAINT: stats.DiscreteDistribution: fix inversion methods

• #22971: MAINT: fix skellam distribution tests

• #22973: BUG: interpolate.make_splrep: raise error when residuals.sum()…

• #22976: ENH: stats: Implement _munp for gennorm

• #22982: BUG: signal: fix incorrect vendoring of npp_polyval

• #22984: MAINT: special: remove test_compiles_in_cupy

• #22987: DOC: sparse: sparray migration guide updates

• #22992: ENH: signal.cspline1d_eval,qspline1d_eval throw exception…

• #22994: DOC: signal.csd: Small fixes to docstr

• #22997: CI: temporarily disable free-threaded job with parallel-threads

• #22998: BUG: Fix duplicate --pyargs due to double addition in SciPy…

• #22999: MAINT: bump up array-api-compat and array-api-extra

• #23000: ENH/DOC/TST: cluster.vq: use xp_capabilities

• #23001: DOC: special: update top-level docs to reflect xp_capabilities

• #23005: BUG: sparse: fix mean/sum change in return of np.matrix for sparse…