Evgeni Burovski

### Is your feature request related to a problem? Please describe. Add an option to specify boundary conditions to `make_lsq_spline`. A first useful step would be to allow specifying the...

Tracking issue for temporary test skips/xfails to revert

5

(As suggested in https://github.com/scipy/scipy/pull/17273#issuecomment-1288059660) Once in a while, we skip or xfail tests to quickly restore the CI to the green status, when we either lack the immediate bandwidth to...

how to properly import cupyx submodules

2

### Description Are cupyx.scipy submodules meant to be available if - the top-level `cupyx` name is imported - the `cupyx.scipy` name is imported - the `cupyx.scipy.` name is imported The...

Otherwise, all array API implementors resort to `exec` via `array_api_compat`: https://github.com/data-apis/array-api-compat/blob/main/array_api_compat/cupy/fft.py#L2

### Description `scipy.ndimage.gaussian_filter1d` and `gaussian_filter` accept a `radius` keyword, https://github.com/scipy/scipy/blob/v1.14.0/scipy/ndimage/_filters.py#L277, which the `cupyx.scipy.ndimage` analogs do not. ### Additional Information _No response_

### Description scipy.ndimage.correlate support `mode='grid-constant'` but the cupyx analog does not ``` In [1]: import cupy as xp In [2]: from cupyx.scipy import ndimage as c_ndi In [3]: c_ndi.correlate(xp.array([[1, 2,...

### Description `ndimage.rank_filter` and other filters accept an optional `output=` argument for the output array. If the output overlaps with the input array, `rank_filter` does not perform filtering and simply...

### Description ``` In [1]: import cupy as xp In [2]: from cupyx.scipy import ndimage as c_ndi In [3]: c_ndi.correlate1d(xp.array([1.+0.j, 0.+2.j, 3.+0.j, 1.+4.j, 5.+0.j, 0.+6.j], dtype=xp.complex64), xp.array([1, 0, 1], dtype=xp.int8),...

### Description `cupyx.scipy.ndimage.gaussian_filter` seems to lose precision in float32, worse than the scipy analog. The reproducer below is from https://github.com/scipy/scipy/blob/main/scipy/ndimage/tests/test_filters.py#L663, which has a comment ``` # input.sum() is 49995000.0. With...

### Description In some cases, `cupyx.scipy.ndimage.uniform_filter` gives wrong results. In the reproducer below, `filter_size=2` agrees with scipy, but `filter_size=6` does not. The scipy test has roundoff_errors in the name, so...