v0.3.16

What's Changed

• Adds RBFInterpolator, a radial basis function interpolator for scattered data in
  N dimensions with an API matching scipy.interpolate.RBFInterpolator. Supports all
  the scipy kernels, smoothing, and an added polynomial term. Passing neighbors
  builds the interpolant at each query point from only the nearest data points, using a
  KD-tree from the new jaxkd dependency.
• Adds CubicSmoothingSpline, a cubic smoothing spline for noisy data. The amount
  of smoothing is set by a normalized parameter p in [0, 1], as in csaps, and all
  the boundary conditions of CubicSpline are supported, including periodic.
• Lots of under the hood changes to reduce JIT compile overhead and speed up
  constructing and evaluating splines.
• Adds even_spacing option to interpXd and InterpolatorXD. If the knots are
  evenly spaced, the interval containing each query point is computed directly instead
  of by binary search. InterpolatorXD detects evenly spaced knots automatically by
  default, once at construction. interpXd uses a binary search by default; pass True
  to assume even spacing, or None to check on each call.
• The binary search used for unevenly spaced knots is unrolled, which is several times
  faster on GPU.
• Bump maximum JAX version to 0.12.x
• Adds support for Python 3.15, drops support for Python 3.10

New Contributors

• @mtruel made their first contribution in #111
• @YigitElma made their first contribution in #148

Full Changelog: v0.3.15...v0.3.16