Source code for geomstats.numerics.interpolation

"""Interpolation machinery."""

from abc import ABC, abstractmethod

import geomstats.backend as gs


[docs] class Interpolator(ABC): """Abstract class for interpolator.""" def __call__(self, t): """Interpolate data. Parameters ---------- t : array-like, shape=[n_time] Interpolation time. Returns ------- point : array-like, shape=[..., n_time, *point_shape] """ return self.interpolate(t)
[docs] @abstractmethod def interpolate(self, t): """Interpolate data. Parameters ---------- t : array-like, shape=[n_time] Interpolation time. Returns ------- point : array-like, shape=[..., n_time, *point_shape] """
class _LinearInterpolator1D(Interpolator, ABC): def __init__(self, data, point_ndim=1): self.data = data self.point_ndim = point_ndim time_axis = -(point_ndim + 1) self._n_times = self.data.shape[time_axis] @abstractmethod def _from_t_to_interval(self, t): """Get interval index from time. Parameters ---------- t : array-like, shape=[n_time] Interpolation time. Returns ------- interval_index : array-like, shape=[n_times] """ def _get_ratio(self, t, interval_index, end_index): """Get ratio within interval. Parameters ---------- t : array-like, shape=[n_time] Interpolation time. interval_index : array-like, shape=[n_times] end_index : array-like, shape=[n_times] Returns ------- ratio : array-like, shape=[n_time] Ratio of t within interval. """ def interpolate(self, t): """Interpolate data. Parameters ---------- t : array-like, shape=[n_time] Interpolation time. Returns ------- point : array-like, shape=[..., n_time, *point_shape] """ if not gs.is_array(t): t = gs.array([t]) if gs.ndim(t) == 0: t = gs.expand_dims(t, axis=0) interval_index = self._from_t_to_interval(t) max_bound_reached = interval_index == self._n_times - 1 end_index = gs.where(max_bound_reached, interval_index, interval_index + 1) point_ndim_slc = (slice(None),) * self.point_ndim initial_point = self.data[..., interval_index, *point_ndim_slc] end_point = self.data[..., end_index, *point_ndim_slc] ratio = self._get_ratio(t, interval_index, end_index) diff = end_point - initial_point ijk = "ijk"[: self.point_ndim] return initial_point + gs.einsum(f"t,...t{ijk}->...t{ijk}", ratio, diff)
[docs] class UniformUnitIntervalLinearInterpolator(_LinearInterpolator1D): """A 1D linear interpolator. Assumes interpolation occurs in the unit interval and data is uniformly sampled. Parameters ---------- data : array-like, [..., *point_shape] point_ndim : int Dimension of point. """ def __init__(self, data, point_ndim=1): super().__init__(data, point_ndim=point_ndim) self._delta = 1 / (self._n_times - 1) def _from_t_to_interval(self, t): """Get interval index from time. Parameters ---------- t : array-like, shape=[n_time] Interpolation time. Returns ------- interval_index : array-like, shape=[n_times] """ return gs.cast( t // self._delta, dtype=gs.int32, ) def _get_ratio(self, t, interval_index, end_index): """Get ratio within interval. Parameters ---------- t : array-like, shape=[n_time] Interpolation time. interval_index : array-like, shape=[n_times] Ignored. end_index : array-like, shape=[n_times] Ignored. Returns ------- ratio : array-like, shape=[n_time] Ratio of t within interval. """ return gs.mod(t, self._delta) / self._delta
[docs] class LinearInterpolator1D(_LinearInterpolator1D): """A 1D linear interpolator. Assumes interpolation occurs in the unit interval. Parameters ---------- times : array-like, [n_times] Times. Must be sorted. data : array-like, [..., *point_shape] point_ndim : int Dimension of point. """ def __init__(self, times, data, point_ndim=1): super().__init__(data, point_ndim=point_ndim) self.times = times self._delta = self.times[1:] - self.times[:-1] def _from_t_to_interval(self, t): """Get interval index from time. Parameters ---------- t : array-like, shape=[n_time] Interpolation time. Returns ------- interval_index : array-like, shape=[n_times] """ indices = gs.searchsorted(self.times, t) - 1 return gs.where(indices < 0, 0, indices) def _get_ratio(self, t, interval_index, end_index): """Get ratio within interval. Parameters ---------- t : array-like, shape=[n_time] Interpolation time. interval_index : array-like, shape=[n_times] end_index : array-like, shape=[n_times] Returns ------- ratio : array-like, shape=[n_time] Ratio of t within interval. """ delta = self._delta[interval_index] return (delta - (self.times[end_index] - t)) / delta