PointCloud.vec_apply_cost(arr, axis=0, fn=None)[source]#

Apply the geometry’s cost matrix in a vectorised way.

This performs either: output = C arr (if axis=1) output = C’ arr (if axis=0) where C is [num_a, num_b]

This function can be used when the cost matrix is squared euclidean and fn is a linear map.

  • arr (ndarray) – jnp.ndarray [num_a or num_b, p], vector that will be multiplied by the cost matrix.

  • axis (int) – standard cost matrix if axis=1, transport if 0

  • fn (Optional[Callable[[ndarray], ndarray]]) – function optionally applied to cost matrix element-wise, before the apply

Return type



A jnp.ndarray, [num_b, p] if axis=0 or [num_a, p] if axis=1