build_input_jets#

t3toolbox.backend.sampling_derivatives.build_input_jets(xis, dxis)#
def build_input_jets(
        xis:    typ.Sequence[NDArray],  # frame projected probes,        len=d, elm_shape=W+C+(nUi,)
        dxis:   typ.Sequence[NDArray],  # projected perturbation dirs,  len=d, elm_shape=W+C+(nUi,)
) -> typ.Tuple[NDArray, ...]:           # xi_jets. len=d, elm_shape=(2,)+W+C+(nUi,): order 0 = xi, 1 = dxi

Input jets: stack each (value, direction) pair on a leading order axis.

Since x + s p is affine in s, an input vector’s jet is just (x, p, 0, ...) – value at order 0, direction at order 1, zero above. Stored at size 2 (orders 0,1); the pushthrough slices the binomial tensor to s in {0,1} accordingly.

Parameters:

  • xis (t3toolbox.backend.common.typ.Sequence[NDArray])

  • dxis (t3toolbox.backend.common.typ.Sequence[NDArray])

Return type:

t3toolbox.backend.common.typ.Tuple[NDArray, Ellipsis]