tv_entries#

t3toolbox.backend.entries.tv_entries(index, variation, frame)#

def tv_entries(
        index:      NDArray,                # int, shape=(d,)+W (index stack W)
        variation:  typ.Tuple[typ.Sequence[NDArray], typ.Sequence[NDArray]],  # (var_tucker, var_tt)
        frame:       typ.Tuple[typ.Sequence[NDArray], typ.Sequence[NDArray],
                              typ.Sequence[NDArray], typ.Sequence[NDArray]],   # (up, down, left, right)
) -> NDArray:                               # entries of the dense tangent at ``index``; shape = W + K + C

Extract entries of the dense tangent at index (= apply with unit vectors, by slicing).

Identical to tv_apply() except the up-index edge variables come from slicing the Tucker-core fibers (U_i[..., index_i] and dU_i[..., index_i]) rather than contracting with vectors – so there is no contraction with unit basis vectors and no N factor.