uniform_corewise_ops#
- t3toolbox.backend.uniform_fitting.uniform_corewise_ops(x0_data)#
def uniform_corewise_ops( x0_data: typ.Tuple, # UniformTuckerTensorTrain.data = (tk_sc, tt_sc, shape, (tucker_mask, tt_mask)) ) -> bopt.GeometryOps: # the uniform corewise geometry ops (masks of x0's fixed rank closed over)
The uniform corewise
GeometryOpsatx0’s fixed rank – the raw-supercore twin oft3toolbox.uniform_manifold.UNIFORM_COREWISE.frame= the(U, G, G, G)non-orthonormal frame (Section 6.3(P, Q, O) -> G);project= the identity (no gauge on the core space);retract= the additive retraction (cores += var);inner= the masked coordinate dot (the Euclidean metric here). Masks are loop-invariant and closed over.- Parameters:
x0_data (t3toolbox.backend.common.typ.Tuple)
- Return type: