UT3Weights.reciprocal
=====================

.. py:method:: t3toolbox.uniform_tucker_tensor_train.UT3Weights.reciprocal()

   .. code-block:: python

      def reciprocal(self) -> 'UT3Weights':


   Elementwise ``1/w`` on the real slots (e.g. to form inverse-singular-value weights); the
   padding stays a canonical, **finite** zero rather than becoming ``inf``. Masks unchanged. See
   :py:func:`~t3toolbox.backend.ut3_operations.ut3_reciprocal_weights` for why that guard is
   load-bearing (and why real-slot zeros are deliberately *not* guarded).