flip.nvflare.components.custom_percentile_privacy
=================================================

.. py:module:: flip.nvflare.components.custom_percentile_privacy


Classes
-------

.. autoapisummary::

   flip.nvflare.components.custom_percentile_privacy.PercentilePrivacy


Module Contents
---------------

.. py:class:: PercentilePrivacy(percentile=10, gamma=0.01, data_kinds: List[str] | None = None, off: bool = False)

   Bases: :py:obj:`nvflare.apis.dxo_filter.DXOFilter`


   
   Implementation of "largest percentile to share" privacy preserving policy.

   Shokri and Shmatikov, Privacy-preserving deep learning, CCS '15

   :param percentile: Only abs diff greater than this percentile is updated.
                      Allowed range 0..100.  Defaults to 10.
   :type percentile: int, optional
   :param gamma: The upper limit to truncate abs values of weight diff. Defaults to 0.01.
   :type gamma: float, optional
   :param Any weight diff with abs<gamma will become 0.:
   :param data_kinds: kinds of DXO to filter
   :param off: If True, the filter is turned off. Defaults to False.
   :type off: bool, optional


   .. py:attribute:: percentile
      :value: 10



   .. py:attribute:: gamma
      :value: 0.01



   .. py:attribute:: off
      :value: False



   .. py:method:: process_dxo(dxo: nvflare.apis.dxo.DXO, shareable: nvflare.apis.shareable.Shareable, fl_ctx: nvflare.apis.fl_context.FLContext) -> Union[None, nvflare.apis.dxo.DXO]

      Compute the percentile on the abs delta_W.

      Only share the params where absolute delta_W greater than
      the percentile value

      :param dxo: information from client
      :param shareable: that the dxo belongs to
      :param fl_ctx: context provided by workflow

      Returns: filtered dxo