Clarification needed: Is this library to be used strictly on list of elements?

[amanjeev]

Hello writers and maintainers:

The documentation pointing to dp.tools has functions that act on some kind of aggregation of array or array_like "vectors". I wonder if there is a way to have the private outcome applied to individual elements of such a vector rather than some kind of aggregation like std, sum etc.

I am more than happy to contribute such a feature if you guide me whether it belongs here. My suggestion is on the lines of a utility function in tools/utils.py

def individual_application(array, epsilon=1.0, accountant=None, ...):
  """works on each individual numeric item in the `array`.
  this could also take a dict instead of an `array`.
  """
  # implemenation ...

Thank you,
Aman

[naoise-h]

Hi Aman,

You could apply a mechanism to every element of an array as follows:

def individual_application(array):
  mech = Laplace(epsilon=1, sensitivity=1)
  return [mech.randomise(val) for val in array]

In general, this is not recommended as it has an unfavourable privacy/accuracy tradeoff. It's not something we are considering adding to the library at present.

[amanjeev]

Thank you @naoise-h !