MPC Tensor Prototype

All functionality in this notebook is just a proof of concept. The tensor is not stable. For a full survey of the tech behind this tensor, see the following blogpost

mortendahl.github.io/2017/04/17/private-deep-learning-with-mpc/



In [1]:

    
import numpy as np
from syft.mpc.rss import MPCRepo
from syft.mpc.rss.tensor import RSSMPCTensor



In [2]:

    
bob = MPCRepo()
alice = MPCRepo()
sam = MPCRepo()

bob.set_siblings(alice,sam)
alice.set_siblings(sam,bob)
sam.set_siblings(bob,alice)



In [3]:

    
r = RSSMPCTensor(alice,np.random.rand(10))



In [4]:

    
r









    Out[4]:





RSSMPCTensor: array([0.50083556, 0.96614524, 0.9928066, 0.21695686, 0.09230825,
       0.41738036, 0.93996725, 0.26363021, 0.74307937, 0.02905819], dtype=object)



In [5]:

    
r+r









    Out[5]:





RSSMPCTensor: array([1.00167112, 1.93229048, 1.9856132, 0.43391372, 0.1846165,
       0.83476072, 1.8799345, 0.52726042, 1.48615874, 0.05811638], dtype=object)



In [6]:

    
r*r









    Out[6]:





RSSMPCTensor: array([0.25083626, 0.93343663, 0.98566495, 0.04707028, 0.00852081,
       0.17420636, 0.88353843, 0.06950089, 0.55216695, 0.00084438], dtype=object)



In [7]:

    
r - r









    Out[7]:





RSSMPCTensor: array([0, 0, 0, 0, 0, 0, 0, 0, 0, 0], dtype=object)



In [8]:

    
r.dot(r)









    Out[8]:





3.90578595



In [10]:

    
r.sum()









    Out[10]:





5.16216789



In [ ]: