'''

Day 20: Fri 19 Jul 2019 @ Beijing

1. Secure and Private AI Challenge

- Review PATE analysis, found out I cannot recall the content by just reading and looking the codes…! How important reviewing is in terms of studying new things!!!

- repo: https://github.com/marofmar/60daysofudacity/blob/master/Day20_review_PATE.py

- OMG...? please,,, at leat github should work in here.

'''

import numpy as np

num_teachers = 10

num_examples = 10000

num_labels = 10

preds = (np.random.rand(num_teachers, num_examples) * num_labels).astype(int).transpose(1,0)

new_labels = list()

for an_image in preds:

label_counts = np.bincount(an_image, minlength = num_labels)

epsilon = 0.1

beta = 1/ epsilon

for i in range(len(label_counts)):

label_counts[i] += np.random.laplace(0, beta, 1)

nes_label = np.argmax(label_counts)

new_labels.append(new_label)

labels = np.array([9,9,3,6,9,9,9,9,8,2])

counts = np.bincount(lables, minlength = 10)

query_result = np.argmax(counts)

print(query_result)

'''

conda activate syft

pip install pysyft

pip install numpy

'''

from syft.frameworks.torch.differential_privacy import pate

num_teachers, num_exmples, num_labels = (100,100,10)

preds = (np.random.rand(num_teachers, num_examples)*num_labels).astype(int)

indices = (np.random.rand(num_exmaples)*num_labels).astype(int)

data_dep_eps, data_ind_eps = pate.perform_analysis(teacher_preds = preds, indices = indices, noise_eps = 0.1, delta = 1e-5)

print("Data Independent Epsilon: ", data_ind_eps)

print("Data Dependent Epsilon: ", data_dep_eps)

preds[:, 0:5]*= 0

data_dep_eps, data_ind_eps = pate.perform_anlysis(teacher_pred = preds, indices = indices, noise_eps = 0.1, delta = 1e-5)

print("Data Independent Epsilon ", data_ind_eps)

print("Data Dependent Epsilon: ", data_dep_eps)

preds[:. 0:50] *= 0

data_dep_eps, data_ind_eps = pate.perform_analysis(teacher_preds = preds, indices = indices, noise_eps = 0.1, delta = 1e-5, moments = 20)

prit("Data Independent Epsilon: ", data_ind_eps)

print("Data Dependent Epsilon: ", data_dep_eps)