def empirical_eps_DE(protocol, epsilon, n):
vars = []
for step in range(NUMBEROFAXIS): # 9 values
d = 2**(2 + step * 2)
x = protocol(d, epsilon) # init
users = zipf.zipf(1.1, d, n)
for i in range(len(users)):
x.PE(users[i])
x.aggregation() # estimate couterEsti
f = zipf.probList(1.1, d, n)
var = x.var_empirical(f, n)
vars.append(math.log10(var)) # log10(var)
return vars
def empirical_eps_BLH(protocol, epsilon, n):
vars = []
for step in range(NUMBEROFAXIS):
d = 2**(2 + step * 2)
x = protocol(d, epsilon, g=2) # init
users = zipf.zipf(1.1, d, n)
for i in range(len(users)):
if i % 1000 == 0:
print(step, i, 'BLH')
x.PE(users[i])
x.aggregation() # estimate couterEsti
f = zipf.probList(1.1, d, n)
var = x.var_empirical(f, n)
vars.append(math.log10(var))
return vars
def empirical_eps_OLH(protocol, epsilon, n):
vars = []
g = 53 # round(math.exp(epsilon)+1)效果不好，要取最近的素数（因为universal hashing！！)
for step in range(NUMBEROFAXIS): #### 9 values
d = 2**(2 + step * 2)
x = protocol(d, epsilon, g) # init
users = zipf.zipf(1.1, d, n)
for i in range(len(users)):
if i % 1000 == 0:
print(step, i, 'OLH')
x.PE(users[i])
x.aggregation() # estimate couterEsti
f = zipf.probList(1.1, d, n)
var = x.var_empirical(f, n)
vars.append(math.log10(var))
return vars
def empirical_OLH(protocol, d, n):
vars = []
primeList = [3, 3, 5, 7, 13, 23, 37, 53, 89,
149] # according to epsilon=[0.5, 1.0, ..., 5.0]
for step in range(NUMBEROFAXIS): # 10 values
epsilon = 0.5 + 0.5 * step
x = protocol(
d, epsilon, g=primeList[step]
) # round(math.exp(epsilon)+1)效果不好，要取最近的素数（因为universal hashing！！)
users = zipf.zipf(1.1, d, n)
for i in range(len(users)):
if i % 1000 == 0:
print(step, i, "OLH")
x.PE(users[i])
x.aggregation() # estimate couterEsti
f = zipf.probList(1.1, d, n)
var = x.var_empirical(f, n)
vars.append(math.log10(var))
return vars
def empirical_eps(protocol, epsilon, n):
vars = []
for step in range(7): # 9 values
d = 2**(2 + step * 2)
#f = []
x = protocol(d, epsilon, n)
users = zipf.zipf(1.1, d, n)
for i in range(n): # random PE
if i % 1000 == 0:
print(i, step)
#m = random.randint(1,d)
#x.PE(m)
x.PE(users[i] + 1)
f = zipf.probList(1.1, d, n)
#for i in range(d):
# f.append(1.0/d)
#print(f)
x.aggregation()
vars.append(x.var_empirical(f))
return vars