[Python] Numpy vectorized version of my deterministic Python function returns different set of...

In the following runnable Python program the arrays values and check_values contain different elements. On my machine the complete output is the following:

Received k = 0, generating value n = 0; 3k < n <==> False
Received k = 0, generating value n = 0; 3k < n <==> False
Received k = 1, generating value n = 2; 3k < n <==> False
Received k = 2, generating value n = 2; 3k < n <==> False
Received k = 3, generating value n = 4; 3k < n <==> False
Received k = 0, generating value n = 0; 3k < n <==> False
Received k = 1, generating value n = 2; 3k < n <==> False
Received k = 2, generating value n = 2; 3k < n <==> False
Received k = 3, generating value n = 4; 3k < n <==> False
values: [0 0 1 0]
check_values: [0. 0.125 1.81097638 0.98330026]


The list check_values contains correct values; the list values does not. I have no clue what could be the reason for this and I could not find help online. All help is appreciated!

The code:

import numpy as np
import math
from math import comb as ncr

def bound_1(k, n):
t = 3 * k - n
if t % 2 != 0:
t -= 1
return int(t / 2)


def bound_2(k, n):
t = 3 * k - n
ret = 0
if t % 2 == 0:
ret = t / 2 - 1
else:
ret = (t - 1) / 2
return int(ret)


def binomial_comb_sum(s, e, v):
sum = 0
for m in range(s, e + 1):
sum += ncr(v, m)
return sum


def binary_entropy(p):
if p <= 0 or p >= 1: return 0
return -(p * math.log2(p) + (1 - p) * math.log2(1 - p))


def n_value_from_k(k):
return 2 * int(math.pow(k, 0.75))


def energy(k, n):
if 3 * k < n: return 0
b1, b2 = bound_1(k, n), bound_2(k, n)
sum1, sum2 = 0, 0

for p in range(1, b1 + 1):
t = n + 2 * p
d = binary_entropy(p / t) - 1
sum1 += math.pow(2, t * d)

for p in range(0, b2 + 1):
t = n + 2 * p + 1
d = binary_entropy(p / t) - 1
sum2 += math.pow(2, t * d)

return sum1 + sum2

def energy_wrt_k(k):
n = n_value_from_k(k)
print(f"Received k = {k}, generating value n = {n}; 3k < n <==> {3 * k < n}")
return energy(k, n)

N = 3
k_vals = np.arange(N + 1)

vectorized_energy_wrt_k = np.vectorize(energy_wrt_k)

values = vectorized_energy_wrt_k(k_vals)

check_values = np.array([energy_wrt_k(k) for k in k_vals])


print("values:", values)
print("check_values:", check_values)

Continue reading...