I think the bit_decomp should be:
def bit_decomp(vector, q):
l = ceil(log2(q))
result = []
for elem in vector:
elem = elem % q
for i in range(l):
result.append((elem // 2 ** i) % 2)
return np.array(result)
The old version can be wrong if there is any elem in the vector is larger than the modulus q
I think the bit_decomp should be:
The old version can be wrong if there is any elem in the vector is larger than the modulus q