import random
from math import ceil, sqrt
import numpy as np
[docs]
def BernoulliSparse(
dimensions: int, dtype: type = np.int32, probability: float = None
) -> np.ndarray:
"""BernoulliSparse X∈{0,1}, X_i ~ B(p << 1)
Bernoulli distribution of binary numbers, either 0 or 1 with variable
probability. If undefined probability is given as p = 1/sqrt(dimensions),
which creates a sparsely populated array
:param dimensions: number of dimensions in the vector
:type dimensions: int
:param probability: probability of the Bernoulli distribution, defaults to 1/sqrt(d)
:type probability: float, optional
:return: hypervector
:rtype: np.ndarray
"""
if probability is None:
probability = 1 / sqrt(dimensions)
return np.random.binomial(size=dimensions, n=1, p=probability).astype(dtype)
[docs]
def SparseSegmented(
dimensions: int, dtype: type = np.int32, probability: float = None
) -> np.ndarray:
"""SparseSegmented X∈{0,1}, X_i ~ B(p << 1)
Sparsely segmented binary numbers, either 0 or 1 with variable
probability. If undefined probability is given as p = 1/sqrt(dimensions).
Hypervector is split into s, dimensions * probability, segments. Each segment
populated with exactly 1 non-zero element uniformly distributed throughout the
segment.
NOTE: Since s = dimensions * probability is not garunteed to evenly divide into
the hypervector dimensions, the segment, s, is rounded up and the final hypervector trimmed.
This means the non-zero value in the last segment may be trimmed and not be present in the
final hypervector.
:param dimensions: number of dimensions in the vector
:type dimensions: int
:param probability: probability of the Bernoulli distribution, defaults to 1/sqrt(d)
:type probability: float, optional
:return: hypervector
:rtype: np.ndarray
"""
if probability is None:
probability = 1 / sqrt(dimensions)
s = ceil(dimensions * probability)
seg_d = ceil(dimensions / s)
hypervector = np.zeros(0, dtype=np.int8)
for i in range(0, s):
vector = np.zeros(seg_d, dtype=np.int8)
index = random.randrange(seg_d)
vector[index] = 1
hypervector = np.append(hypervector, vector)
return hypervector[:dimensions].astype(dtype)