def distributedVector(self, s):
# h = int(hashlib.md5(s.encode()).hexdigest(),16) % 100000000 #probably too slow and not necessary ??
# h = abs(mmh3.hash(s)) % 1000000
h = abs(op.hash(s)) % 4294967295
# h = np.abs(hash(s)) #devo hashare s in qualche modo (controllare che basti) e
np.random.seed(h) #inizializzare np.random.seed()
if self.mode == "binary":
q = 1 / np.sqrt(self.dimension)
return np.random.choice([-q, q], self.dimension)
else:
return op.random_vector(self.dimension, normalized=False)
def distributedVector(self, s):
if s in self.random_cache:
return self.random_cache[s]
# h = int(hashlib.md5(s.encode()).hexdigest(),16) % 100000000 #probably too slow and not necessary ??
# h = abs(mmh3.hash(s)) % 1000000
h = abs(op.hash(s)) % 4294967295
# h = np.abs(hash(s)) #devo hashare s in qualche modo (controllare che basti) e
np.random.seed(h) #inizializzare np.random.seed()
v_ = op.random_vector(self.dimension,normalized=False)
self.random_cache[s] = v_
return v_
__author__ = 'lorenzo'
import kerMIT.operation
import numpy
def covariance(a, b):
return numpy.cov(a.outer(b))
if __name__ == '__main__':
dimension = 100
size = 10000
M = numpy.zeros((dimension, size))
for i in range(size):
a = operation.random_vector(dimension)
b = operation.random_vector(dimension)
c = operation.circular_convolution(a, b)
M[:, i] = c
C = numpy.cov(M)
print(numpy.round(C * dimension, 2))
print(sum(x for x in numpy.nditer(C)) * dimension) #sum of all elements
print(numpy.trace(
C * dimension)) #trace.. should be approximately equal to dimension
print(numpy.round(numpy.diag(C) * dimension, 1))