def setUp(self):
self.prng = np.random.RandomState(10)
n = 8
I = matrix.Identity(n)
O = matrix.Ones(n, n)
P = workload.Prefix(n)
V = matrix.VStack([I, -0.5 * O])
H = matrix.HStack([I, -0.5 * O])
S = matrix.Sum([I, -0.5 * O])
K = matrix.Kronecker([I, V, P])
W = matrix.Weighted(K, 3.0)
M = workload.DimKMarginals((2, 3, 4), 2)
D = workload.Disjuncts([P, I])
N = workload.AllNormK(n, 2)
self.matrices = [I, O, P, V, H, K, O, W, D, N, M]
def select(self):
N = self.domain_shape[self.hb_dim]
domains = list(self.domain_shape)
del domains[self.hb_dim]
if self.impl == 'MM':
I = matrix.Identity(int(np.prod(domains)))
H = HB((N, )).select()
elif self.impl == 'sparse':
I = matrix.Identity(int(np.prod(domains))).sparse_matrix()
H = HB((N, )).select().sparse_matrix()
elif self.impl == 'dense':
I = matrix.Identity(int(np.prod(domains))).dense_matrix()
H = HB((N, )).select().dense_matrix()
else:
print("Invalid measurement type", self.impl)
exit(1)
return matrix.Kronecker([I, H])
def strategy(self):
return matrix.Kronecker([T.strategy() for T in self._templates])
def select(self):
if len(self.domain_shape) == 1:
return matrix.Haar(self.domain_shape[0])
return matrix.Kronecker([matrix.Haar(n) for n in self.domain_shape])