def _resample_Z_gsl(self, data=[]):
"""
Resample the parents given the bias_model, weight_model, and impulse_model.
:param bias_model:
:param weight_model:
:param impulse_model:
:return:
"""
from gslrandom import multinomial
bias_model, weight_model, impulse_model = \
self.model.bias_model, self.model.weight_model, self.model.impulse_model
K, B = self.K, self.B
# Make a big matrix of size T x (K*B + 1)
for k2, (Sk, Fk, Tk,
Zk) in enumerate(zip(self.Ss, self.Fs, self.Ts, self.Z)):
P = np.zeros((Tk, 1 + self.K * self.B))
P[:, 0] = bias_model.lambda0[k2]
Wk2 = np.repeat(weight_model.W_effective[:, k2], B)
Gk2 = impulse_model.g[:, k2, :].reshape((K * B, ), order="C")
P[:, 1:] = Wk2 * Gk2 * Fk.reshape((Tk, K * B))
# Normalize the rows
P = P / P.sum(1)[:, None]
# Sample parents from P with counts S[:,k2]
multinomial(self.pyrngs, Sk, P, out=Zk)
def _resample_Z_gsl(self, data=[]):
"""
Resample the parents given the bias_model, weight_model, and impulse_model.
:param bias_model:
:param weight_model:
:param impulse_model:
:return:
"""
from gslrandom import multinomial
bias_model, weight_model, impulse_model = \
self.model.bias_model, self.model.weight_model, self.model.impulse_model
K, B = self.K, self.B
# Make a big matrix of size T x (K*B + 1)
for k2, (Sk, Fk, Tk, Zk) in enumerate(zip(self.Ss, self.Fs, self.Ts, self.Z)):
P = np.zeros((Tk, 1+self.K * self.B))
P[:,0] = bias_model.lambda0[k2]
Wk2 = np.repeat(weight_model.W_effective[:,k2], B)
Gk2 = impulse_model.g[:,k2,:].reshape((K*B,), order="C")
P[:,1:] = Wk2 * Gk2 * Fk.reshape((Tk, K*B))
# Normalize the rows
P = P / P.sum(1)[:,None]
# Sample parents from P with counts S[:,k2]
multinomial(self.pyrngs, Sk, P, out=Zk)
def test_one_rng_one_N_one_p_with_out():
K = 5
N = 10
p_K = np.ones(K) / K
rng = PyRNG(rn.randint(2**16))
n_iter = 10000
z_K = np.zeros(K)
for _ in xrange(n_iter):
n_K = np.zeros(K, dtype=np.float)
multinomial(rng, N, p_K, out=n_K)
assert n_K.sum() == N
z_K += n_K
assert np.allclose(z_K / z_K.sum(), p_K, atol=1e-2)
def test_one_rng_one_N_one_p_with_out():
K = 5
N = 10
p_K = np.ones(K) / K
rng = PyRNG(rn.randint(2**16))
n_iter = 10000
z_K = np.zeros(K)
for _ in xrange(n_iter):
n_K = np.zeros(K, dtype=np.float)
multinomial(rng, N, p_K, out=n_K)
assert n_K.sum() == N
z_K += n_K
assert np.allclose(z_K / z_K.sum(), p_K, atol=1e-2)
def test_simple_with_out():
# One N count, one p array, out structure provided
N = 10
K = 5
p = 1./K * np.ones(K)
rng = PyRNG()
n_iter = 10000
z = np.zeros(K)
for _ in xrange(n_iter):
out = np.zeros(K, dtype=np.float)
multinomial(rng, N, p, out)
assert out.sum() == N
z += out
assert (np.abs(z/z.sum() - p) < 1e-2).all()
def test_multi_N_single_p_with_out():
# Multiple N counts, one p array, out structure provided
L = 10
N = np.arange(L) + 10
K = 5
p = 1./K * np.ones(K)
rng = PyRNG()
n_iter = 10000
z = np.zeros((L,K))
for _ in xrange(n_iter):
out = np.zeros((L,K), dtype=np.uint32)
multinomial(rng, N, p, out)
assert (out.sum(axis=1) == N).all()
z += out
assert (np.abs(z/z.sum(axis=1)[:,np.newaxis] - p) < 1e-2).all()
def test_one_rng_multi_N_one_p_with_out():
K = 5
I = 10
N_I = np.ones(I) * 10
p_K = np.ones(K) / K
rng = PyRNG(rn.randint(2**16))
n_iter = 10000
z_IK = np.zeros((I, K))
for _ in xrange(n_iter):
n_IK = np.zeros((I, K))
multinomial(rng, N_I, p_K, out=n_IK)
assert np.allclose(n_IK.sum(axis=1), N_I)
z_IK += n_IK
norm_z_IK = z_IK.astype(float) / np.sum(z_IK, axis=1, keepdims=True)
p_IK = np.ones((I, K)) * p_K
assert np.allclose(norm_z_IK, p_IK, atol=1e-2)
def test_one_rng_multi_N_multi_p_with_out():
K = 5
I = 3
N_I = np.arange(1, I+1) * 10
p_IK = np.ones((I, K)) / K
p_IK[0, :] = [0.5, 0.25, 0.05, 0.1, 0.1] # make one non-uniform
rng = PyRNG(rn.randint(2**16))
n_iter = 10000
z_IK = np.zeros((I, K))
for _ in xrange(n_iter):
n_IK = np.zeros((I, K))
multinomial(rng, N_I, p_IK, out=n_IK)
np.allclose(n_IK.sum(axis=1), N_I)
z_IK += n_IK
norm_z_IK = z_IK.astype(float) / np.sum(z_IK, axis=1, keepdims=True)
assert np.allclose(norm_z_IK, p_IK, atol=1e-2)
def test_one_rng_multi_N_one_p_with_out():
K = 5
I = 10
N_I = np.ones(I) * 10
p_K = np.ones(K) / K
rng = PyRNG(rn.randint(2**16))
n_iter = 10000
z_IK = np.zeros((I, K))
for _ in xrange(n_iter):
n_IK = np.zeros((I, K))
multinomial(rng, N_I, p_K, out=n_IK)
assert np.allclose(n_IK.sum(axis=1), N_I)
z_IK += n_IK
norm_z_IK = z_IK.astype(float) / np.sum(z_IK, axis=1, keepdims=True)
p_IK = np.ones((I, K)) * p_K
assert np.allclose(norm_z_IK, p_IK, atol=1e-2)
def test_one_rng_multi_N_multi_p_with_out():
K = 5
I = 3
N_I = np.arange(1, I + 1) * 10
p_IK = np.ones((I, K)) / K
p_IK[0, :] = [0.5, 0.25, 0.05, 0.1, 0.1] # make one non-uniform
rng = PyRNG(rn.randint(2**16))
n_iter = 10000
z_IK = np.zeros((I, K))
for _ in xrange(n_iter):
n_IK = np.zeros((I, K))
multinomial(rng, N_I, p_IK, out=n_IK)
np.allclose(n_IK.sum(axis=1), N_I)
z_IK += n_IK
norm_z_IK = z_IK.astype(float) / np.sum(z_IK, axis=1, keepdims=True)
assert np.allclose(norm_z_IK, p_IK, atol=1e-2)
def test_single_N_multi_p_with_out():
# One N count, multiple p arrays, out structure provided
N = 10
K = 5
p = np.zeros((2, K))
p[0,:] = 1./K * np.ones(K)
p[1,:] = np.asarray([0.5, 0.25, 0.05, 0.1, 0.1])
rng = PyRNG()
n_iter = 10000
z = np.zeros((2,K))
for _ in xrange(n_iter):
out = np.zeros((2,K), dtype=np.uint32)
multinomial(rng, N, p, out)
assert out.shape == (2,K)
assert (out.sum(axis=1) == N).all()
z += out
assert (np.abs(z/z.sum(axis=1)[:,np.newaxis] - p) < 1e-2).all()
def test_multi_rng_multi_N_multi_p_with_out():
K = 5
A = 3
B = 2
N_AB = (np.arange(1, A*B+1) * 10).reshape((A, B))
p_ABK = np.ones((A, B, K)) / K
p_ABK[0, 1, :] = [0.5, 0.25, 0.05, 0.1, 0.1] # make one non-uniform
p_ABK[1, 0, :] = [0.9, 0.05, 0.03, 0.01, 0.01] # make one really non-uniform
rngs = [PyRNG(rn.randint(2**16)) for _ in xrange(get_omp_num_threads())]
n_iter = 10000
z_ABK = np.zeros((A, B, K))
for _ in xrange(n_iter):
n_ABK = np.zeros((A, B, K), dtype=np.uint32)
multinomial(rngs, N_AB, p_ABK, out=n_ABK)
np.allclose(n_ABK.sum(axis=-1), N_AB)
z_ABK += n_ABK
norm_z_ABK = z_ABK.astype(float) / np.sum(z_ABK, axis=-1, keepdims=True)
assert np.allclose(norm_z_ABK, p_ABK, atol=1e-2)
def test_multi_rng_multi_N_multi_p_with_out():
K = 5
A = 3
B = 2
N_AB = (np.arange(1, A * B + 1) * 10).reshape((A, B))
p_ABK = np.ones((A, B, K)) / K
p_ABK[0, 1, :] = [0.5, 0.25, 0.05, 0.1, 0.1] # make one non-uniform
p_ABK[1, 0, :] = [0.9, 0.05, 0.03, 0.01,
0.01] # make one really non-uniform
rngs = [PyRNG(rn.randint(2**16)) for _ in xrange(get_omp_num_threads())]
n_iter = 10000
z_ABK = np.zeros((A, B, K))
for _ in xrange(n_iter):
n_ABK = np.zeros((A, B, K), dtype=np.uint32)
multinomial(rngs, N_AB, p_ABK, out=n_ABK)
np.allclose(n_ABK.sum(axis=-1), N_AB)
z_ABK += n_ABK
norm_z_ABK = z_ABK.astype(float) / np.sum(z_ABK, axis=-1, keepdims=True)
assert np.allclose(norm_z_ABK, p_ABK, atol=1e-2)
def test_multi_N_single_p():
# Multiple N counts, one p array, out structure NOT provided
L = 10
N = np.arange(L) + 10
K = 5
p = 1./K * np.ones(K)
rng = PyRNG()
n_iter = 10000
z = np.zeros((L,K))
for _ in xrange(n_iter):
n = multinomial(rng, N, p)
assert n.shape == (L,K)
assert (n.sum(axis=1) == N).all()
z += n
assert (np.abs(z/z.sum(axis=1)[:,np.newaxis] - p) < 1e-2).all()
def test_one_rng_one_N_multi_p_no_out():
K = 5
I = 3
N = 10
p_IK = np.ones((I, K)) / K
p_IK[0, :] = [0.5, 0.25, 0.05, 0.1, 0.1] # make one non-uniform
rng = PyRNG(rn.randint(2**16))
n_iter = 10000
z_IK = np.zeros((I, K))
for _ in xrange(n_iter):
n_IK = multinomial(rng, N, p_IK)
assert n_IK.shape == (I, K)
assert (n_IK.sum(axis=1) == N).all()
z_IK += n_IK
norm_z_IK = z_IK.astype(float) / np.sum(z_IK, axis=1, keepdims=True)
assert np.allclose(norm_z_IK, p_IK, atol=1e-2)
def test_simple():
# One N count, one p array, out structure NOT provided
N = 10
K = 5
p = 1./K * np.ones(K)
rng = PyRNG(np.random.randint(2**16))
# rng = PyRNG(0)
n_iter = 10000
z = np.zeros(K)
for _ in xrange(n_iter):
n = multinomial(rng, N, p)
assert n.sum() == N
z += n
print n
assert (np.abs(z/z.sum() - p) < 1e-2).all()
def test_single_N_multi_p():
# One N count, multiple p arrays, out structure NOT provided
N = 10
K = 5
p = np.zeros((2, K))
p[0,:] = 1./K * np.ones(K)
p[1,:] = np.asarray([0.5, 0.25, 0.05, 0.1, 0.1])
rng = PyRNG()
n_iter = 10000
z = np.zeros((2,K))
for _ in xrange(n_iter):
n = multinomial(rng, N, p)
assert n.shape == (2,K)
assert (n.sum(axis=1) == N).all()
z += n
assert (np.abs(z/z.sum(axis=1)[:,np.newaxis] - p) < 1e-2).all()
def test_one_rng_one_N_multi_p_no_out():
K = 5
I = 3
N = 10
p_IK = np.ones((I, K)) / K
p_IK[0, :] = [0.5, 0.25, 0.05, 0.1, 0.1] # make one non-uniform
rng = PyRNG(rn.randint(2**16))
n_iter = 10000
z_IK = np.zeros((I, K))
for _ in xrange(n_iter):
n_IK = multinomial(rng, N, p_IK)
assert n_IK.shape == (I, K)
assert (n_IK.sum(axis=1) == N).all()
z_IK += n_IK
norm_z_IK = z_IK.astype(float) / np.sum(z_IK, axis=1, keepdims=True)
assert np.allclose(norm_z_IK, p_IK, atol=1e-2)
def resample_z(self):
topicprobs = self._get_topicprobs()
multinomial(self.pyrngs, self.data.data, topicprobs, self.z)
self._update_counts()
def coupling_random_sample(n, l, s, threads):
P = np.empty((s, len(l)), dtype=np.uint32)
l_tile = np.tile(l, (s, 1))
rngs = [PyRNG(np.random.randint(2**16)) for _ in range(threads)]
return multinomial(rngs, n, l_tile, P)
K = 1000
N_I = rn.poisson(rn.gamma(2, 500, size=I)).astype(np.uint32)
P_IK = 1. / K * np.ones((I, K), dtype=np.float)
N_IK = np.ones((I, K), dtype=np.uint32)
s = time.time()
seeded_multinomial(N_I, P_IK, N_IK)
print '%fs: No PyRNG parallel version with %d cores' % (time.time() - s,
get_omp_num_threads())
assert (N_IK.sum(axis=1) == N_I).all()
rngs = [PyRNG(rn.randint(2**16)) for _ in xrange(get_omp_num_threads())]
N_IK = np.ones((I, K), dtype=np.uint32)
s = time.time()
multinomial(rngs, N_I, P_IK, N_IK)
print '%fs: PyRNG parallel version with %d cores' % (time.time() - s,
len(rngs))
assert (N_IK.sum(axis=1) == N_I).all()
rng = PyRNG(rn.randint(2**16))
N_IK = np.ones((I, K), dtype=np.uint32)
s = time.time()
multinomial(rng, N_I, P_IK, N_IK)
print '%fs: PyRNG version with 1 core' % (time.time() - s)
assert (N_IK.sum(axis=1) == N_I).all()
rng = PyRNG()
N_IK = np.ones((I, K), dtype=np.uint32)
s = time.time()
for i in xrange(I):
I = 100000
K = 1000
N_I = rn.poisson(rn.gamma(2, 500, size=I)).astype(np.uint32)
P_IK = 1./K * np.ones((I, K), dtype=np.float)
N_IK = np.ones((I, K), dtype=np.uint32)
s = time.time()
seeded_multinomial(N_I, P_IK, N_IK)
print '%fs: No PyRNG parallel version with %d cores' % (time.time() - s, get_omp_num_threads())
assert (N_IK.sum(axis=1) == N_I).all()
rngs = [PyRNG(rn.randint(2**16)) for _ in xrange(get_omp_num_threads())]
N_IK = np.ones((I, K), dtype=np.uint32)
s = time.time()
multinomial(rngs, N_I, P_IK, N_IK)
print '%fs: PyRNG parallel version with %d cores' % (time.time() - s, len(rngs))
assert (N_IK.sum(axis=1) == N_I).all()
rng = PyRNG(rn.randint(2**16))
N_IK = np.ones((I, K), dtype=np.uint32)
s = time.time()
multinomial(rng, N_I, P_IK, N_IK)
print '%fs: PyRNG version with 1 core' % (time.time() - s)
assert (N_IK.sum(axis=1) == N_I).all()
rng = PyRNG()
N_IK = np.ones((I, K), dtype=np.uint32)
s = time.time()
for i in xrange(I):
N_IK[i, :] = rn.multinomial(N_I[i], P_IK[i, :])
