def test_update(device, f, dtype):
npdtype = dtype.as_numpy_dtype
M, K, tau, F = (20, 30), 3, 0.1, 2
npU = (np.random.normal(size=(K, M[0])).astype(npdtype),
np.random.normal(size=(K, M[1])).astype(npdtype))
U = [tf.constant(npU[0]), tf.constant(npU[1])]
npnoise = np.random.normal(size=M).astype(npdtype)
npdata = np.dot(npU[0].T, npU[1]) + npnoise
data = tf.constant(npdata, dtype=dtype)
lh = Normal2dLikelihood(M=M, K=K, tau=tau, dtype=dtype)
lh.init(data=data)
properties = Properties(persistent=True,
dtype=dtype)
prior = Uniform(dummy=tf.constant(np.random.random(K).astype(npdtype),
dtype=dtype),
properties=properties)
postUf = PostU(lh, prior, f)
Ufupdated = postUf.update(U, data, transform=False)
for g in range(F):
assert(Ufupdated.dtype == dtype)
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
npUfupdated = sess.run(Ufupdated)
assert(not np.allclose(npU[f], npUfupdated))
tf.reset_default_graph()
def cond(self) -> Exponential:
tau = self.tau
name = self.name + "Cond"
properties = Properties(name=name,
drawType=self.drawType,
updateType=self.updateType,
persistent=False)
cond = Exponential(beta=1. / tau, properties=properties)
return (cond)
def cond(self) -> CenLaplace:
tau = self.tau
name = self.name + "Cond"
properties = Properties(name=name,
drawType=self.drawType,
updateType=self.updateType,
persistent=False)
cond = CenLaplace(beta=1. / tau, properties=properties)
return (cond)
def productParams(self, d0: Distribution, d1: Distribution):
name = self.name(d0, d1)
drawType = self.drawType(d0, d1)
updateType = self.updateType(d0, d1)
persistent = False
properties = Properties(name=name,
drawType=drawType,
updateType=updateType,
persistent=persistent)
params = {"properties": properties}
return (params)
def __init__(self, M: Tuple[int, ...], K: int=1, tau: float = 1./1e2,
drawType: DrawType = DrawType.SAMPLE,
updateType: UpdateType = UpdateType.ALL,
dtype=tf.float32) -> None:
Likelihood.__init__(self, M, K)
self.__tauInit = tau
self.__dtype = dtype
self.__properties = Properties(name='likelihood',
drawType=drawType,
dtype=dtype,
updateType=updateType,
persistent=True)
def cond(self) -> Normal:
mu = self.mu
Psi = self.Psi
tau = self.tau
mu = tf.ones_like(tau) * mu
Psi = tf.ones_like(tau) * Psi
name = self.name + "Cond"
properties = Properties(name=name,
drawType=self.drawType,
updateType=self.updateType,
persistent=False)
cond = Normal(mu=mu, tau=tau / Psi, properties=properties)
return (cond)
def lhUfk(self, Uf: Tensor, prepVars: Tuple[Tensor, ...], f: int,
k: Tensor) -> Distribution:
XVT, VVT, alpha = prepVars
XvT = XVT[:, k]
VvT = VVT[..., k]
vvT = VVT[..., k, k]
Ufk = Uf[k]
UVvT = tf.reduce_sum(tf.transpose(Uf) * VvT, axis=-1)
uvvT = Ufk * vvT
Xtildev = XvT - UVvT + uvvT
mu = Xtildev / vvT
tau = vvT * alpha
properties = Properties(name=f"lhU{f}k",
drawType=self.noiseDistribution.drawType,
updateType=self.noiseDistribution.updateType,
persistent=False)
lhUfk = Normal(mu=mu, tau=tau, properties=properties)
return (lhUfk)
def cond(self) -> CenNnFullyElasticNetCond:
b = self.b
mu = self.mu
tau = self.tau
betaExponential = self.betaExponential
tauLomax = self.tauLomax
b = tf.ones_like(tauLomax) * b
mu = tf.ones_like(tauLomax) * mu
tau = tf.ones_like(tauLomax) * tau
betaExponential = tf.ones_like(tauLomax) * betaExponential
name = self.name + "Cond"
properties = Properties(name=name,
drawType=self.drawType,
updateType=self.updateType,
persistent=False)
cond = CenNnFullyElasticNetCond(b=b,
mu=mu,
tau=tau,
betaExponential=betaExponential,
beta=1. / tauLomax,
properties=properties)
return (cond)