def setUp(self):
# simple Gaussian-emission HMM
def model():
p_latent = pyro.param("p1", Variable(torch.Tensor([[0.7], [0.3]])))
p_obs = pyro.param("p2", Variable(torch.Tensor([[0.9], [0.1]])))
latents = [Variable(torch.ones(1, 1))]
observes = []
for t in range(self.model_steps):
latents.append(
pyro.sample(
"latent_{}".format(str(t)),
Bernoulli(
torch.index_select(p_latent, 0,
latents[-1].view(-1).long()))))
observes.append(
pyro.observe(
"observe_{}".format(str(t)),
Bernoulli(
torch.index_select(p_obs, 0,
latents[-1].view(-1).long())),
self.data[t]))
return torch.sum(torch.cat(latents))
self.model_steps = 3
self.data = [pyro.ones(1, 1) for _ in range(self.model_steps)]
self.model = model
def log_pdf(self, x):
"""
gamma log-likelihood
"""
ll_1 = -self.beta * x
ll_2 = (self.alpha - pyro.ones([1])) * torch.log(x)
ll_3 = self.alpha * torch.log(self.beta)
ll_4 = -log_gamma(self.alpha)
return ll_1 + ll_2 + ll_3 + ll_4
def _enter_poutine(self, *args, **kwargs):
"""
When model execution begins
"""
super(VIGuideCo, self)._enter_poutine(*args, **kwargs)
# trace structure:
# site = {"sample": sample, "logq": logq, "reparam": reparam}
self.trace = OrderedDict()
self.batch = []
# TODO: make this cleaner via a trace data structure
self.score_multiplier = pyro.ones(1)
def batch_log_pdf(self, x, batch_size=1):
"""
gamma log-likelihood
"""
if x.dim() == 1 and self.beta.dim() == 1 and batch_size == 1:
return self.log_pdf(x)
elif x.dim() == 1:
x = x.expand(batch_size, x.size(0))
ll_1 = -self.beta * x
ll_2 = (self.alpha - pyro.ones(x.size())) * torch.log(x)
ll_3 = self.alpha * torch.log(self.beta)
ll_4 = -log_gamma(self.alpha)
return ll_1 + ll_2 + ll_3 + ll_4
def log_gamma(xx):
"""
quick and dirty log gamma copied from webppl
"""
gamma_coeff = [
76.18009172947146, -86.50532032941677, 24.01409824083091,
-1.231739572450155, 0.1208650973866179e-2, -0.5395239384953e-5
]
magic1 = 1.000000000190015
magic2 = 2.5066282746310005
x = xx - 1.0
t = x + 5.5
t = t - (x + 0.5) * torch.log(t)
ser = pyro.ones(x.size()) * magic1
for c in gamma_coeff:
x = x + 1.0
ser = ser + torch.pow(x / c, -1)
return torch.log(ser * magic2) - t
def model():
ps = pyro.param("ps", Variable(torch.Tensor([[0.8], [0.3]])))
mu = pyro.param("mu", Variable(torch.Tensor([[-0.1], [0.9]])))
sigma = Variable(torch.ones(1, 1))
latents = [Variable(torch.ones(1))]
observes = []
for t in range(3):
latents.append(
pyro.sample("latent_{}".format(str(t)),
Bernoulli(ps[latents[-1][0].long().data])))
observes.append(
pyro.observe("observe_{}".format(str(t)),
Normal(mu[latents[-1][0].long().data], sigma),
pyro.ones(1)))
return latents
def model():
ps = pyro.param("ps", Variable(torch.Tensor([[0.8], [0.3]])))
mu = pyro.param("mu", Variable(torch.Tensor([[-0.1], [0.9]])))
sigma = Variable(torch.ones(1, 1))
latents = [Variable(torch.ones(1))]
observes = []
for t in range(3):
latents.append(
pyro.sample("latent_{}".format(str(t)),
Bernoulli(ps[latents[-1][0].long().data])))
observes.append(
pyro.observe("observe_{}".format(str(t)),
Normal(mu[latents[-1][0].long().data], sigma),
pyro.ones(1)))
return latents
def setUp(self):
# simple Gaussian-emission HMM
def model():
p_latent = pyro.param("p1", Variable(torch.Tensor([[0.7], [0.3]])))
p_obs = pyro.param("p2", Variable(torch.Tensor([[0.9], [0.1]])))
latents = [Variable(torch.ones(1, 1))]
observes = []
for t in range(self.model_steps):
latents.append(
pyro.sample("latent_{}".format(str(t)),
Bernoulli(torch.index_select(p_latent, 0, latents[-1].view(-1).long()))))
observes.append(
pyro.observe("observe_{}".format(str(t)),
Bernoulli(torch.index_select(p_obs, 0, latents[-1].view(-1).long())),
self.data[t]))
return torch.sum(torch.cat(latents))
self.model_steps = 3
self.data = [pyro.ones(1, 1) for _ in range(self.model_steps)]
self.model = model