def setup_logp(self):
"""Set up the prior and likelihood logp functions, and derivatives."""
shared = make_shared_replacements(self.variables, self.model)
self.prior_logp_func = logp_forw([self.model.varlogpt], self.variables,
shared)
self.prior_dlogp_func = logp_forw(
[gradient(self.model.varlogpt, self.variables)], self.variables,
shared)
self.likelihood_logp_func = logp_forw([self.model.datalogpt],
self.variables, shared)
self.posterior_logp_func = logp_forw([self.model.logpt],
self.variables, shared)
self.posterior_dlogp_func = logp_forw(
[gradient(self.model.logpt, self.variables)], self.variables,
shared)
self.posterior_hessian_func = logp_forw(
[hessian(self.model.logpt, self.variables)], self.variables,
shared)
self.posterior_logp_nojac = logp_forw([self.model.logp_nojact],
self.variables, shared)
self.posterior_dlogp_nojac = logp_forw(
[gradient(self.model.logp_nojact, self.variables)], self.variables,
shared)
self.posterior_hessian_nojac = logp_forw(
[hessian(self.model.logp_nojact, self.variables)], self.variables,
shared)
def setup_logp(self):
"""Set up the likelihood logp function based on the chosen kernel."""
shared = make_shared_replacements(self.variables, self.model)
self.prior_logp_func = logp_forw([self.model.varlogpt], self.variables,
shared)
self.likelihood_logp_func = logp_forw([self.model.datalogpt],
self.variables, shared)
self.posterior_logp_func = logp_forw([self.model.logpt],
self.variables, shared)
self.posterior_dlogp_func = logp_forw(
[gradient(self.model.logpt, self.variables)], self.variables,
shared)
self.prior_dlogp_func = logp_forw(
[gradient(self.model.varlogpt, self.variables)], self.variables,
shared)
self.likelihood_dlogp_func = logp_forw(
[gradient(self.model.datalogpt, self.variables)], self.variables,
shared)
def _theano_hamiltonian(model_vars, shared, logpt, potential):
"""Creates a Hamiltonian with shared inputs.
Parameters
----------
model_vars : array of variables to be sampled
shared : theano tensors that are already shared
logpt : model log probability
potential : hamiltonian potential
Returns
-------
Hamiltonian : namedtuple with log pdf, gradient of log pdf, and potential functions
q : Starting position variable.
"""
dlogp = gradient(logpt, model_vars)
(logp, dlogp), q = join_nonshared_inputs([logpt, dlogp], model_vars,
shared)
logp = CallableTensor(logp)
dlogp = CallableTensor(dlogp)
return Hamiltonian(logp, dlogp, potential), q
def _theano_hamiltonian(model_vars, shared, logpt, potential):
"""Creates a Hamiltonian with shared inputs.
Parameters
----------
model_vars : array of variables to be sampled
shared : theano tensors that are already shared
logpt : model log probability
potential : hamiltonian potential
Returns
-------
Hamiltonian : namedtuple with log pdf, gradient of log pdf, and potential functions
q : Starting position variable.
"""
dlogp = gradient(logpt, model_vars)
(logp, dlogp), q = join_nonshared_inputs([logpt, dlogp], model_vars, shared)
dlogp_func = theano.function(inputs=[q], outputs=dlogp)
dlogp_func.trust_input = True
logp = CallableTensor(logp)
dlogp = CallableTensor(dlogp)
return Hamiltonian(logp, dlogp, potential), q, dlogp_func
def _theano_hamiltonian(model_vars, shared, logpt, potential):
"""Create a Hamiltonian with shared inputs.
Parameters
----------
model_vars : array of variables to be sampled
shared : theano tensors that are already shared
logpt : model log probability
potential : hamiltonian potential
Returns
-------
Hamiltonian : namedtuple with log pdf, gradient of log pdf, and potential functions
q : Initial position for Hamiltonian Monte Carlo
dlogp_func: theano function that computes the gradient of a log pdf at a point
"""
dlogp = gradient(logpt, model_vars)
(logp, dlogp), q = join_nonshared_inputs([logpt, dlogp], model_vars, shared)
dlogp_func = theano.function(inputs=[q], outputs=dlogp)
dlogp_func.trust_input = True
logp = CallableTensor(logp)
dlogp = CallableTensor(dlogp)
return Hamiltonian(logp, dlogp, potential), q, dlogp_func
def jacobian_det(self, x):
grad = tt.reshape(gradient(tt.sum(self.backward(x)), [x]), x.shape)
return tt.log(tt.abs_(grad))
