def reweight_particles(self):
wgts = self.wgts
logG = self.fk.logG(self.t, self.X)
new_wgts = wgts.add(logG)
loglt = rs.log_mean_exp(new_wgts.lw) - rs.log_mean_exp(wgts.lw)
self.loglt += [loglt]
self.logLt += loglt
self.wgts = new_wgts
def reset_weights(self):
"""Reset weights after a resampling step.
"""
if self.fk.isAPF:
lw = (rs.log_mean_exp(self.logetat, W=self.W) -
self.logetat[self.A])
self.wgts = rs.Weights(lw=lw)
else:
self.wgts = rs.Weights()
def compute_summaries(self):
if self.t > 0:
prec_log_mean_w = self.log_mean_w
self.log_mean_w = rs.log_mean_exp(self.wgts.lw)
if self.t==0 or self.rs_flag:
self.loglt = self.log_mean_w
else:
self.loglt = self.log_mean_w - prec_log_mean_w
self.logLt += self.loglt
if self.verbose:
print(self)
if self.summaries:
self.summaries.collect(self)
if self.hist:
self.hist.save(X=self.X, w=self.wgts, A=self.A)
def compute_summaries(self):
if self.t > 0:
prec_log_mean_w = self.log_mean_w
self.log_mean_w = rs.log_mean_exp(self.wgts.lw)
if self.t == 0 or self.rs_flag:
self.loglt = self.log_mean_w
else:
self.loglt = self.log_mean_w - prec_log_mean_w
self.logLt += self.loglt
if self.verbose:
print(self)
if self.hist:
self.hist.save(self)
# must collect summaries *after* history, because a collector (e.g.
# FixedLagSmoother) may needs to access history
if self.summaries:
self.summaries.collect(self)
def run(self, N=100):
"""
Parameter
---------
N: int
number of particles
Returns
-------
wgts: Weights object
The importance weights (with attributes lw, W, and ESS)
X: ThetaParticles object
The N particles (with attributes theta, logpost)
norm_cst: float
Estimate of the normalising constant of the target
"""
th = self.proposal.rvs(size=N)
self.X = ThetaParticles(theta=th, lpost=None)
self.X.lpost = self.model.logpost(th)
lw = self.X.lpost - self.proposal.logpdf(th)
self.wgts = rs.Weights(lw=lw)
self.norm_cst = rs.log_mean_exp(lw)
def logLT(self) -> float:
"""
returns the local logLT (for debugging)
"""
# noinspection PyTypeChecker
return rs.log_mean_exp(self.logLTs_of_M1)
def logLT(self) -> float:
"""
Calculate the log of the normalizing constant of the model. Can be useful for debugging (i.e., by comparing its with the estimates returned by the SMCNew executor).
"""
return sum([rs.log_mean_exp(s) for s in self._logG])
