def _get_mc_stats(self, op):
loc = _np.empty(self._samples.shape[0:2], dtype=_np.complex128)
for i, sample in enumerate(self._samples):
_local_values(op, self._machine, sample, out=loc[i])
# notice that loc.T is passed to statistics, since that function assumes
# that the first index is the batch index.
return loc, _statistics(loc.T)
def _get_mc_superop_stats(self, op):
samples_r = self._samples.reshape((-1, self._samples.shape[-1]))
loc = _local_values(op, self._machine,
samples_r).reshape(self._samples.shape[0:2])
# notice that loc.T is passed to statistics, since that function assumes
# that the first index is the batch index.
return loc, _statistics(abs(loc.T)**2)
def estimate(op):
lvs = _local_values(op, psi, samples)
stats = nst.statistics(lvs)
if compute_gradients:
grad = nst.covariance_sv(lvs, der_logs)
return stats, grad
else:
return stats
def _get_mc_obs_stats(self, op):
if not self._obs_samples_valid:
self.sweep_diagonal()
samples_r = self._samples_obs.reshape(
(-1, self._samples_obs.shape[-1]))
loc = _local_values(op, self._machine,
samples_r).reshape(self._samples_obs.shape[0:2])
# notice that loc.T is passed to statistics, since that function assumes
# that the first index is the batch index.
return loc, _statistics(loc.T)
def estimate(op):
lvs = _local_values(op, psi, samples)
stats = _statistics(lvs.T)
if compute_gradients:
samples_r = samples.reshape((-1, samples.shape[-1]))
eloc_r = (lvs - _mean(lvs)).reshape(-1, 1)
grad = sampler.machine.vector_jacobian_prod(
samples_r,
eloc_r / n_samples,
)
return stats, grad
else:
return stats
