def importance_sample(self, logL_new, action='add', inplace=False):
"""Perform importance re-weighting on the log-likelihood.
Parameters
----------
logL_new: np.array
New log-likelihood values. Should have the same shape as `logL`.
action: str, optional
Can be any of {'add', 'replace', 'mask'}.
* add: Add the new `logL_new` to the current `logL`.
* replace: Replace the current `logL` with the new `logL_new`.
* mask: treat `logL_new` as a boolean mask and only keep the
corresponding (True) samples.
default: 'add'
inplace: bool, optional
Indicates whether to modify the existing array, or return a new
frame with importance sampling applied.
default: False
Returns
-------
samples: NestedSamples
Importance re-weighted samples.
"""
samples = super().importance_sample(logL_new, action=action)
samples = samples[samples.logL > samples.logL_birth].recompute()
return modify_inplace(self, samples, inplace)
def recompute(self, logL_birth=None, inplace=False):
"""Re-calculate the nested sampling contours and live points.
Parameters
----------
logL_birth: array-like or int, optional
array-like: the birth contours.
int: the number of live points.
default: use the existing birth contours to compute nlive
inplace: bool, optional
Indicates whether to modify the existing array, or return a new
frame with contours resorted and nlive recomputed
default: False
"""
samples = self.sort_values('logL').reset_index(drop=True)
if is_int(logL_birth):
nlive = logL_birth
samples['nlive'] = nlive
descending = np.arange(nlive, 0, -1)
samples.loc[len(samples) - nlive:, 'nlive'] = descending
else:
if logL_birth is not None:
samples['logL_birth'] = logL_birth
samples.tex['logL_birth'] = r'$\log\mathcal{L}_{\rm birth}$'
if 'logL_birth' not in samples:
raise RuntimeError("Cannot recompute run without "
"birth contours logL_birth.")
invalid = samples.logL <= samples.logL_birth
n_bad = invalid.sum()
n_equal = (samples.logL == samples.logL_birth).sum()
if n_bad:
warnings.warn(
"%i out of %i samples have logL <= logL_birth,"
"\n%i of which have logL == logL_birth."
"\nThis may just indicate numerical rounding "
"errors at the peak of the likelihood, but "
"further investigation of the chains files is "
"recommended."
"\nDropping the invalid samples." %
(n_bad, len(samples), n_equal), RuntimeWarning)
samples = samples[~invalid].reset_index(drop=True)
samples['nlive'] = compute_nlive(samples.logL, samples.logL_birth)
samples.tex['nlive'] = r'$n_{\rm live}$'
samples.beta = samples._beta
return modify_inplace(self, samples, inplace)
def importance_sample(self, logL_new, action='add', inplace=False):
"""Perform importance re-weighting on the log-likelihood.
Parameters
----------
logL_new: np.array
New log-likelihood values. Should have the same shape as `logL`.
action: str, optional
Can be any of {'add', 'replace', 'mask'}.
* add: Add the new `logL_new` to the current `logL`.
* replace: Replace the current `logL` with the new `logL_new`.
* mask: treat `logL_new` as a boolean mask and only keep the
corresponding (True) samples.
default: 'add'
inplace: bool, optional
Indicates whether to modify the existing array, or return a new
frame with importance sampling applied.
default: False
Returns
-------
samples: MCMCSamples
Importance re-weighted samples.
"""
samples = self.copy()
if action == 'add':
samples.weights *= np.exp(logL_new - logL_new.max())
samples.logL += logL_new
elif action == 'replace':
logL_new2 = logL_new - samples.logL
samples.weights *= np.exp(logL_new2 - logL_new2.max())
samples.logL = logL_new
elif action == 'mask':
samples = samples[logL_new]
else:
raise NotImplementedError("`action` needs to be one of "
"{'add', 'replace', 'mask'}, but '%s' "
"was requested." % action)
return modify_inplace(self, samples, inplace)
def recompute(self, logL_birth=None, inplace=False):
"""Re-calculate the nested sampling contours and live points.
Parameters
----------
logL_birth, array-like or int, optional
array-like: the birth contours.
int: the number of live points.
default: use the existing birth contours to compute nlive
inplace: bool, optional
Indicates whether to modify the existing array, or return a new
frame with contours resorted and nlive recomputed
default: False
"""
samples = self.sort_values('logL').reset_index(drop=True)
if is_int(logL_birth):
nlive = logL_birth
samples['nlive'] = nlive
descending = np.arange(nlive, 0, -1)
samples.loc[len(samples) - nlive:, 'nlive'] = descending
else:
if logL_birth is not None:
samples['logL_birth'] = logL_birth
samples.tex['logL_birth'] = r'$\log\mathcal{L}_{\rm birth}$'
if 'logL_birth' not in samples:
raise RuntimeError("Cannot recompute run without "
"birth contours logL_birth.")
if (samples.logL <= samples.logL_birth).any():
raise RuntimeError("Not a valid nested sampling run. "
"Require logL > logL_birth.")
samples['nlive'] = compute_nlive(samples.logL, samples.logL_birth)
samples.tex['nlive'] = r'$n_{\rm live}$'
samples.beta = samples._beta
return modify_inplace(self, samples, inplace)