|
Explorer = class Explorer(builtins.object) |
| |
Explorer(ns, threads=False)
Explorer is a helper class of NestedSampler, which contains and runs the
diffusion engines.
It uses Threads to parallelise the diffusion engines.
Attributes
----------
walkers : WalkerList
walkers to be explored
engines : [engine]
list of engines to be used
errdis : ErrorDistribution
to be used
rng : numpy.random.RandomState
random number generator
rate : float (1.0)
governs processing speed (vs precision)
maxtrials : int (5)
number of trials
verbose : int (0)
level of blabbering
lowLhood : float
present low likelihood level
iteration : int
counting explorer calls
Author Do Kester.
|
| |
Constructor:
- Explorer( ns, threads=False )
- Construct Explorer from a NestedSampler object.
Parameters
----------
ns : NestedSampler
the calling NestedSampler. It provides the attributes.
Methods defined here:
- allEngines( iteration )
- Always use all engines.
Parameters
----------
iteration : int
iteration number
- checkWalkers()
- explore( worst, lowLhood, iteration )
- Explore the likelihood function, using threads.
Parameters
----------
worst : [int]
list of walkers to be explored/updated
lowLhood : float
level of the low likelihood
- exploreWalker( kw, lowLhood, engines, rng )
- Move the walker around until it is randomly distributed over the prior and
higher in logL then lowLhood
Parameters
----------
kw : int
index in walkerlist, of the walker to be explored
lowLhood : float
minimum value for the log likelihood
engine : list of Engine
to be used
rng : RandomState
random number generator
- logLcheck( walker )
- Sanity check when no moves are found, if the LogL is still the same as the stored logL.
Parameters
----------
walker : Walker
the one with the stored logL
Raises
------
ValueError at inconsistency.
- selEngines( iteration )
- Select engines with slowly changing parameters once per so many iterations.
Parameter
---------
iteration : int
iteration number
|
|