#!/usr/bin/env python2
# -*- coding: utf-8 -*-
'''
Created on Wed Jan 17 09:08:13 2018
@author: prmiles
'''
# import required packages
import numpy as np
from datetime import datetime
[docs]class SimulationOptions:
'''
MCMC simulation options.
Attributes:
* :meth:`~define_simulation_options`
* :meth:`~display_simulation_options`
'''
def __init__(self):
# initialize simulation option variables
# self.options = BaseSimulationOptions()
self.description = 'simulation_options'
self.__options_set = False
[docs] def define_simulation_options(self, nsimu=int(1e4), adaptint=None, ntry=None, method='dram',
printint=None, adaptend=0, lastadapt=0, burnintime=0,
waitbar=True, debug=0, qcov=None, updatesigma=False,
stats=0, drscale=np.array([5, 4, 3], dtype=float),
adascale=None, savesize=0, maxmem=0, chainfile='chainfile',
s2chainfile='s2chainfile', sschainfile='sschainfile',
covchainfile='covchainfile', savedir=None,
save_to_bin=False, skip=1, label=None, RDR=None, verbosity=1, maxiter=None,
priorupdatestart=0, qcov_adjust=1e-8, burnin_scale=10,
alphatarget=0.234, etaparam=0.7, initqcovn=None,
doram=None, rndseq=None, results_filename=None,
save_to_json=False, save_to_txt=False,
json_restart_file=None, save_lightly=False):
'''
Define simulation options.
Args:
* **nsimu** (:py:class:`int`): Number of parameter samples to simulate. Default is 1e4.
* **adaptint** (:py:class:`int`): Number of interates between adaptation. Default is method dependent.
* **ntry** (:py:class:`int`): Number of tries to take before rejection. Default is method dependent.
* **method** (:py:class:`str`): Sampling method (:code:`'mh', 'am', 'dr', 'dram'`).\
Default is :code:`'dram'`.
* **printint** (:py:class:`int`): Printing interval.
* **adaptend** (:py:class:`int`): Obsolete.
* **lastadapt** (:py:class:`int`): Last adaptation iteration (i.e., no more adaptation beyond this point).
* **burnintime** (:py:class:`int`):
* **waitbar** (:py:class:`int`): Flag to use progress bar. Default is 1 -> on (otherwise -> off).
* **debug** (:py:class:`int`): Flag to perform debug. Default is 0 -> off.
* **qcov** (:class:`~numpy.ndarray`): Proposal parameter covariance matrix.
* **updatesigma** (:py:class:`bool`): Flag for updating measurement error variance.\
Default is 0 -> off (1 -> on).
* **stats** (:py:class:`int`): Calculate convergence statistics. Default is 0 -> off (1 -> on).
* **drscale** (:class:`~numpy.ndarray`): Reduced scale for sampling in DR algorithm. Default is [5,4,3].
* **adascale** (:py:class:`float`): User defined covariance scale. Default is method dependent (untested).
* **savesize** (:py:class:`int`): Size of chain segments when saving to log files. Default is 0.
* **maxmem** (:py:class:`int`): Maximum memory available in mega bytes (Obsolete).
* **chainfile** (:py:class:`str`): File name for :code:`chain` log file.
* **sschainfile** (:py:class:`str`): File name for :code:`sschain` log file.
* **s2chainfile** (:py:class:`str`): File name for :code:`s2chain` log file.
* **covchainfile** (:py:class:`str`): File name for :code:`qcov` log file.
* **savedir** (:py:class:`str`): Output directory of log files. Default is current directory.
* **save_to_bin** (:py:class:`bool`): Save log files to binary. Default is False.
* **save_to_txt** (:py:class:`bool`): Save log files to text. Default is False.
* **skip** (:py:class:`int`):
* **label** (:py:class:`str`):
* **RDR** (:class:`~numpy.ndarray`): R matrix for each stage of DR.
* **verbosity** (:py:class:`int`): Verbosity of display output.
* **maxiter** (:py:class:`int`): Obsolete.
* **priorupdatestart**
* **qcov_adjust** (:py:class:`float`): Adjustment scale for covariance matrix.
* **burnin_scale** (:py:class:`float`): Scale for burnin.
* **alphatarget** (:py:class:`float`): Acceptance ratio target.
* **etaparam** (:py:class:`float`):
* **initqcovn** (:py:class:`float`): Proposal covariance weight in update.
* **doram** (:py:class:`int`): Flag to perform :code:`'ram'` algorithm (Obsolete).
* **rndseq** (:class:`~numpy.ndarray`): Random number sequence (Obsolete).
* **results_filename** (:py:class:`str`): Output file name when saving results structure with json.
* **save_to_json** (:py:class:`bool`): Save results structure to json file. Default is False.
* **json_restart_file** (:py:class:`str`): Extract parameter covariance and last sample\
value from saved json file.
* **save_lightly** (:py:class:`bool`): Flag to indicate whether
results json file should include chains.
.. note::
For the log file names :code:`chainfile, sschainfile, s2chainfile` and :code:`covchainfile` \
do not include the extension.
By specifying whether to save to text or to binary, the appropriate extension will be added.
'''
method_dictionary = {
'mh': {'adaptint': 0, 'ntry': 1, 'doram': 0, 'adascale': adascale},
'am': {'adaptint': 100, 'ntry': 1, 'doram': 0, 'adascale': adascale},
'dr': {'adaptint': 0, 'ntry': 2, 'doram': 0, 'adascale': adascale},
'dram': {'adaptint': 100, 'ntry': 2, 'doram': 0, 'adascale': adascale},
'ram': {'adaptint': 1, 'ntry': 1, 'doram': 1, 'adascale': 1.},
}
# define items from dictionary
if adaptint is None:
self.adaptint = method_dictionary[method]['adaptint'] # update interval for adaptation
elif method == 'mh' or method == 'dr':
self.adaptint = method_dictionary[method]['adaptint'] # no adaptation - enforce!
else:
self.adaptint = adaptint
if ntry is None:
self.ntry = method_dictionary[method]['ntry']
else:
self.ntry = ntry
if adascale is None:
self.adascale = method_dictionary[method]['adascale'] # qcov_scale
else:
self.adascale = adascale
if doram is None:
self.doram = method_dictionary[method]['doram']
else:
self.doram = doram
self.nsimu = int(nsimu) # length of chain to simulate
self.method = method
self.dodram = 0
self.printint = printint # print interval
self.adaptend = adaptend # last adapt
self.lastadapt = lastadapt # last adapt
self.burnintime = burnintime
self.waitbar = bool(waitbar) # use waitbar
self.debug = debug # show some debug information
self.qcov = qcov # proposal covariance
self.initqcovn = initqcovn # proposal covariance weight in update
self.updatesigma = bool(updatesigma) #
self.priorupdatestart = priorupdatestart
self.qcov_adjust = qcov_adjust # eps adjustment
self.burnin_scale = burnin_scale
self.alphatarget = alphatarget # acceptance ratio target
self.etaparam = etaparam #
self.stats = stats # convergence statistics
self.drscale = drscale
self.skip = skip
datestr = datetime.now().strftime('%Y%m%d_%H%M%S')
self.datestr = datestr
if label is None:
self.label = str('MCMC run at {}'.format(datestr))
else:
self.label = label
self.RDR = RDR
self.verbosity = verbosity # amount of information to print
self.maxiter = maxiter
# log settings
self.savesize = savesize
self.maxmem = maxmem
self.chainfile = chainfile
self.s2chainfile = s2chainfile
self.sschainfile = sschainfile
self.covchainfile = covchainfile
if savedir is None:
self.savedir = str('{}_{}'.format(datestr, 'chain_log'))
else:
self.savedir = savedir
self.save_to_bin = save_to_bin
self.save_to_txt = save_to_txt
self.results_filename = results_filename
self.save_to_json = save_to_json
self.json_restart_file = json_restart_file
self.save_lightly = check_lightly_save(save_lightly, save_to_json,
save_to_bin, save_to_txt)
self.__options_set = True # options have been defined
def _check_dependent_simulation_options(self, model):
'''
Check dependent parameters.
- Checks that :code:`savesize` is between 0 and :code:`nsimu`.
- If :code:`ntry` greater than 1, then assume delayed rejection was wanted.
- If :code:`lastadapt` less than 1, then set equal to :code:`nsimu`.
- Update :code:`printint` based on size of :code:`adaptint`.
- If :code:`N0` not None, turn on :code:`updatesigma`.
Args:
* **model**: (:class:`~.ModelSettings`): MCMC model settings.
'''
# save options
if self.savesize <= 0 or self.savesize > self.nsimu:
self.savesize = self.nsimu
# turn on DR if ntry > 1
if self.ntry > 1:
self.dodram = 1
else:
self.dodram = 0
if self.lastadapt < 1:
self.lastadapt = self.nsimu
if self.printint is None:
self.printint = max(100, min(1000, self.adaptint))
# if N0 given, then also turn on updatesigma
if model.N0 is not None:
self.updatesigma = True
[docs] def display_simulation_options(self, print_these=None):
'''
Display subset of the simulation options.
Args:
* **print_these** (:py:class:`list`): List of strings corresponding to keywords. Default below.
::
print_these = ['nsimu', 'adaptint', 'ntry', 'method', 'printint', 'lastadapt', 'drscale', 'qcov']
'''
if print_these is None:
print_these = ['nsimu', 'adaptint', 'ntry', 'method', 'printint', 'lastadapt', 'drscale', 'qcov']
print('simulation options:')
for ptii in print_these:
print('\t{} = {}'.format(ptii, getattr(self, ptii)))
return print_these
[docs]def check_lightly_save(save_lightly, save_to_json, save_to_bin, save_to_txt):
'''
Check settings for lightly save
If saving to json, chains will be removed if already being stored via
one of the logging methods, binary or text. If logging methods not
being used, then chains will be included in json file.
Args:
* **save_lightly** (:py:class:`bool`): Flag to save results w/out arrays
* **save_to_json** (:py:class:`bool`): Flag to save to json
* **save_to_bin** (:py:class:`bool`): Flag to save to binary
* **save_to_txt** (:py:class:`bool`): Flag to save to text
'''
if save_to_json is True:
if save_to_bin is True or save_to_txt is True:
return True
else:
return False
else:
return save_lightly