mcmc_based_searches.py 90.7 KB
Newer Older
Gregory Ashton's avatar
Gregory Ashton committed
1
""" Searches using MCMC-based methods """
2
from __future__ import division, absolute_import, print_function
Gregory Ashton's avatar
Gregory Ashton committed
3

4
import sys
Gregory Ashton's avatar
Gregory Ashton committed
5
import os
6
import copy
Gregory Ashton's avatar
Gregory Ashton committed
7
import logging
8
from collections import OrderedDict
9
import subprocess
10
11
12
13
14
15
16
17

import numpy as np
import matplotlib
import matplotlib.pyplot as plt
import emcee
import corner
import dill as pickle

18
import pyfstat.core as core
19
20
from pyfstat.core import tqdm, args, read_par
import pyfstat.optimal_setup_functions as optimal_setup_functions
21
import pyfstat.helper_functions as helper_functions
22
23


24
class MCMCSearch(core.BaseSearchClass):
Gregory Ashton's avatar
Gregory Ashton committed
25
    """MCMC search using ComputeFstat
26
27
28
29
30
31
32
33
34
35
36
37

    Parameters
    ----------
    label, outdir: str
        A label and directory to read/write data from/to
    theta_prior: dict
        Dictionary of priors and fixed values for the search parameters.
        For each parameters (key of the dict), if it is to be held fixed
        the value should be the constant float, if it is be searched, the
        value should be a dictionary of the prior.
    tref, minStartTime, maxStartTime: int
        GPS seconds of the reference time, start time and end time
Gregory Ashton's avatar
Gregory Ashton committed
38
39
40
41
42
43
    sftfilepattern: str
        Pattern to match SFTs using wildcards (*?) and ranges [0-9];
        mutiple patterns can be given separated by colons.
    detectors: str
        Two character reference to the detectors to use, specify None for no
        contraint and comma separate for multiple references.
44
45
46
47
    nsteps: list (2,)
        Number of burn-in and production steps to take, [nburn, nprod]. See
        `pyfstat.MCMCSearch.setup_initialisation()` for details on adding
        initialisation steps.
48
49
50
51
52
    nwalkers, ntemps: int,
        The number of walkers and temperates to use in the parallel
        tempered PTSampler.
    log10temperature_min float < 0
        The  log_10(tmin) value, the set of betas passed to PTSampler are
Gregory Ashton's avatar
Gregory Ashton committed
53
54
        generated from `np.logspace(0, log10temperature_min, ntemps)`.
    theta_initial: dict, array, (None)
55
56
        A dictionary of distribution about which to distribute the
        initial walkers about
Gregory Ashton's avatar
Gregory Ashton committed
57
    rhohatmax: float,
58
59
60
        Upper bound for the SNR scale parameter (required to normalise the
        Bayes factor) - this needs to be carefully set when using the
        evidence.
Gregory Ashton's avatar
Gregory Ashton committed
61
    binary: bool
62
        If true, search over binary parameters
Gregory Ashton's avatar
Gregory Ashton committed
63
64
65
66
    BSGL: bool
        If true, use the BSGL statistic
    SSBPrec: int
        SSBPrec (SSB precision) to use when calling ComputeFstat
67
68
69
    minCoverFreq, maxCoverFreq: float
        Minimum and maximum instantaneous frequency which will be covered
        over the SFT time span as passed to CreateFstatInput
Gregory Ashton's avatar
Gregory Ashton committed
70
71
72
73
74
    injectSources: dict
        If given, inject these properties into the SFT files before running
        the search
    assumeSqrtSX: float
        Don't estimate noise-floors, but assume (stationary) per-IFO sqrt{SX}
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89

    Attributes
    ----------
    symbol_dictionary: dict
        Key, val pairs of the parameters (i.e. `F0`, `F1`), to Latex math
        symbols for plots
    unit_dictionary: dict
        Key, val pairs of the parameters (i.e. `F0`, `F1`), and the
        units (i.e. `Hz`)
    transform_dictionary: dict
        Key, val pairs of the parameters (i.e. `F0`, `F1`), where the key is
        itself a dictionary which can item `multiplier`, `subtractor`, or
        `unit` by which to transform by and update the units.

    """
90
91

    symbol_dictionary = dict(
92
        F0='$f$', F1='$\dot{f}$', F2='$\ddot{f}$', Alpha=r'$\alpha$',
93
94
        Delta='$\delta$', asini='asini', period='P', ecc='ecc', tp='tp',
        argp='argp')
95
    unit_dictionary = dict(
96
97
        F0='Hz', F1='Hz/s', F2='Hz/s$^2$', Alpha=r'rad', Delta='rad',
        asini='', period='s', ecc='', tp='', argp='')
98
    transform_dictionary = {}
99

Gregory Ashton's avatar
Gregory Ashton committed
100
    @helper_functions.initializer
Gregory Ashton's avatar
Gregory Ashton committed
101
    def __init__(self, label, outdir, theta_prior, tref, minStartTime,
Gregory Ashton's avatar
Gregory Ashton committed
102
103
104
                 maxStartTime, sftfilepattern=None, detectors=None,
                 nsteps=[100, 100], nwalkers=100, ntemps=1,
                 log10temperature_min=-5, theta_initial=None,
105
                 rhohatmax=1000, binary=False, BSGL=False,
Gregory Ashton's avatar
Gregory Ashton committed
106
                 SSBprec=None, minCoverFreq=None, maxCoverFreq=None,
107
                 injectSources=None, assumeSqrtSX=None):
108

Gregory Ashton's avatar
Gregory Ashton committed
109
110
        if os.path.isdir(outdir) is False:
            os.mkdir(outdir)
111
        self._add_log_file()
112
        logging.info('Set-up MCMC search for model {}'.format(self.label))
113
114
        if sftfilepattern:
            logging.info('Using data {}'.format(self.sftfilepattern))
115
        else:
116
            logging.info('No sftfilepattern given')
117
118
        if injectSources:
            logging.info('Inject sources: {}'.format(injectSources))
119
        self.pickle_path = '{}/{}_saved_data.p'.format(self.outdir, self.label)
120
        self._unpack_input_theta()
121
        self.ndim = len(self.theta_keys)
122
123
124
125
        if self.log10temperature_min:
            self.betas = np.logspace(0, self.log10temperature_min, self.ntemps)
        else:
            self.betas = None
126

127
128
129
        if args.clean and os.path.isfile(self.pickle_path):
            os.rename(self.pickle_path, self.pickle_path+".old")

130
        self._set_likelihoodcoef()
131
        self._log_input()
132
133
134

    def _set_likelihoodcoef(self):
        self.likelihoodcoef = np.log(70./self.rhohatmax**4)
135

136
    def _log_input(self):
137
        logging.info('theta_prior = {}'.format(self.theta_prior))
138
        logging.info('nwalkers={}'.format(self.nwalkers))
139
140
141
        logging.info('nsteps = {}'.format(self.nsteps))
        logging.info('ntemps = {}'.format(self.ntemps))
        logging.info('log10temperature_min = {}'.format(
142
            self.log10temperature_min))
143

144
    def _initiate_search_object(self):
145
        logging.info('Setting up search object')
146
        self.search = core.ComputeFstat(
147
            tref=self.tref, sftfilepattern=self.sftfilepattern,
148
            minCoverFreq=self.minCoverFreq, maxCoverFreq=self.maxCoverFreq,
149
            detectors=self.detectors, BSGL=self.BSGL, transient=False,
150
            minStartTime=self.minStartTime, maxStartTime=self.maxStartTime,
151
            binary=self.binary, injectSources=self.injectSources,
152
            assumeSqrtSX=self.assumeSqrtSX, SSBprec=self.SSBprec)
153
154

    def logp(self, theta_vals, theta_prior, theta_keys, search):
155
        H = [self._generic_lnprior(**theta_prior[key])(p) for p, key in
156
157
158
159
160
161
             zip(theta_vals, theta_keys)]
        return np.sum(H)

    def logl(self, theta, search):
        for j, theta_i in enumerate(self.theta_idxs):
            self.fixed_theta[theta_i] = theta[j]
162
163
        FS = search.compute_fullycoherent_det_stat_single_point(
            *self.fixed_theta)
164
        return FS + self.likelihoodcoef
165

166
    def _unpack_input_theta(self):
167
        full_theta_keys = ['F0', 'F1', 'F2', 'Alpha', 'Delta']
168
169
170
        if self.binary:
            full_theta_keys += [
                'asini', 'period', 'ecc', 'tp', 'argp']
171
172
        full_theta_keys_copy = copy.copy(full_theta_keys)

173
174
        full_theta_symbols = ['$f$', '$\dot{f}$', '$\ddot{f}$', r'$\alpha$',
                              r'$\delta$']
175
176
        if self.binary:
            full_theta_symbols += [
177
                'asini', 'period', 'ecc', 'tp', 'argp']
178

179
180
        self.theta_keys = []
        fixed_theta_dict = {}
181
        for key, val in self.theta_prior.iteritems():
182
183
            if type(val) is dict:
                fixed_theta_dict[key] = 0
Gregory Ashton's avatar
Gregory Ashton committed
184
                self.theta_keys.append(key)
185
186
187
188
189
190
            elif type(val) in [float, int, np.float64]:
                fixed_theta_dict[key] = val
            else:
                raise ValueError(
                    'Type {} of {} in theta not recognised'.format(
                        type(val), key))
Gregory Ashton's avatar
Gregory Ashton committed
191
            full_theta_keys_copy.pop(full_theta_keys_copy.index(key))
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206

        if len(full_theta_keys_copy) > 0:
            raise ValueError(('Input dictionary `theta` is missing the'
                              'following keys: {}').format(
                                  full_theta_keys_copy))

        self.fixed_theta = [fixed_theta_dict[key] for key in full_theta_keys]
        self.theta_idxs = [full_theta_keys.index(k) for k in self.theta_keys]
        self.theta_symbols = [full_theta_symbols[i] for i in self.theta_idxs]

        idxs = np.argsort(self.theta_idxs)
        self.theta_idxs = [self.theta_idxs[i] for i in idxs]
        self.theta_symbols = [self.theta_symbols[i] for i in idxs]
        self.theta_keys = [self.theta_keys[i] for i in idxs]

207
    def _check_initial_points(self, p0):
208
209
210
211
212
213
214
215
216
217
218
219
220
        for nt in range(self.ntemps):
            logging.info('Checking temperature {} chains'.format(nt))
            initial_priors = np.array([
                self.logp(p, self.theta_prior, self.theta_keys, self.search)
                for p in p0[nt]])
            number_of_initial_out_of_bounds = sum(initial_priors == -np.inf)

            if number_of_initial_out_of_bounds > 0:
                logging.warning(
                    'Of {} initial values, {} are -np.inf due to the prior'
                    .format(len(initial_priors),
                            number_of_initial_out_of_bounds))

221
                p0 = self._generate_new_p0_to_fix_initial_points(
222
223
                    p0, nt, initial_priors)

224
    def _generate_new_p0_to_fix_initial_points(self, p0, nt, initial_priors):
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
        logging.info('Attempting to correct intial values')
        idxs = np.arange(self.nwalkers)[initial_priors == -np.inf]
        count = 0
        while sum(initial_priors == -np.inf) > 0 and count < 100:
            for j in idxs:
                p0[nt][j] = (p0[nt][np.random.randint(0, self.nwalkers)]*(
                             1+np.random.normal(0, 1e-10, self.ndim)))
            initial_priors = np.array([
                self.logp(p, self.theta_prior, self.theta_keys,
                          self.search)
                for p in p0[nt]])
            count += 1

        if sum(initial_priors == -np.inf) > 0:
            logging.info('Failed to fix initial priors')
        else:
            logging.info('Suceeded to fix initial priors')

        return p0
244

245
246
    def setup_burnin_convergence_testing(
            self, n=10, test_type='autocorr', windowed=False, **kwargs):
247
        """ Set up convergence testing during the MCMC simulation
248
249
250

        Parameters
        ----------
251
252
253
254
255
256
257
258
259
        n: int
            Number of steps after which to test convergence
        test_type: str ['autocorr', 'GR']
            If 'autocorr' use the exponential autocorrelation time (kwargs
            passed to `get_autocorr_convergence`). If 'GR' use the Gelman-Rubin
            statistic (kwargs passed to `get_GR_convergence`)
        windowed: bool
            If True, only calculate the convergence test in a window of length
            `n`
260
261
262
263
        **kwargs:
            Passed to either `_test_autocorr_convergence()` or
            `_test_GR_convergence()` depending on `test_type`.

264
        """
265
        logging.info('Setting up convergence testing')
266
267
268
269
        self.convergence_n = n
        self.convergence_windowed = windowed
        self.convergence_test_type = test_type
        self.convergence_kwargs = kwargs
270
271
        self.convergence_diagnostic = []
        self.convergence_diagnosticx = []
272
        if test_type in ['autocorr']:
273
            self._get_convergence_test = self._test_autocorr_convergence
274
        elif test_type in ['GR']:
275
            self._get_convergence_test = self._test_GR_convergence
276
277
278
        else:
            raise ValueError('test_type {} not understood'.format(test_type))

279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
    def setup_initialisation(self, nburn0, scatter_val=1e-10):
        """ Add an initialisation step to the MCMC run

        If called prior to `run()`, adds an intial step in which the MCMC
        simulation is run for `nburn0` steps. After this, the MCMC simulation
        continues in the usual manner (i.e. for nburn and nprod steps), but the
        walkers are reset scattered around the maximum likelihood position
        of the initialisation step.

        Parameters
        ----------
        nburn0: int
            Number of initialisation steps to take
        scatter_val: float
            Relative number to scatter walkers around the maximum likelihood
            position after the initialisation step

        """

        logging.info('Setting up initialisation with nburn0={}, scatter_val={}'
                     .format(nburn0, scatter_val))
        self.nsteps = [nburn0] + self.nsteps
        self.scatter_val = scatter_val

303
    def _test_autocorr_convergence(self, i, sampler, test=True, n_cut=5):
304
305
306
307
308
309
310
311
312
313
314
        try:
            acors = np.zeros((self.ntemps, self.ndim))
            for temp in range(self.ntemps):
                if self.convergence_windowed:
                    j = i-self.convergence_n
                else:
                    j = 0
                x = np.mean(sampler.chain[temp, :, j:i, :], axis=0)
                acors[temp, :] = emcee.autocorr.exponential_time(x)
            c = np.max(acors, axis=0)
        except emcee.autocorr.AutocorrError:
Gregory Ashton's avatar
Gregory Ashton committed
315
316
317
318
            logging.info('Failed to calculate exponential autocorrelation')
            c = np.zeros(self.ndim) + np.nan
        except AttributeError:
            logging.info('Unable to calculate exponential autocorrelation')
319
320
321
322
323
324
325
326
            c = np.zeros(self.ndim) + np.nan

        self.convergence_diagnosticx.append(i - self.convergence_n/2.)
        self.convergence_diagnostic.append(list(c))

        if test:
            return i > n_cut * np.max(c)

327
    def _test_GR_convergence(self, i, sampler, test=True, R=1.1):
328
329
330
331
332
        if self.convergence_windowed:
            s = sampler.chain[0, :, i-self.convergence_n+1:i+1, :]
        else:
            s = sampler.chain[0, :, :i+1, :]
        N = float(self.convergence_n)
333
334
        M = float(self.nwalkers)
        W = np.mean(np.var(s, axis=1), axis=0)
335
336
        per_walker_mean = np.mean(s, axis=1)
        mean = np.mean(per_walker_mean, axis=0)
337
338
        B = N / (M-1.) * np.sum((per_walker_mean-mean)**2, axis=0)
        Vhat = (N-1)/N * W + (M+1)/(M*N) * B
339
        c = np.sqrt(Vhat/W)
340
        self.convergence_diagnostic.append(c)
341
        self.convergence_diagnosticx.append(i - self.convergence_n/2.)
342

343
344
345
        if test and np.max(c) < R:
            return True
        else:
346
            return False
347
348
349
350

    def _test_convergence(self, i, sampler, **kwargs):
        if np.mod(i+1, self.convergence_n) == 0:
            return self._get_convergence_test(i, sampler, **kwargs)
351
        else:
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
            return False

    def _run_sampler_with_conv_test(self, sampler, p0, nprod=0, nburn=0):
        logging.info('Running {} burn-in steps with convergence testing'
                     .format(nburn))
        iterator = tqdm(sampler.sample(p0, iterations=nburn), total=nburn)
        for i, output in enumerate(iterator):
            if self._test_convergence(i, sampler, test=True,
                                      **self.convergence_kwargs):
                logging.info(
                    'Converged at {} before max number {} of steps reached'
                    .format(i, nburn))
                self.convergence_idx = i
                break
        iterator.close()
        logging.info('Running {} production steps'.format(nprod))
        j = nburn
        iterator = tqdm(sampler.sample(output[0], iterations=nprod),
                        total=nprod)
        for result in iterator:
            self._test_convergence(j, sampler, test=False,
                                   **self.convergence_kwargs)
            j += 1
        return sampler
376

377
    def _run_sampler(self, sampler, p0, nprod=0, nburn=0):
378
379
        if hasattr(self, 'convergence_n'):
            self._run_sampler_with_conv_test(sampler, p0, nprod, nburn)
380
381
382
383
        else:
            for result in tqdm(sampler.sample(p0, iterations=nburn+nprod),
                               total=nburn+nprod):
                pass
384

385
386
        self.mean_acceptance_fraction = np.mean(
            sampler.acceptance_fraction, axis=1)
387
        logging.info("Mean acceptance fraction: {}"
388
                     .format(self.mean_acceptance_fraction))
389
        if self.ntemps > 1:
390
            self.tswap_acceptance_fraction = sampler.tswap_acceptance_fraction
391
392
393
            logging.info("Tswap acceptance fraction: {}"
                         .format(sampler.tswap_acceptance_fraction))
        try:
394
            self.autocorr_time = sampler.get_autocorr_time(c=4)
395
            logging.info("Autocorrelation length: {}".format(
396
                self.autocorr_time))
397
        except emcee.autocorr.AutocorrError as e:
398
            self.autocorr_time = np.nan
399
400
401
402
403
            logging.warning(
                'Autocorrelation calculation failed with message {}'.format(e))

        return sampler

404
    def _estimate_run_time(self):
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
        """ Print the estimated run time

        Uses timing coefficients based on a Lenovo T460p Intel(R)
        Core(TM) i5-6300HQ CPU @ 2.30GHz.

        """
        # Todo: add option to time on a machine, and move coefficients to
        # ~/.pyfstat.conf
        if (type(self.theta_prior['Alpha']) == dict or
                type(self.theta_prior['Delta']) == dict):
            tau0S = 7.3e-5
            tau0LD = 4.2e-7
        else:
            tau0S = 5.0e-5
            tau0LD = 6.2e-8
420
        Nsfts = (self.maxStartTime - self.minStartTime) / 1800.
421
422
423
        numb_evals = np.sum(self.nsteps)*self.nwalkers*self.ntemps
        a = tau0S * numb_evals
        b = tau0LD * Nsfts * numb_evals
424
425
426
        logging.info('Estimated run-time = {} s = {:1.0f}:{:1.0f} m'.format(
            a+b, *divmod(a+b, 60)))

427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
    def run(self, proposal_scale_factor=2, create_plots=True, c=5, **kwargs):
        """ Run the MCMC simulatation

        Parameters
        ----------
        proposal_scale_factor: float
            The proposal scale factor used by the sampler, see Goodman & Weare
            (2010). If the acceptance fraction is too low, you can raise it by
            decreasing the a parameter; and if it is too high, you can reduce
            it by increasing the a parameter [Foreman-Mackay (2013)].
        create_plots: bool
            If true, save trace plots of the walkers
        c: int
            The minimum number of autocorrelation times needed to trust the
            result when estimating the autocorrelation time (see
            emcee.autocorr.integrated_time for further details. Default is 5
        **kwargs:
            Passed to _plot_walkers to control the figures

446
447
448
449
450
        Returns
        -------
        sampler: emcee.ptsampler.PTSampler
            The emcee ptsampler object

451
        """
452

453
        self.old_data_is_okay_to_use = self._check_old_data_is_okay_to_use()
454
455
456
        if self.old_data_is_okay_to_use is True:
            logging.warning('Using saved data from {}'.format(
                self.pickle_path))
457
            d = self.get_saved_data_dictionary()
458
459
460
            self.samples = d['samples']
            self.lnprobs = d['lnprobs']
            self.lnlikes = d['lnlikes']
461
            self.all_lnlikelihood = d['all_lnlikelihood']
462
463
            return

464
        self._initiate_search_object()
465
        self._estimate_run_time()
466
467
468
469

        sampler = emcee.PTSampler(
            self.ntemps, self.nwalkers, self.ndim, self.logl, self.logp,
            logpargs=(self.theta_prior, self.theta_keys, self.search),
470
            loglargs=(self.search,), betas=self.betas, a=proposal_scale_factor)
471

472
473
474
        p0 = self._generate_initial_p0()
        p0 = self._apply_corrections_to_p0(p0)
        self._check_initial_points(p0)
475

476
        # Run initialisation steps if required
477
478
479
        ninit_steps = len(self.nsteps) - 2
        for j, n in enumerate(self.nsteps[:-2]):
            logging.info('Running {}/{} initialisation with {} steps'.format(
Gregory Ashton's avatar
Gregory Ashton committed
480
                j, ninit_steps, n))
481
            sampler = self._run_sampler(sampler, p0, nburn=n)
482
            if create_plots:
483
                fig, axes = self._plot_walkers(sampler,
484
485
                                               symbols=self.theta_symbols,
                                               **kwargs)
486
487
                fig.tight_layout()
                fig.savefig('{}/{}_init_{}_walkers.png'.format(
Gregory Ashton's avatar
Gregory Ashton committed
488
                    self.outdir, self.label, j))
489

490
491
492
            p0 = self._get_new_p0(sampler)
            p0 = self._apply_corrections_to_p0(p0)
            self._check_initial_points(p0)
493
494
            sampler.reset()

Gregory Ashton's avatar
Gregory Ashton committed
495
496
497
498
        if len(self.nsteps) > 1:
            nburn = self.nsteps[-2]
        else:
            nburn = 0
499
500
501
        nprod = self.nsteps[-1]
        logging.info('Running final burn and prod with {} steps'.format(
            nburn+nprod))
502
        sampler = self._run_sampler(sampler, p0, nburn=nburn, nprod=nprod)
503
        if create_plots:
504
            fig, axes = self._plot_walkers(sampler, symbols=self.theta_symbols,
505
                                           nprod=nprod, **kwargs)
506
507
            fig.tight_layout()
            fig.savefig('{}/{}_walkers.png'.format(self.outdir, self.label),
Gregory Ashton's avatar
Gregory Ashton committed
508
                        )
509
510
511
512

        samples = sampler.chain[0, :, nburn:, :].reshape((-1, self.ndim))
        lnprobs = sampler.lnprobability[0, :, nburn:].reshape((-1))
        lnlikes = sampler.lnlikelihood[0, :, nburn:].reshape((-1))
513
        all_lnlikelihood = sampler.lnlikelihood[:, :, nburn:]
514
515
516
        self.samples = samples
        self.lnprobs = lnprobs
        self.lnlikes = lnlikes
517
518
        self.all_lnlikelihood = all_lnlikelihood
        self._save_data(sampler, samples, lnprobs, lnlikes, all_lnlikelihood)
Gregory Ashton's avatar
Gregory Ashton committed
519
        return sampler
520

521
    def _get_rescale_multiplier_for_key(self, key):
522
        """ Get the rescale multiplier from the transform_dictionary
523
524
525
526
527

        Can either be a float, a string (in which case it is interpretted as
        a attribute of the MCMCSearch class, e.g. minStartTime, or non-existent
        in which case 0 is returned
        """
528
        if key not in self.transform_dictionary:
529
530
            return 1

531
532
        if 'multiplier' in self.transform_dictionary[key]:
            val = self.transform_dictionary[key]['multiplier']
533
534
535
            if type(val) == str:
                if hasattr(self, val):
                    multiplier = getattr(
536
                        self, self.transform_dictionary[key]['multiplier'])
537
538
539
540
541
542
543
544
545
                else:
                    raise ValueError(
                        "multiplier {} not a class attribute".format(val))
            else:
                multiplier = val
        else:
            multiplier = 1
        return multiplier

546
    def _get_rescale_subtractor_for_key(self, key):
547
        """ Get the rescale subtractor from the transform_dictionary
548
549
550
551
552

        Can either be a float, a string (in which case it is interpretted as
        a attribute of the MCMCSearch class, e.g. minStartTime, or non-existent
        in which case 0 is returned
        """
553
        if key not in self.transform_dictionary:
554
555
            return 0

556
557
        if 'subtractor' in self.transform_dictionary[key]:
            val = self.transform_dictionary[key]['subtractor']
558
559
560
            if type(val) == str:
                if hasattr(self, val):
                    subtractor = getattr(
561
                        self, self.transform_dictionary[key]['subtractor'])
562
563
564
565
566
567
568
569
570
                else:
                    raise ValueError(
                        "subtractor {} not a class attribute".format(val))
            else:
                subtractor = val
        else:
            subtractor = 0
        return subtractor

571
    def _scale_samples(self, samples, theta_keys):
572
        """ Scale the samples using the transform_dictionary """
573
        for key in theta_keys:
574
            if key in self.transform_dictionary:
575
576
                idx = theta_keys.index(key)
                s = samples[:, idx]
577
                subtractor = self._get_rescale_subtractor_for_key(key)
578
                s = s - subtractor
579
                multiplier = self._get_rescale_multiplier_for_key(key)
580
                s *= multiplier
581
582
                samples[:, idx] = s

583
584
        return samples

585
    def _get_labels(self):
586
        """ Combine the units, symbols and rescaling to give labels """
587

588
589
590
591
592
593
        labels = []
        for key in self.theta_keys:
            label = None
            s = self.symbol_dictionary[key]
            s.replace('_{glitch}', r'_\textrm{glitch}')
            u = self.unit_dictionary[key]
594
595
596
597
598
599
600
            if key in self.transform_dictionary:
                if 'symbol' in self.transform_dictionary[key]:
                    s = self.transform_dictionary[key]['symbol']
                if 'label' in self.transform_dictionary[key]:
                    label = self.transform_dictionary[key]['label']
                if 'unit' in self.transform_dictionary[key]:
                    u = self.transform_dictionary[key]['unit']
601
602
603
604
            if label is None:
                label = '{} \n [{}]'.format(s, u)
            labels.append(label)
        return labels
605

606
607
    def plot_corner(self, figsize=(7, 7), add_prior=False, nstds=None,
                    label_offset=0.4, dpi=300, rc_context={},
608
                    tglitch_ratio=False, fig_and_axes=None, save_fig=True,
609
                    **kwargs):
610
611
612
613
614
615
616
617
618
        """ Generate a corner plot of the posterior

        Using the `corner` package (https://pypi.python.org/pypi/corner/),
        generate estimates of the posterior from the production samples.

        Parameters
        ----------
        figsize: tuple (7, 7)
            Figure size in inches (passed to plt.subplots)
619
620
621
        add_prior: bool, str
            If true, plot the prior as a red line. If 'full' then for uniform
            priors plot the full extent of the prior.
622
623
624
625
626
627
628
629
630
631
632
633
634
635
        nstds: float
            The number of standard deviations to plot centered on the mean
        label_offset: float
            Offset the labels from the plot: useful to precent overlapping the
            tick labels with the axis labels
        dpi: int
            Passed to plt.savefig
        rc_context: dict
            Dictionary of rc values to set while generating the figure (see
            matplotlib rc for more details)
        tglitch_ratio: bool
            If true, and tglitch is a parameter, plot posteriors as the
            fractional time at which the glitch occurs instead of the actual
            time
636
637
638
639
640
        fig_and_axes: tuple
            fig and axes to plot on, the axes must be of the right shape,
            namely (ndim, ndim)
        save_fig: bool
            If true, save the figure, else return the fig, axes
641
642
        **kwargs:
            Passed to corner.corner
643

644
645
646
647
        Returns
        -------
        fig, axes:
            The matplotlib figure and axes, only returned if save_fig = False
648
649

        """
650

651
652
653
654
        if 'truths' in kwargs and len(kwargs['truths']) != self.ndim:
            logging.warning('len(Truths) != ndim, Truths will be ignored')
            kwargs['truths'] = None

Gregory Ashton's avatar
Gregory Ashton committed
655
656
        if self.ndim < 2:
            with plt.rc_context(rc_context):
657
658
659
660
                if fig_and_axes is None:
                    fig, ax = plt.subplots(figsize=figsize)
                else:
                    fig, ax = fig_and_axes
Gregory Ashton's avatar
Gregory Ashton committed
661
662
663
664
665
666
667
                ax.hist(self.samples, bins=50, histtype='stepfilled')
                ax.set_xlabel(self.theta_symbols[0])

            fig.savefig('{}/{}_corner.png'.format(
                self.outdir, self.label), dpi=dpi)
            return

668
        with plt.rc_context(rc_context):
669
670
671
672
673
            if fig_and_axes is None:
                fig, axes = plt.subplots(self.ndim, self.ndim,
                                         figsize=figsize)
            else:
                fig, axes = fig_and_axes
674
675

            samples_plt = copy.copy(self.samples)
676
            labels = self._get_labels()
677

678
            samples_plt = self._scale_samples(samples_plt, self.theta_keys)
679
680
681
682
683

            if tglitch_ratio:
                for j, k in enumerate(self.theta_keys):
                    if k == 'tglitch':
                        s = samples_plt[:, j]
684
685
686
                        samples_plt[:, j] = (
                            s - self.minStartTime)/(
                                self.maxStartTime - self.minStartTime)
687
                        labels[j] = r'$R_{\textrm{glitch}}$'
688
689
690
691
692
693
694

            if type(nstds) is int and 'range' not in kwargs:
                _range = []
                for j, s in enumerate(samples_plt.T):
                    median = np.median(s)
                    std = np.std(s)
                    _range.append((median - nstds*std, median + nstds*std))
695
696
            elif 'range' in kwargs:
                _range = kwargs.pop('range')
697
698
699
            else:
                _range = None

700
701
702
703
            hist_kwargs = kwargs.pop('hist_kwargs', dict())
            if 'normed' not in hist_kwargs:
                hist_kwargs['normed'] = True

704
            fig_triangle = corner.corner(samples_plt,
705
                                         labels=labels,
706
707
708
709
710
711
712
713
714
                                         fig=fig,
                                         bins=50,
                                         max_n_ticks=4,
                                         plot_contours=True,
                                         plot_datapoints=True,
                                         label_kwargs={'fontsize': 8},
                                         data_kwargs={'alpha': 0.1,
                                                      'ms': 0.5},
                                         range=_range,
715
                                         hist_kwargs=hist_kwargs,
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
                                         **kwargs)

            axes_list = fig_triangle.get_axes()
            axes = np.array(axes_list).reshape(self.ndim, self.ndim)
            plt.draw()
            for ax in axes[:, 0]:
                ax.yaxis.set_label_coords(-label_offset, 0.5)
            for ax in axes[-1, :]:
                ax.xaxis.set_label_coords(0.5, -label_offset)
            for ax in axes_list:
                ax.set_rasterized(True)
                ax.set_rasterization_zorder(-10)
            plt.tight_layout(h_pad=0.0, w_pad=0.0)
            fig.subplots_adjust(hspace=0.05, wspace=0.05)

            if add_prior:
732
                self._add_prior_to_corner(axes, self.samples, add_prior)
733

734
735
736
737
738
            if save_fig:
                fig_triangle.savefig('{}/{}_corner.png'.format(
                    self.outdir, self.label), dpi=dpi)
            else:
                return fig, axes
739

740
    def _add_prior_to_corner(self, axes, samples, add_prior):
741
742
743
        for i, key in enumerate(self.theta_keys):
            ax = axes[i][i]
            s = samples[:, i]
744
745
746
747
748
749
750
751
752
753
            lnprior = self._generic_lnprior(**self.theta_prior[key])
            if add_prior == 'full' and self.theta_prior[key]['type'] == 'unif':
                lower = self.theta_prior[key]['lower']
                upper = self.theta_prior[key]['upper']
                r = upper-lower
                xlim = [lower-0.05*r, upper+0.05*r]
                x = np.linspace(xlim[0], xlim[1], 1000)
            else:
                xlim = ax.get_xlim()
                x = np.linspace(s.min(), s.max(), 1000)
754
755
            multiplier = self._get_rescale_multiplier_for_key(key)
            subtractor = self._get_rescale_subtractor_for_key(key)
756
757
758
759
760
761
762
763
            ax.plot((x-subtractor)*multiplier,
                    [np.exp(lnprior(xi)) for xi in x], '-C3',
                    label='prior')

            for j in range(i, self.ndim):
                axes[j][i].set_xlim(xlim[0], xlim[1])
            for k in range(0, i):
                axes[i][k].set_ylim(xlim[0], xlim[1])
764

765
766
767
768
769
770
771
772
    def plot_prior_posterior(self, normal_stds=2):
        """ Plot the posterior in the context of the prior """
        fig, axes = plt.subplots(nrows=self.ndim, figsize=(8, 4*self.ndim))
        N = 1000
        from scipy.stats import gaussian_kde

        for i, (ax, key) in enumerate(zip(axes, self.theta_keys)):
            prior_dict = self.theta_prior[key]
773
            prior_func = self._generic_lnprior(**prior_dict)
774
775
776
777
778
            if prior_dict['type'] == 'unif':
                x = np.linspace(prior_dict['lower'], prior_dict['upper'], N)
                prior = prior_func(x)
                prior[0] = 0
                prior[-1] = 0
Gregory Ashton's avatar
Gregory Ashton committed
779
780
781
782
783
            elif prior_dict['type'] == 'log10unif':
                upper = prior_dict['log10upper']
                lower = prior_dict['log10lower']
                x = np.linspace(lower, upper, N)
                prior = [prior_func(xi) for xi in x]
784
785
786
787
788
            elif prior_dict['type'] == 'norm':
                lower = prior_dict['loc'] - normal_stds * prior_dict['scale']
                upper = prior_dict['loc'] + normal_stds * prior_dict['scale']
                x = np.linspace(lower, upper, N)
                prior = prior_func(x)
789
790
791
792
793
            elif prior_dict['type'] == 'halfnorm':
                lower = prior_dict['loc']
                upper = prior_dict['loc'] + normal_stds * prior_dict['scale']
                x = np.linspace(lower, upper, N)
                prior = [prior_func(xi) for xi in x]
Gregory Ashton's avatar
Gregory Ashton committed
794
795
796
797
798
            elif prior_dict['type'] == 'neghalfnorm':
                upper = prior_dict['loc']
                lower = prior_dict['loc'] - normal_stds * prior_dict['scale']
                x = np.linspace(lower, upper, N)
                prior = [prior_func(xi) for xi in x]
799
800
801
            else:
                raise ValueError('Not implemented for prior type {}'.format(
                    prior_dict['type']))
802
            priorln = ax.plot(x, prior, 'C3', label='prior')
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
            ax.set_xlabel(self.theta_symbols[i])

            s = self.samples[:, i]
            while len(s) > 10**4:
                # random downsample to avoid slow calculation of kde
                s = np.random.choice(s, size=int(len(s)/2.))
            kde = gaussian_kde(s)
            ax2 = ax.twinx()
            postln = ax2.plot(x, kde.pdf(x), 'k', label='posterior')
            ax2.set_yticklabels([])
            ax.set_yticklabels([])

        lns = priorln + postln
        labs = [l.get_label() for l in lns]
        axes[0].legend(lns, labs, loc=1, framealpha=0.8)

        fig.savefig('{}/{}_prior_posterior.png'.format(
            self.outdir, self.label))

822
    def plot_cumulative_max(self, **kwargs):
823
824
825
826
        """ Plot the cumulative twoF for the maximum posterior estimate

        See the pyfstat.core.plot_twoF_cumulative function for further details
        """
Gregory Ashton's avatar
Gregory Ashton committed
827
828
829
830
        d, maxtwoF = self.get_max_twoF()
        for key, val in self.theta_prior.iteritems():
            if key not in d:
                d[key] = val
831
832

        if hasattr(self, 'search') is False:
833
            self._initiate_search_object()
834
835
836
        if self.binary is False:
            self.search.plot_twoF_cumulative(
                self.label, self.outdir, F0=d['F0'], F1=d['F1'], F2=d['F2'],
837
                Alpha=d['Alpha'], Delta=d['Delta'],
838
                tstart=self.minStartTime, tend=self.maxStartTime,
839
                **kwargs)
840
841
842
843
844
        else:
            self.search.plot_twoF_cumulative(
                self.label, self.outdir, F0=d['F0'], F1=d['F1'], F2=d['F2'],
                Alpha=d['Alpha'], Delta=d['Delta'], asini=d['asini'],
                period=d['period'], ecc=d['ecc'], argp=d['argp'], tp=d['argp'],
845
                tstart=self.minStartTime, tend=self.maxStartTime, **kwargs)
Gregory Ashton's avatar
Gregory Ashton committed
846

847
    def _generic_lnprior(self, **kwargs):
848
849
850
851
        """ Return a lambda function of the pdf

        Parameters
        ----------
852
        **kwargs:
853
854
855
856
            A dictionary containing 'type' of pdf and shape parameters

        """

Gregory Ashton's avatar
Gregory Ashton committed
857
        def log_of_unif(x, a, b):
858
859
860
861
862
863
864
865
866
867
868
869
870
            above = x < b
            below = x > a
            if type(above) is not np.ndarray:
                if above and below:
                    return -np.log(b-a)
                else:
                    return -np.inf
            else:
                idxs = np.array([all(tup) for tup in zip(above, below)])
                p = np.zeros(len(x)) - np.inf
                p[idxs] = -np.log(b-a)
                return p

Gregory Ashton's avatar
Gregory Ashton committed
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
        def log_of_log10unif(x, log10lower, log10upper):
            log10x = np.log10(x)
            above = log10x < log10upper
            below = log10x > log10lower
            if type(above) is not np.ndarray:
                if above and below:
                    return -np.log(x*np.log(10)*(log10upper-log10lower))
                else:
                    return -np.inf
            else:
                idxs = np.array([all(tup) for tup in zip(above, below)])
                p = np.zeros(len(x)) - np.inf
                p[idxs] = -np.log(x*np.log(10)*(log10upper-log10lower))
                return p

        def log_of_halfnorm(x, loc, scale):
887
            if x < loc:
888
889
890
891
892
893
894
895
896
897
898
899
900
901
                return -np.inf
            else:
                return -0.5*((x-loc)**2/scale**2+np.log(0.5*np.pi*scale**2))

        def cauchy(x, x0, gamma):
            return 1.0/(np.pi*gamma*(1+((x-x0)/gamma)**2))

        def exp(x, x0, gamma):
            if x > x0:
                return np.log(gamma) - gamma*(x - x0)
            else:
                return -np.inf

        if kwargs['type'] == 'unif':
Gregory Ashton's avatar
Gregory Ashton committed
902
903
904
905
            return lambda x: log_of_unif(x, kwargs['lower'], kwargs['upper'])
        if kwargs['type'] == 'log10unif':
            return lambda x: log_of_log10unif(
                x, kwargs['log10lower'], kwargs['log10upper'])
906
        elif kwargs['type'] == 'halfnorm':
Gregory Ashton's avatar
Gregory Ashton committed
907
            return lambda x: log_of_halfnorm(x, kwargs['loc'], kwargs['scale'])
908
        elif kwargs['type'] == 'neghalfnorm':
Gregory Ashton's avatar
Gregory Ashton committed
909
910
            return lambda x: log_of_halfnorm(
                -x, kwargs['loc'], kwargs['scale'])
911
912
913
914
915
916
917
        elif kwargs['type'] == 'norm':
            return lambda x: -0.5*((x - kwargs['loc'])**2/kwargs['scale']**2
                                   + np.log(2*np.pi*kwargs['scale']**2))
        else:
            logging.info("kwargs:", kwargs)
            raise ValueError("Print unrecognise distribution")

918
    def _generate_rv(self, **kwargs):
919
920
921
        dist_type = kwargs.pop('type')
        if dist_type == "unif":
            return np.random.uniform(low=kwargs['lower'], high=kwargs['upper'])
Gregory Ashton's avatar
Gregory Ashton committed
922
923
924
        if dist_type == "log10unif":
            return 10**(np.random.uniform(low=kwargs['log10lower'],
                                          high=kwargs['log10upper']))
925
926
927
928
929
        if dist_type == "norm":
            return np.random.normal(loc=kwargs['loc'], scale=kwargs['scale'])
        if dist_type == "halfnorm":
            return np.abs(np.random.normal(loc=kwargs['loc'],
                                           scale=kwargs['scale']))
930
931
932
        if dist_type == "neghalfnorm":
            return -1 * np.abs(np.random.normal(loc=kwargs['loc'],
                                                scale=kwargs['scale']))
933
934
935
936
937
938
        if dist_type == "lognorm":
            return np.random.lognormal(
                mean=kwargs['loc'], sigma=kwargs['scale'])
        else:
            raise ValueError("dist_type {} unknown".format(dist_type))

939
    def _plot_walkers(self, sampler, symbols=None, alpha=0.8, color="k",
940
941
                      temp=0, lw=0.1, nprod=0, add_det_stat_burnin=False,
                      fig=None, axes=None, xoffset=0, plot_det_stat=False,
942
                      context='ggplot', subtractions=None, labelpad=0.05):
943
944
        """ Plot all the chains from a sampler """

945
946
947
948
949
        if context not in plt.style.available:
            raise ValueError((
                'The requested context {} is not available; please select a'
                ' context from `plt.style.available`').format(context))

950
951
952
        if np.ndim(axes) > 1:
            axes = axes.flatten()

953
954
955
956
957
958
959
960
961
962
963
964
965
        shape = sampler.chain.shape
        if len(shape) == 3:
            nwalkers, nsteps, ndim = shape
            chain = sampler.chain[:, :, :]
        if len(shape) == 4:
            ntemps, nwalkers, nsteps, ndim = shape
            if temp < ntemps:
                logging.info("Plotting temperature {} chains".format(temp))
            else:
                raise ValueError(("Requested temperature {} outside of"
                                  "available range").format(temp))
            chain = sampler.chain[temp, :, :, :]

966
967
        if subtractions is None:
            subtractions = [0 for i in range(ndim)]
968
969
970
        else:
            if len(subtractions) != self.ndim:
                raise ValueError('subtractions must be of length ndim')
971

972
973
974
975
        if plot_det_stat:
            extra_subplots = 1
        else:
            extra_subplots = 0
976
        with plt.style.context((context)):
Gregory Ashton's avatar
Gregory Ashton committed
977
            plt.rcParams['text.usetex'] = True
Gregory Ashton's avatar
Gregory Ashton committed
978
            if fig is None and axes is None:
979
                fig = plt.figure(figsize=(4, 3.0*ndim))
980
981
                ax = fig.add_subplot(ndim+extra_subplots, 1, 1)
                axes = [ax] + [fig.add_subplot(ndim+extra_subplots, 1, i)
Gregory Ashton's avatar
Gregory Ashton committed
982
                               for i in range(2, ndim+1)]
983

Gregory Ashton's avatar
Gregory Ashton committed
984
            idxs = np.arange(chain.shape[1])
985
986
987
988
989
            burnin_idx = chain.shape[1] - nprod
            if hasattr(self, 'convergence_idx'):
                convergence_idx = self.convergence_idx
            else:
                convergence_idx = burnin_idx
990
991
            if ndim > 1:
                for i in range(ndim):
992
                    axes[i].ticklabel_format(useOffset=False, axis='y')
Gregory Ashton's avatar
Gregory Ashton committed
993
                    cs = chain[:, :, i].T
994
                    if burnin_idx > 0:
995
996
                        axes[i].plot(xoffset+idxs[:convergence_idx+1],
                                     cs[:convergence_idx+1]-subtractions[i],
997
                                     color="C3", alpha=alpha,
Gregory Ashton's avatar
Gregory Ashton committed
998
                                     lw=lw)
999
                        axes[i].axvline(xoffset+convergence_idx,
1000
                                        color='k', ls='--', lw=0.25)
1001
1002
                    axes[i].plot(xoffset+idxs[burnin_idx:],
                                 cs[burnin_idx:]-subtractions[i],
Gregory Ashton's avatar
Gregory Ashton committed
1003
                                 color="k", alpha=alpha, lw=lw)
Gregory Ashton's avatar
Gregory Ashton committed
1004
1005

                    axes[i].set_xlim(0, xoffset+idxs[-1])
1006
                    if symbols:
1007
                        if subtractions[i] == 0:
1008
                            axes[i].set_ylabel(symbols[i], labelpad=labelpad)
1009
1010
                        else:
                            axes[i].set_ylabel(
1011
1012
                                symbols[i]+'$-$'+symbols[i]+'$_0$',
                                labelpad=labelpad)
1013

1014
1015
                    if hasattr(self, 'convergence_diagnostic'):
                        ax = axes[i].twinx()
1016
1017
                        axes[i].set_zorder(ax.get_zorder()+1)
                        axes[i].patch.set_visible(False)
1018
1019
                        c_x = np.array(self.convergence_diagnosticx)
                        c_y = np.array(self.convergence_diagnostic)
1020
                        break_idx = np.argmin(np.abs(c_x - burnin_idx))
1021
1022
1023
1024
                        ax.plot(c_x[:break_idx], c_y[:break_idx, i], '-C0',
                                zorder=-10)
                        ax.plot(c_x[break_idx:], c_y[break_idx:, i], '-C0',
                                zorder=-10)
1025
1026
1027
1028
                        if self.convergence_test_type == 'autocorr':
                            ax.set_ylabel(r'$\tau_\mathrm{exp}$')
                        elif self.convergence_test_type == 'GR':
                            ax.set_ylabel('PSRF')
1029
                        ax.ticklabel_format(useOffset=False)
1030
            else:
Gregory Ashton's avatar
Gregory Ashton committed
1031
                axes[0].ticklabel_format(useOffset=False, axis='y')
Gregory Ashton's avatar
Gregory Ashton committed
1032
                cs = chain[:, :, temp].T
Gregory Ashton's avatar
Gregory Ashton committed
1033
1034
                if burnin_idx:
                    axes[0].plot(idxs[:burnin_idx], cs[:burnin_idx],
1035
                                 color="C3", alpha=alpha, lw=lw)
Gregory Ashton's avatar
Gregory Ashton committed
1036
1037
1038
                axes[0].plot(idxs[burnin_idx:], cs[burnin_idx:], color="k",
                             alpha=alpha, lw=lw)
                if symbols:
1039
                    axes[0].set_ylabel(symbols[0], labelpad=labelpad)
1040

Gregory Ashton's avatar
Gregory Ashton committed
1041
1042
            axes[-1].set_xlabel(r'$\textrm{Number of steps}$', labelpad=0.2)

1043
            if plot_det_stat:
1044
1045
1046
                if len(axes) == ndim:
                    axes.append(fig.add_subplot(ndim+1, 1, ndim+1))

1047
1048
1049
                lnl = sampler.lnlikelihood[temp, :, :]
                if burnin_idx and add_det_stat_burnin:
                    burn_in_vals = lnl[:, :burnin_idx].flatten()
1050
                    try:
1051
1052
1053
1054
                        twoF_burnin = (burn_in_vals[~np.isnan(burn_in_vals)]
                                       - self.likelihoodcoef)
                        axes[-1].hist(twoF_burnin, bins=50, histtype='step',
                                      color='C3')
1055
1056
1057
1058
                    except ValueError:
                        logging.info('Det. Stat. hist failed, most likely all '
                                     'values where the same')
                        pass
1059
                else:
1060
                    twoF_burnin = []
1061
                prod_vals = lnl[:, burnin_idx:].flatten()
1062
                try:
1063
1064
                    twoF = prod_vals[~np.isnan(prod_vals)]-self.likelihoodcoef
                    axes[-1].hist(twoF, bins=50, histtype='step', color='k')