This file is indexed.

/usr/lib/python3/dist-packages/emcee/tests.py is in python3-emcee 2.2.1-1.

This file is owned by root:root, with mode 0o644.

The actual contents of the file can be viewed below.

  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
#!/usr/bin/env python
# encoding: utf-8
"""
Defines various nose unit tests

"""

import numpy as np

from .mh import MHSampler
from .ensemble import EnsembleSampler
from .ptsampler import PTSampler

logprecision = -4


def lnprob_gaussian(x, icov):
    return -np.dot(x, np.dot(icov, x)) / 2.0


def lnprob_gaussian_nan(x, icov):
    # if walker's parameters are zeros => return NaN
    if not (np.array(x)).any():
        result = np.nan
    else:
        result = -np.dot(x, np.dot(icov, x)) / 2.0

    return result


def log_unit_sphere_volume(ndim):
    if ndim % 2 == 0:
        logfactorial = 0.0
        for i in range(1, ndim / 2 + 1):
            logfactorial += np.log(i)
        return ndim / 2.0 * np.log(np.pi) - logfactorial
    else:
        logfactorial = 0.0
        for i in range(1, ndim + 1, 2):
            logfactorial += np.log(i)
        return (ndim + 1) / 2.0 * np.log(2.0) \
            + (ndim - 1) / 2.0 * np.log(np.pi) - logfactorial


class LogLikeGaussian(object):
    def __init__(self, icov):
        """Initialize a gaussian PDF with the given inverse covariance
        matrix.  If not ``None``, ``cutoff`` truncates the PDF at the
        given number of sigma from the origin (i.e. the PDF is
        non-zero only on an ellipse aligned with the principal axes of
        the distribution).  Without this cutoff, thermodynamic
        integration with a flat prior is logarithmically divergent."""

        self.icov = icov

    def __call__(self, x):
        dist2 = lnprob_gaussian(x, self.icov)

        return dist2


class LogPriorGaussian(object):
    def __init__(self, icov, cutoff=None):
        self.icov = icov
        self.cutoff = cutoff

    def __call__(self, x):
        dist2 = lnprob_gaussian(x, self.icov)

        if self.cutoff is not None:
            if -dist2 > self.cutoff * self.cutoff / 2.0:
                return float('-inf')
            else:
                return 0.0
        else:
            return 0.0


def ln_flat(x):
    return 0.0


class Tests:

    def setUp(self):
        np.random.seed(42)

        self.nwalkers = 100
        self.ndim = 5

        self.ntemp = 20

        self.N = 1000

        self.mean = np.zeros(self.ndim)
        self.cov = 0.5 - np.random.rand(self.ndim ** 2) \
            .reshape((self.ndim, self.ndim))
        self.cov = np.triu(self.cov)
        self.cov += self.cov.T - np.diag(self.cov.diagonal())
        self.cov = np.dot(self.cov, self.cov)
        self.icov = np.linalg.inv(self.cov)
        self.p0 = [0.1 * np.random.randn(self.ndim)
                   for i in range(self.nwalkers)]
        self.truth = np.random.multivariate_normal(self.mean, self.cov, 100000)

    def check_sampler(self, N=None, p0=None):
        if N is None:
            N = self.N
        if p0 is None:
            p0 = self.p0

        for i in self.sampler.sample(p0, iterations=N):
            pass

        assert np.mean(self.sampler.acceptance_fraction) > 0.25
        assert np.all(self.sampler.acceptance_fraction > 0)

        chain = self.sampler.flatchain
        maxdiff = 10. ** (logprecision)
        assert np.all((np.mean(chain, axis=0) - self.mean) ** 2 / self.N ** 2
                      < maxdiff)
        assert np.all((np.cov(chain, rowvar=0) - self.cov) ** 2 / self.N ** 2
                      < maxdiff)

    def check_pt_sampler(self, cutoff, N=None, p0=None):
        if N is None:
            N = self.N
        if p0 is None:
            p0 = self.p0

        for i in self.sampler.sample(p0, iterations=N):
            pass

        # Weaker assertions on acceptance fraction
        assert np.mean(self.sampler.acceptance_fraction) > 0.1, \
            "acceptance fraction < 0.1"
        assert np.mean(self.sampler.tswap_acceptance_fraction) > 0.1, \
            "tswap acceptance frac < 0.1"

        maxdiff = 10.0 ** logprecision

        chain = np.reshape(self.sampler.chain[0, ...],
                           (-1, self.sampler.chain.shape[-1]))

        # np.savetxt('/tmp/chain.dat', chain)

        log_volume = self.ndim * np.log(cutoff) \
            + log_unit_sphere_volume(self.ndim) \
            + 0.5 * np.log(np.linalg.det(self.cov))
        gaussian_integral = self.ndim / 2.0 * np.log(2.0 * np.pi) \
            + 0.5 * np.log(np.linalg.det(self.cov))

        lnZ, dlnZ = self.sampler.thermodynamic_integration_log_evidence()
        print(self.sampler.get_autocorr_time(c=2))

        assert np.abs(lnZ - (gaussian_integral - log_volume)) < 3 * dlnZ, \
            ("evidence incorrect: {0:g} versus correct {1:g} (uncertainty "
             "{2:g})").format(lnZ, gaussian_integral - log_volume, dlnZ)
        assert np.all((np.mean(chain, axis=0) - self.mean) ** 2.0 / N ** 2.0
                      < maxdiff), 'mean incorrect'
        assert np.all((np.cov(chain, rowvar=0) - self.cov) ** 2.0 / N ** 2.0
                      < maxdiff), 'covariance incorrect'

    def test_mh(self):
        self.sampler = MHSampler(self.cov, self.ndim, lnprob_gaussian,
                                 args=[self.icov])
        self.check_sampler(N=self.N * self.nwalkers, p0=self.p0[0])

    def test_ensemble(self):
        self.sampler = EnsembleSampler(self.nwalkers, self.ndim,
                                       lnprob_gaussian, args=[self.icov])
        self.check_sampler()

    def test_nan_lnprob(self):
        self.sampler = EnsembleSampler(self.nwalkers, self.ndim,
                                       lnprob_gaussian_nan,
                                       args=[self.icov])

        # If a walker is right at zero, ``lnprobfn`` returns ``np.nan``.
        p0 = self.p0
        p0[0] = 0.0

        try:
            self.check_sampler(p0=p0)
        except ValueError:
            # This should fail *immediately* with a ``ValueError``.
            return

        assert False, "We should never get here."

    def test_inf_nan_params(self):
        self.sampler = EnsembleSampler(self.nwalkers, self.ndim,
                                       lnprob_gaussian, args=[self.icov])

        # Set one of the walkers to have a ``np.nan`` value.
        p0 = self.p0
        p0[0][0] = np.nan

        try:
            self.check_sampler(p0=p0)

        except ValueError:
            # This should fail *immediately* with a ``ValueError``.
            pass

        else:
            assert False, "The sampler should have failed by now."

        # Set one of the walkers to have a ``np.inf`` value.
        p0[0][0] = np.inf

        try:
            self.check_sampler(p0=p0)

        except ValueError:
            # This should fail *immediately* with a ``ValueError``.
            pass

        else:
            assert False, "The sampler should have failed by now."

        # Set one of the walkers to have a ``np.inf`` value.
        p0[0][0] = -np.inf

        try:
            self.check_sampler(p0=p0)

        except ValueError:
            # This should fail *immediately* with a ``ValueError``.
            pass

        else:
            assert False, "The sampler should have failed by now."

    def test_parallel(self):
        self.sampler = EnsembleSampler(self.nwalkers, self.ndim,
                                       lnprob_gaussian, args=[self.icov],
                                       threads=2)
        self.check_sampler()

    def test_pt_sampler(self):
        cutoff = 10.0
        self.sampler = PTSampler(self.ntemp, self.nwalkers, self.ndim,
                                 LogLikeGaussian(self.icov),
                                 LogPriorGaussian(self.icov, cutoff=cutoff))
        p0 = np.random.multivariate_normal(mean=self.mean, cov=self.cov,
                                           size=(self.ntemp, self.nwalkers))
        self.check_pt_sampler(cutoff, p0=p0, N=1000)

    def test_blobs(self):
        lnprobfn = lambda p: (-0.5 * np.sum(p ** 2), np.random.rand())
        self.sampler = EnsembleSampler(self.nwalkers, self.ndim, lnprobfn)
        self.check_sampler()

        # Make sure that the shapes of everything are as expected.
        assert (self.sampler.chain.shape == (self.nwalkers, self.N, self.ndim)
                and len(self.sampler.blobs) == self.N
                and len(self.sampler.blobs[0]) == self.nwalkers), \
            "The blob dimensions are wrong."

        # Make sure that the blobs aren't all the same.
        blobs = self.sampler.blobs
        assert np.any([blobs[-1] != blobs[i] for i in range(len(blobs) - 1)])

    def test_run_mcmc_resume(self):

        self.sampler = s = EnsembleSampler(self.nwalkers, self.ndim,
                                           lnprob_gaussian, args=[self.icov])

        # first time around need to specify p0
        try:
            s.run_mcmc(None, self.N)
        except ValueError:
            pass

        s.run_mcmc(self.p0, N=self.N)
        assert s.chain.shape[1] == self.N

        # this doesn't actually check that it resumes with the right values, as
        # that's non-trivial... so we just make sure it does *something* when
        # None is given and that it records whatever it does
        s.run_mcmc(None, N=self.N)
        assert s.chain.shape[1] == 2 * self.N