sugar.py 11.6 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation, either version 3 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program.  If not, see <http://www.gnu.org/licenses/>.
Theo Steininger's avatar
Theo Steininger committed
13
14
15
16
17
#
# Copyright(C) 2013-2017 Max-Planck-Society
#
# NIFTy is being developed at the Max-Planck-Institut fuer Astrophysik
# and financially supported by the Studienstiftung des deutschen Volkes.
18

19
import numpy as np
Martin Reinecke's avatar
Martin Reinecke committed
20
from . import Space,\
Martin Reinecke's avatar
Martin Reinecke committed
21
22
23
24
25
26
27
28
              PowerSpace,\
              Field,\
              ComposedOperator,\
              DiagonalOperator,\
              PowerProjectionOperator,\
              FFTOperator,\
              sqrt,\
              DomainTuple
29
from . import nifty_utilities as utilities
Martin Reinecke's avatar
Martin Reinecke committed
30
from . import dobj
31

32
33
34
35
__all__ = ['power_analyze',
           'power_synthesize',
           'power_synthesize_special',
           'create_power_field',
36
           'create_power_operator',
37
38
           'generate_posterior_sample',
           'create_composed_fft_operator']
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
def _single_power_analyze(field, idx, binbounds):
    from .operators.power_projection_operator import PowerProjectionOperator
    power_domain = PowerSpace(field.domain[idx], binbounds)
    ppo = PowerProjectionOperator(field.domain, power_domain, idx)
    return ppo(field.weight(-1))


def power_analyze(field, spaces=None, binbounds=None,
                  keep_phase_information=False):
    """ Computes the square root power spectrum for a subspace of `field`.

    Creates a PowerSpace for the space addressed by `spaces` with the given
    binning and computes the power spectrum as a Field over this
    PowerSpace. This can only be done if the subspace to  be analyzed is a
    harmonic space. The resulting field has the same units as the initial
    field, corresponding to the square root of the power spectrum.

    Parameters
    ----------
    field : Field
        The field to be analyzed
    spaces : int *optional*
        The subspace for which the powerspectrum shall be computed.
        (default : None).
    binbounds : array-like *optional*
        Inner bounds of the bins (default : None).
        if binbounds==None : bins are inferred.
    keep_phase_information : boolean, *optional*
        If False, return a real-valued result containing the power spectrum
        of the input Field.
        If True, return a complex-valued result whose real component
        contains the power spectrum computed from the real part of the
        input Field, and whose imaginary component contains the power
        spectrum computed from the imaginary part of the input Field.
        The absolute value of this result should be identical to the output
        of power_analyze with keep_phase_information=False.
        (default : False).

    Raise
    -----
    TypeError
        Raised if any of the input field's domains is not harmonic

    Returns
    -------
    out : Field
        The output object. Its domain is a PowerSpace and it contains
        the power spectrum of 'field'.
    """

    # check if all spaces in `field.domain` are either harmonic or
    # power_space instances
    for sp in field.domain:
        if not sp.harmonic and not isinstance(sp, PowerSpace):
            print("WARNING: Field has a space in `domain` which is "
                  "neither harmonic nor a PowerSpace.")

    # check if the `spaces` input is valid
    if spaces is None:
        spaces = range(len(field.domain))
    else:
        spaces = utilities.cast_iseq_to_tuple(spaces)

    if len(spaces) == 0:
        raise ValueError("No space for analysis specified.")

    if keep_phase_information:
        parts = [field.real*field.real, field.imag*field.imag]
    else:
        parts = [field.real*field.real + field.imag*field.imag]

    parts = [part.weight(1, spaces) for part in parts]
    for space_index in spaces:
        parts = [_single_power_analyze(field=part,
                                       idx=space_index,
                                       binbounds=binbounds)
                 for part in parts]

    return parts[0] + 1j*parts[1] if keep_phase_information else parts[0]


def _compute_spec(field, spaces):
    from .operators.power_projection_operator import PowerProjectionOperator
    from .basic_arithmetics import sqrt
    if spaces is None:
        spaces = range(len(field.domain))
    else:
        spaces = utilities.cast_iseq_to_tuple(spaces)

    # create the result domain
    result_domain = list(field.domain)

    spec = sqrt(field)
    for i in spaces:
        result_domain[i] = field.domain[i].harmonic_partner
        ppo = PowerProjectionOperator(result_domain, field.domain[i], i)
        spec = ppo.adjoint_times(spec)

    return spec


def power_synthesize(field, spaces=None, real_power=True, real_signal=True):
    """ Yields a sampled field with `field`**2 as its power spectrum.

    This method draws a Gaussian random field in the harmonic partner
    domain of this field's domains, using this field as power spectrum.

    Parameters
    ----------
    field : Field
        The input field containing the square root of the power spectrum
    spaces : {tuple, int, None} *optional*
        Specifies the subspace containing all the PowerSpaces which
        should be converted (default : None).
        if spaces==None : Tries to convert the whole domain.
    real_power : boolean *optional*
        Determines whether the power spectrum is treated as intrinsically
        real or complex (default : True).
    real_signal : boolean *optional*
        True will result in a purely real signal-space field
        (default : True).

    Returns
    -------
    out : Field
        The output object. A random field created with the power spectrum
        stored in the `spaces` in `field`.

    Notes
    -----
    For this the spaces specified by `spaces` must be a PowerSpace.
    This expects this field to be the square root of a power spectrum, i.e.
    to have the unit of the field to be sampled.

    Raises
    ------
    ValueError : If domain specified by `spaces` is not a PowerSpace.

    """

    spec = _compute_spec(field, spaces)

    # create random samples: one or two, depending on whether the
    # power spectrum is real or complex
    result = [field.from_random('normal', mean=0., std=1.,
                                domain=spec.domain,
                                dtype=np.float if real_signal
                                else np.complex)
              for x in range(1 if real_power else 2)]

    # MR: dummy call - will be removed soon
    if real_signal:
        field.from_random('normal', mean=0., std=1.,
                          domain=spec.domain, dtype=np.float)

    # apply the rescaler to the random fields
    result[0] *= spec.real
    if not real_power:
        result[1] *= spec.imag

    return result[0] if real_power else result[0] + 1j*result[1]


def power_synthesize_special(field, spaces=None):
    spec = _compute_spec(field, spaces)

    # MR: dummy call - will be removed soon
    field.from_random('normal', mean=0., std=1.,
                      domain=spec.domain, dtype=np.complex)

    return spec.real


214
def create_power_field(domain, power_spectrum, dtype=None):
Martin Reinecke's avatar
tweaks    
Martin Reinecke committed
215
216
217
218
219
220
221
222
223
224
225
    if not callable(power_spectrum):  # we have a Field living on a PowerSpace
        if not isinstance(power_spectrum, Field):
            raise TypeError("Field object expected")
        if len(power_spectrum.domain) != 1:
            raise ValueError("exactly one domain required")
        if not isinstance(power_spectrum.domain[0], PowerSpace):
            raise TypeError("PowerSpace required")
        power_domain = power_spectrum.domain[0]
        fp = Field(power_domain, val=power_spectrum.val, dtype=dtype)
    else:
        power_domain = PowerSpace(domain)
Martin Reinecke's avatar
Martin Reinecke committed
226
        fp = Field(power_domain,
Martin Reinecke's avatar
Martin Reinecke committed
227
                   val=dobj.from_global_data(power_spectrum(power_domain.k_lengths)),
Martin Reinecke's avatar
tweaks    
Martin Reinecke committed
228
                   dtype=dtype)
Martin Reinecke's avatar
Martin Reinecke committed
229
230
    P = PowerProjectionOperator(domain, power_domain)
    f = P.adjoint_times(fp)
231
232
233
234
235
236

    if not issubclass(fp.dtype.type, np.complexfloating):
        f = f.real

    return f

237

Martin Reinecke's avatar
Martin Reinecke committed
238
def create_power_operator(domain, power_spectrum, space=None, dtype=None):
Theo Steininger's avatar
Theo Steininger committed
239
    """ Creates a diagonal operator with the given power spectrum.
240

241
    Constructs a diagonal operator that lives over the specified domain.
242

243
244
245
    Parameters
    ----------
    domain : DomainObject
246
        Domain over which the power operator shall live.
Martin Reinecke's avatar
Martin Reinecke committed
247
248
249
250
    power_spectrum : callable of Field
        An object that implements the power spectrum as a function of k.
    space : int
            the domain index on which the power operator will work
Theo Steininger's avatar
Theo Steininger committed
251
    dtype : type *optional*
252
        dtype that the field holding the power spectrum shall use
Theo Steininger's avatar
Theo Steininger committed
253
254
255
        (default : None).
        if dtype == None: the dtype of `power_spectrum` will be used.

256
257
    Returns
    -------
Theo Steininger's avatar
Theo Steininger committed
258
    DiagonalOperator : An operator that implements the given power spectrum.
259

260
    """
Martin Reinecke's avatar
Martin Reinecke committed
261
262
    domain = DomainTuple.make(domain)
    if space is None:
Martin Reinecke's avatar
Martin Reinecke committed
263
        if len(domain) != 1:
Martin Reinecke's avatar
Martin Reinecke committed
264
265
266
267
            raise ValueError("space keyword must be set")
        else:
            space = 0
    space = int(space)
268
    return DiagonalOperator(
Martin Reinecke's avatar
Martin Reinecke committed
269
270
271
272
        create_power_field(domain[space],
                           power_spectrum, dtype).weight(1),
        domain=domain,
        spaces=space)
273

274

275
276
277
def generate_posterior_sample(mean, covariance):
    """ Generates a posterior sample from a Gaussian distribution with given
    mean and covariance
278

279
280
281
    This method generates samples by setting up the observation and
    reconstruction of a mock signal in order to obtain residuals of the right
    correlation which are added to the given mean.
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297

    Parameters
    ----------
    mean : Field
        the mean of the posterior Gaussian distribution
    covariance : WienerFilterCurvature
        The posterior correlation structure consisting of a
        response operator, noise covariance and prior signal covariance

    Returns
    -------
    sample : Field
        Returns the a sample from the Gaussian of given mean and covariance.

    """

298
299
300
    S = covariance.op.S
    R = covariance.op.R
    N = covariance.op.N
301

302
    power = sqrt(power_analyze(S.diagonal()))
303
    mock_signal = power_synthesize(power, real_signal=True)
304

305
    noise = N.diagonal().weight(-1)
306

307
    mock_noise = Field.from_random(random_type="normal", domain=N.domain,
Martin Reinecke's avatar
Martin Reinecke committed
308
                                   dtype=noise.dtype.type)
309
310
    mock_noise *= sqrt(noise)

Jakob Knollmueller's avatar
Jakob Knollmueller committed
311
    mock_data = R(mock_signal) + mock_noise
312

Jakob Knollmueller's avatar
Jakob Knollmueller committed
313
    mock_j = R.adjoint_times(N.inverse_times(mock_data))
314
315
316
    mock_m = covariance.inverse_times(mock_j)
    sample = mock_signal - mock_m + mean
    return sample
317
318
319
320
321
322
323


def create_composed_fft_operator(domain, codomain=None, all_to='other'):
    fft_op_list = []

    if codomain is None:
        codomain = [None]*len(domain)
324
    interdomain = list(domain.domains)
Martin Reinecke's avatar
Martin Reinecke committed
325
    for i, space in enumerate(domain):
326
327
328
329
330
        if not isinstance(space, Space):
            continue
        if (all_to == 'other' or
                (all_to == 'position' and space.harmonic) or
                (all_to == 'harmonic' and not space.harmonic)):
331
332
333
334
335
336
337
338
            if codomain[i] is None:
                interdomain[i] = domain[i].get_default_codomain()
            else:
                interdomain[i] = codomain[i]
            fft_op_list += [FFTOperator(domain=domain, target=interdomain,
                                        space=i)]
        domain = interdomain
    return ComposedOperator(fft_op_list)