critical_power_energy.py 4.88 KB
Newer Older
1
2
import numpy as np

3
from nifty.energies.energy import Energy
4
5
from nifty.operators.smoothness_operator import SmoothnessOperator
from nifty.library.critical_filter import CriticalPowerCurvature
6
from nifty.energies.memoization import memo
Jakob Knollmueller's avatar
Jakob Knollmueller committed
7

8
from nifty.sugar import generate_posterior_sample
9
from nifty import Field, exp
10

11

12
class CriticalPowerEnergy(Energy):
Jakob Knollmueller's avatar
Jakob Knollmueller committed
13
    """The Energy of the power spectrum according to the critical filter.
14

15
16
17
18
    It describes the energy of the logarithmic amplitudes of the power spectrum
    of a Gaussian random field under reconstruction uncertainty with smoothness
    and inverse gamma prior. It is used to infer the correlation structure of a
    correlated signal.
19
20
21

    Parameters
    ----------
Jakob Knollmueller's avatar
Jakob Knollmueller committed
22
23
24
25
26
27
28
29
30
    position : Field,
        The current position of this energy.
    m : Field,
        The map whichs power spectrum has to be inferred
    D : EndomorphicOperator,
        The curvature of the Gaussian encoding the posterior covariance.
        If not specified, the map is assumed to be no reconstruction.
        default : None
    alpha : float
31
32
        The spectral prior of the inverse gamma distribution. 1.0 corresponds
        to non-informative.
Jakob Knollmueller's avatar
Jakob Knollmueller committed
33
34
        default : 1.0
    q : float
35
36
37
38
39
        The cutoff parameter of the inverse gamma distribution. 0.0 corresponds
        to non-informative.
        default : 0.0
    smoothness_prior : float
        Controls the strength of the smoothness prior
Jakob Knollmueller's avatar
Jakob Knollmueller committed
40
41
42
43
        default : 0.0
    logarithmic : boolean
        Whether smoothness acts on linear or logarithmic scale.
    samples : integer
44
45
        Number of samples used for the estimation of the uncertainty
        corrections.
Jakob Knollmueller's avatar
Jakob Knollmueller committed
46
47
48
49
50
        default : 3
    w : Field
        The contribution from the map with or without uncertainty. It is used
        to pass on the result of the costly sampling during the minimization.
        default : None
51
52
53
    inverter : ConjugateGradient
        The inversion strategy to invert the curvature and to generate samples.
        default : None
54
55
    """

56
57
    # ---Overwritten properties and methods---

58
59
60
    def __init__(self, position, m, D=None, alpha=1.0, q=0.,
                 smoothness_prior=0., logarithmic=True, samples=3, w=None):
        super(CriticalPowerEnergy, self).__init__(position=position)
61
        self.m = m
62
63
        self.D = D
        self.samples = samples
64
65
        self.alpha = Field(self.position.domain, val=alpha)
        self.q = Field(self.position.domain, val=q)
66
        self.T = SmoothnessOperator(domain=self.position.domain[0],
67
                                    strength=smoothness_prior,
Jakob Knollmueller's avatar
Jakob Knollmueller committed
68
                                    logarithmic=logarithmic)
69
        self.rho = self.position.domain[0].rho
70
        self._w = w if w is not None else None
71

72
73
    # ---Mandatory properties and methods---

74
    def at(self, position):
75
76
        return self.__class__(position, self.m, D=self.D, alpha=self.alpha,
                              q=self.q, smoothness_prior=self.smoothness_prior,
77
                              logarithmic=self.logarithmic,
78
                              w=self.w, samples=self.samples)
79
80
81

    @property
    def value(self):
82
        energy = self._theta.sum()
83
84
        energy += self.position.vdot(self._rho_prime, bare=True)
        energy += 0.5 * self.position.vdot(self._Tt)
85
86
87
88
        return energy.real

    @property
    def gradient(self):
89
        gradient = -self._theta.weight(-1)
90
        gradient += (self._rho_prime).weight(-1)
91
        gradient += self._Tt
92
        gradient.val = gradient.val.real
93
94
95
96
        return gradient

    @property
    def curvature(self):
97
98
        curvature = CriticalPowerCurvature(theta=self._theta.weight(-1),
                                           T=self.T)
99
100
        return curvature

101
102
103
104
105
106
107
108
109
110
    # ---Added properties and methods---

    @property
    def logarithmic(self):
        return self.T.logarithmic

    @property
    def smoothness_prior(self):
        return self.T.strength

111
112
113
114
115
116
117
118
119
    @property
    def w(self):
        if self._w is None:
            w = Field(domain=self.position.domain, val=0., dtype=self.m.dtype)
            if self.D is not None:
                for i in range(self.samples):
                    posterior_sample = generate_posterior_sample(
                                                            self.m, self.D)
                    projected_sample = posterior_sample.power_analyze(
Martin Reinecke's avatar
Martin Reinecke committed
120
                     binbounds=self.position.domain[0].binbounds)
121
122
123
124
                    w += (projected_sample) * self.rho
                w /= float(self.samples)
            else:
                w = self.m.power_analyze(
Martin Reinecke's avatar
Martin Reinecke committed
125
                     binbounds=self.position.domain[0].binbounds)
126
127
128
                w *= self.rho
            self._w = w
        return self._w
129

130
131
132
    @property
    @memo
    def _theta(self):
133
        return exp(-self.position) * (self.q + self.w / 2.)
134
135
136
137
138
139
140
141
142
143

    @property
    @memo
    def _rho_prime(self):
        return self.alpha - 1. + self.rho / 2.

    @property
    @memo
    def _Tt(self):
        return self.T(self.position)