critical_filtering.py 4.95 KB
Newer Older
1
2
3
4
5
6
7
8
9
10

from nifty import *

import plotly.offline as pl
import plotly.graph_objs as go

from mpi4py import MPI
comm = MPI.COMM_WORLD
rank = comm.rank

11
np.random.seed(42)
12

Jakob Knollmueller's avatar
Jakob Knollmueller committed
13

14
def plot_parameters(m,t,p, p_d):
Jakob Knollmueller's avatar
Jakob Knollmueller committed
15
16

    x = log(t.domain[0].kindex)
Jakob Knollmueller's avatar
Jakob Knollmueller committed
17
    m = fft.adjoint_times(m)
18
19
20
21
22
23
    m = m.val.get_full_data().real
    t = t.val.get_full_data().real
    p = p.val.get_full_data().real
    p_d = p_d.val.get_full_data().real
    pl.plot([go.Heatmap(z=m)], filename='map.html')
    pl.plot([go.Scatter(x=x,y=t), go.Scatter(x=x ,y=p), go.Scatter(x=x, y=p_d)], filename="t.html")
Jakob Knollmueller's avatar
Jakob Knollmueller committed
24

Jakob Knollmueller's avatar
Jakob Knollmueller committed
25
26
27
28

class AdjointFFTResponse(LinearOperator):
    def __init__(self, FFT, R, default_spaces=None):
        super(AdjointFFTResponse, self).__init__(default_spaces)
29
        self._domain = FFT.target
Jakob Knollmueller's avatar
Jakob Knollmueller committed
30
31
        self._target = R.target
        self.R = R
32
33
34
        self.FFT = FFT

    def _times(self, x, spaces=None):
Jakob Knollmueller's avatar
Jakob Knollmueller committed
35
        return self.R(self.FFT.adjoint_times(x))
36
37

    def _adjoint_times(self, x, spaces=None):
Jakob Knollmueller's avatar
Jakob Knollmueller committed
38
        return self.FFT(self.R.adjoint_times(x))
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
    @property
    def domain(self):
        return self._domain

    @property
    def target(self):
        return self._target

    @property
    def unitary(self):
        return False

if __name__ == "__main__":

    distribution_strategy = 'not'

    # Set up position space
    s_space = RGSpace([128,128])
    # s_space = HPSpace(32)

    # Define harmonic transformation and associated harmonic space
    fft = FFTOperator(s_space)
    h_space = fft.target[0]

    # Setting up power space
64
    p_space = PowerSpace(h_space, logarithmic=True,
65
                         distribution_strategy=distribution_strategy)
66
67

    # Choosing the prior correlation structure and defining correlation operator
68
    p_spec = (lambda k: (.5 / (k + 1) ** 3))
Jakob Knollmueller's avatar
Jakob Knollmueller committed
69
    S = create_power_operator(h_space, power_spectrum=p_spec,
70
71
72
                              distribution_strategy=distribution_strategy)

    # Drawing a sample sh from the prior distribution in harmonic space
Jakob Knollmueller's avatar
Jakob Knollmueller committed
73
    sp = Field(p_space,  val=p_spec,
74
75
76
77
78
               distribution_strategy=distribution_strategy)
    sh = sp.power_synthesize(real_signal=True)


    # Choosing the measurement instrument
Jakob Knollmueller's avatar
Jakob Knollmueller committed
79
80
    # Instrument = SmoothingOperator(s_space, sigma=0.01)
    Instrument = DiagonalOperator(s_space, diagonal=1.)
Jakob Knollmueller's avatar
Jakob Knollmueller committed
81
    # Instrument._diagonal.val[200:400, 200:400] = 0
Jakob Knollmueller's avatar
Jakob Knollmueller committed
82
    #Instrument._diagonal.val[64:512-64, 64:512-64] = 0
83
84
85


    #Adding a harmonic transformation to the instrument
Jakob Knollmueller's avatar
Jakob Knollmueller committed
86
87
    R = AdjointFFTResponse(fft, Instrument)

88
    noise = 1.
89
90
91
92
93
94
95
96
    N = DiagonalOperator(s_space, diagonal=noise, bare=True)
    n = Field.from_random(domain=s_space,
                          random_type='normal',
                          std=sqrt(noise),
                          mean=0)

    # Creating the mock data
    d = R(sh) + n
Jakob Knollmueller's avatar
Jakob Knollmueller committed
97

98
99
    # The information source
    j = R.adjoint_times(N.inverse_times(d))
Jakob Knollmueller's avatar
Jakob Knollmueller committed
100
    realized_power = log(sh.power_analyze(logarithmic=p_space.config["logarithmic"],
Jakob Knollmueller's avatar
Jakob Knollmueller committed
101
                                          nbin=p_space.config["nbin"]))
102
    data_power = log(fft(d).power_analyze(logarithmic=p_space.config["logarithmic"],
Jakob Knollmueller's avatar
Jakob Knollmueller committed
103
                                          nbin=p_space.config["nbin"]))
104
105
106
107
    d_data = d.val.get_full_data().real
    if rank == 0:
        pl.plot([go.Heatmap(z=d_data)], filename='data.html')

Jakob Knollmueller's avatar
Jakob Knollmueller committed
108
    #  minimization strategy
109
110
111
112
113
114

    def convergence_measure(a_energy, iteration): # returns current energy
        x = a_energy.value
        print (x, iteration)


115
116
    minimizer1 = RelaxedNewton(convergence_tolerance=10e-2,
                              convergence_level=2,
117
                              iteration_limit=3,
118
                              callback=convergence_measure)
119

Jakob Knollmueller's avatar
Jakob Knollmueller committed
120
    minimizer2 = VL_BFGS(convergence_tolerance=0,
Jakob Knollmueller's avatar
Jakob Knollmueller committed
121
                       iteration_limit=7,
Jakob Knollmueller's avatar
Jakob Knollmueller committed
122
                       callback=convergence_measure,
Jakob Knollmueller's avatar
Jakob Knollmueller committed
123
                       max_history_length=3)
124

125
126
127
    inverter = ConjugateGradient(convergence_level=1,
                                 convergence_tolerance=10e-4,
                                 preconditioner=None)
128
129
130
131
    # Setting starting position
    flat_power = Field(p_space,val=10e-8)
    m0 = flat_power.power_synthesize(real_signal=True)

132
133
    t0 = Field(p_space, val=log(1./(1+p_space.kindex)**2))

Jakob Knollmueller's avatar
Jakob Knollmueller committed
134

135

136
    for i in range(500):
Jakob Knollmueller's avatar
Jakob Knollmueller committed
137
        S0 = create_power_operator(h_space, power_spectrum=exp(t0),
138
139
140
                              distribution_strategy=distribution_strategy)

        # Initializing the  nonlinear Wiener Filter energy
141
142
        map_energy = WienerFilterEnergy(position=m0, d=d, R=R, N=N, S=S0, inverter=inverter)
        # Solving the Wiener Filter analytically
143
        D0 = map_energy.curvature
144
        m0 = D0.inverse_times(j)
145
        # Initializing the power energy with updated parameters
146
        power_energy = CriticalPowerEnergy(position=t0, m=m0, D=D0, sigma=10., samples=3, inverter=inverter)
147
148
149

        (power_energy, convergence) = minimizer1(power_energy)

150

151
        # Setting new power spectrum
152
        t0.val  = power_energy.position.val.real
Jakob Knollmueller's avatar
Jakob Knollmueller committed
153

Jakob Knollmueller's avatar
Jakob Knollmueller committed
154
        # Plotting current estimate
155
        plot_parameters(m0,t0,log(sp), data_power)
156

Jakob Knollmueller's avatar
Jakob Knollmueller committed
157