|
1
|
|
|
# -*- coding: utf-8 -*- |
|
2
|
|
|
# vim:fileencoding=utf-8 |
|
3
|
|
|
# |
|
4
|
|
|
# Copyright (c) 2018-2022 Stefan Bender |
|
5
|
|
|
# |
|
6
|
|
|
# This module is part of sciapy. |
|
7
|
|
|
# sciapy is free software: you can redistribute it or modify |
|
8
|
|
|
# it under the terms of the GNU General Public License as published |
|
9
|
|
|
# by the Free Software Foundation, version 2. |
|
10
|
|
|
# See accompanying LICENSE file or http://www.gnu.org/licenses/gpl-2.0.html. |
|
11
|
|
|
"""SCIAMACHY regression module tests |
|
12
|
|
|
""" |
|
13
|
|
|
import numpy as np |
|
14
|
|
|
|
|
15
|
|
|
import pytest |
|
16
|
|
|
|
|
17
|
|
|
try: |
|
18
|
|
|
import pymc3 as pm |
|
19
|
|
|
import arviz as az |
|
20
|
|
|
except ImportError: |
|
21
|
|
|
pytest.skip("Theano/PyMC3 packages not installed", allow_module_level=True) |
|
22
|
|
|
|
|
23
|
|
|
try: |
|
24
|
|
|
from sciapy.regress.models_theano import ( |
|
25
|
|
|
HarmonicModelCosineSine, |
|
26
|
|
|
HarmonicModelAmpPhase, |
|
27
|
|
|
LifetimeModel, |
|
28
|
|
|
ProxyModel, |
|
29
|
|
|
) |
|
30
|
|
|
except ImportError: |
|
31
|
|
|
pytest.skip("Theano/PyMC3 interface not installed", allow_module_level=True) |
|
32
|
|
|
|
|
33
|
|
|
|
|
34
|
|
|
@pytest.fixture(scope="module") |
|
35
|
|
|
def xs(): |
|
36
|
|
|
_xs = np.linspace(0., 11.1, 2048) |
|
37
|
|
|
return np.ascontiguousarray(_xs, dtype=np.float64) |
|
38
|
|
|
|
|
39
|
|
|
|
|
40
|
|
|
def ys(xs, c, s): |
|
41
|
|
|
_ys = c * np.cos(2 * np.pi * xs) + s * np.sin(2 * np.pi * xs) |
|
42
|
|
|
return np.ascontiguousarray(_ys, dtype=np.float64) |
|
43
|
|
|
|
|
44
|
|
|
|
|
45
|
|
|
@pytest.mark.parametrize( |
|
46
|
|
|
"c, s", |
|
47
|
|
|
[ |
|
48
|
|
|
(0.5, 2.0), |
|
49
|
|
|
(1.0, 0.5), |
|
50
|
|
|
(1.0, 1.0), |
|
51
|
|
|
] |
|
52
|
|
|
) |
|
53
|
|
|
def test_harmonics_theano(xs, c, s): |
|
54
|
|
|
# Initialize random number generator |
|
55
|
|
|
np.random.seed(93457) |
|
56
|
|
|
yp = ys(xs, c, s) |
|
57
|
|
|
yp += 0.5 * np.random.randn(xs.shape[0]) |
|
58
|
|
|
|
|
59
|
|
|
with pm.Model() as model1: |
|
60
|
|
|
cos = pm.Normal("cos", mu=0.0, sd=4.0) |
|
61
|
|
|
sin = pm.Normal("sin", mu=0.0, sd=4.0) |
|
62
|
|
|
harm1 = HarmonicModelCosineSine(1., cos, sin) |
|
63
|
|
|
wave1 = harm1.get_value(xs) |
|
64
|
|
|
# add amplitude and phase for comparison |
|
65
|
|
|
pm.Deterministic("amp", harm1.get_amplitude()) |
|
66
|
|
|
pm.Deterministic("phase", harm1.get_phase()) |
|
67
|
|
|
resid1 = yp - wave1 |
|
68
|
|
|
pm.Normal("obs", mu=0.0, observed=resid1) |
|
69
|
|
|
trace1 = pm.sample(tune=800, draws=800, chains=2, return_inferencedata=True) |
|
70
|
|
|
|
|
71
|
|
|
with pm.Model() as model2: |
|
72
|
|
|
amp2 = pm.HalfNormal("amp", sigma=4.0) |
|
73
|
|
|
phase2 = pm.Normal("phase", mu=0.0, sd=4.0) |
|
74
|
|
|
harm2 = HarmonicModelAmpPhase(1., amp2, phase2) |
|
75
|
|
|
wave2 = harm2.get_value(xs) |
|
76
|
|
|
resid2 = yp - wave2 |
|
77
|
|
|
pm.Normal("obs", mu=0.0, observed=resid2) |
|
78
|
|
|
trace2 = pm.sample(tune=800, draws=800, chains=2, return_inferencedata=True) |
|
79
|
|
|
|
|
80
|
|
|
np.testing.assert_allclose( |
|
81
|
|
|
trace1.posterior.median(dim=("chain", "draw"))[["cos", "sin"]].to_array(), |
|
82
|
|
|
(c, s), |
|
83
|
|
|
atol=1e-2, |
|
84
|
|
|
) |
|
85
|
|
|
np.testing.assert_allclose( |
|
86
|
|
|
trace1.posterior.median(dim=("chain", "draw"))[["amp", "phase"]].to_array(), |
|
87
|
|
|
trace2.posterior.median(dim=("chain", "draw"))[["amp", "phase"]].to_array(), |
|
88
|
|
|
atol=3e-3, |
|
89
|
|
|
) |
|
90
|
|
|
|
|
91
|
|
|
|
|
92
|
|
|
def _test_data(xs, c, s): |
|
93
|
|
|
# generate proxy "values" |
|
94
|
|
|
values = ys(xs, c, s) |
|
95
|
|
|
amp = 3. |
|
96
|
|
|
lag = 2. |
|
97
|
|
|
tau0 = 1. |
|
98
|
|
|
harm0 = HarmonicModelCosineSine(1., c, s) |
|
99
|
|
|
tau_lt0 = LifetimeModel(harm0, lower=0.) |
|
100
|
|
|
proxy0 = ProxyModel( |
|
101
|
|
|
xs, values, |
|
102
|
|
|
amp=amp, |
|
103
|
|
|
lag=lag, |
|
104
|
|
|
tau0=tau0, |
|
105
|
|
|
tau_harm=tau_lt0, |
|
106
|
|
|
tau_scan=10, |
|
107
|
|
|
) |
|
108
|
|
|
return proxy0.get_value(xs).eval() |
|
109
|
|
|
|
|
110
|
|
|
|
|
111
|
|
|
@pytest.mark.long |
|
112
|
|
|
def test_proxy_theano(xs, c=3.0, s=1.0): |
|
113
|
|
|
# Initialize random number generator |
|
114
|
|
|
np.random.seed(93457) |
|
115
|
|
|
|
|
116
|
|
|
# proxy "values" |
|
117
|
|
|
values = ys(xs, c, s) |
|
118
|
|
|
|
|
119
|
|
|
yp = _test_data(xs, c, s) |
|
120
|
|
|
yp += 0.5 * np.random.randn(xs.shape[0]) |
|
121
|
|
|
|
|
122
|
|
|
# using "name" prefixes all variables with <name>_ |
|
123
|
|
|
with pm.Model(name="proxy") as model: |
|
124
|
|
|
# amplitude |
|
125
|
|
|
plamp = pm.Normal("log_amp", mu=0.0, sd=np.log(10.0)) |
|
126
|
|
|
pamp = pm.Deterministic("amp", pm.math.exp(plamp)) |
|
127
|
|
|
# lag |
|
128
|
|
|
pllag = pm.Normal("log_lag", mu=0.0, sd=np.log(10.0)) |
|
129
|
|
|
plag = pm.Deterministic("lag", pm.math.exp(pllag)) |
|
130
|
|
|
# lifetime |
|
131
|
|
|
pltau0 = pm.Normal("log_tau0", mu=0.0, sd=np.log(10.0)) |
|
132
|
|
|
ptau0 = pm.Deterministic("tau0", pm.math.exp(pltau0)) |
|
133
|
|
|
cos1 = pm.Normal("tau_cos1", mu=0.0, sd=10.0) |
|
134
|
|
|
sin1 = pm.Normal("tau_sin1", mu=0.0, sd=10.0) |
|
135
|
|
|
harm1 = HarmonicModelCosineSine(1., cos1, sin1) |
|
136
|
|
|
tau1 = LifetimeModel(harm1, lower=0) |
|
137
|
|
|
|
|
138
|
|
|
proxy = ProxyModel( |
|
139
|
|
|
xs, values, |
|
140
|
|
|
amp=pamp, |
|
141
|
|
|
lag=plag, |
|
142
|
|
|
tau0=ptau0, |
|
143
|
|
|
tau_harm=tau1, |
|
144
|
|
|
tau_scan=10, |
|
145
|
|
|
) |
|
146
|
|
|
prox1 = proxy.get_value(xs) |
|
147
|
|
|
# Include "jitter" |
|
148
|
|
|
log_jitter = pm.Normal("log_jitter", mu=0.0, sd=4.0) |
|
149
|
|
|
pm.Normal("obs", mu=prox1, sd=pm.math.exp(log_jitter), observed=yp) |
|
150
|
|
|
|
|
151
|
|
|
maxlp0 = pm.find_MAP() |
|
152
|
|
|
trace = pm.sample( |
|
153
|
|
|
chains=2, |
|
154
|
|
|
draws=1000, |
|
155
|
|
|
tune=1000, |
|
156
|
|
|
init="jitter+adapt_full", |
|
157
|
|
|
random_seed=[286923464, 464329682], |
|
158
|
|
|
return_inferencedata=True, |
|
159
|
|
|
start=maxlp0, |
|
160
|
|
|
target_accept=0.9, |
|
161
|
|
|
) |
|
162
|
|
|
|
|
163
|
|
|
medians = trace.posterior.median(dim=("chain", "draw")) |
|
164
|
|
|
np.testing.assert_allclose( |
|
165
|
|
|
medians[[ |
|
166
|
|
|
"proxy_amp", "proxy_lag", "proxy_tau0", |
|
167
|
|
|
"proxy_tau_cos1", "proxy_tau_sin1", |
|
168
|
|
|
"proxy_log_jitter", |
|
169
|
|
|
]].to_array(), |
|
170
|
|
|
(3., 2., 1., c, s, np.log(0.5)), |
|
171
|
|
|
atol=3e-2, rtol=1e-2, |
|
172
|
|
|
) |
|
173
|
|
|
|
st-bender /
sciapy
