/
test_drift.py
162 lines (133 loc) · 8.2 KB
/
test_drift.py
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
import unittest
import numpy as np
import pygsti
from pygsti.extras import drift
from pygsti.modelpacks.legacy import std1Q_XYI
from ..testutils import BaseTestCase
class DriftTestCase(BaseTestCase):
def test_signal(self):
base = 0.5
p = drift.signal.generate_gaussian_signal(10, 23, 10, 1000, base=base, method='sharp')
p = drift.signal.generate_gaussian_signal(10, 23, 10, 1000, base=base, method='logistic')
p = drift.signal.generate_gaussian_signal(10, 23, 10, 1000, base=base, method=None)
p = drift.signal.generate_flat_signal(100, 2, 1000, candidatefreqs=None, base=0.5, method ='logistic')
p = drift.signal.generate_flat_signal(100, 2, 1000, candidatefreqs=np.arange(1,10), base=0.5, method ='sharp')
a1, b1, c1 = drift.signal.spectrum(p, times=np.arange(1000), returnfrequencies=True)
a2, b2, c2 = drift.signal.spectrum(p, times=np.arange(1000), null_hypothesis=None, transform='lsp')
a3, b3, c3 = drift.signal.spectrum(p, times=np.arange(1000), null_hypothesis=p, transform='dft')
power = drift.signal.bartlett_spectrum(p, 5, counts=1, null_hypothesis=None, transform='dct')
assert(abs(max(c1) - 50) < 1e-10)
assert(abs(sum(c1[0:10]) - 100) < 1e-10)
fnc = drift.signal.dct_basisfunction(3, np.arange(500), 0, 500)
amps = drift.signal.amplitudes_at_frequencies([1, 3], {'0': fnc, '1': - fnc})
assert(abs(amps['0'][0]) < 1e-10)
assert(abs(amps['0'][1] - 1) < 1e-10)
assert(abs(amps['1'][0]) < 1e-10)
assert(abs(amps['1'][1] + 1) < 1e-10)
powerthreshold = drift.signal.power_significance_threshold(0.05, 100, 1)
pvaluethreshold = drift.signal.power_to_pvalue(powerthreshold, 1)
assert(abs(pvaluethreshold - 0.05 / 100) < 1e-10)
drift.signal.maxpower_pvalue(20, 100, 1)
qthreshold = drift.signal.power_significance_quasithreshold(0.05, 100, 1, procedure='Benjamini-Hochberg')
assert(abs(qthreshold[-1] - powerthreshold) < 1e-10)
drift.signal.sparsity(p)
plpf = drift.signal.lowpass_filter(p, max_freq=None)
assert(np.sum(abs(p - plpf)) < 1e-7)
assert(abs(drift.signal.moving_average(np.arange(0,100),width=11)[50] - 50) < 1e-10)
def test_probtrajectory(self):
# Create some fake clickstream data, from a constant probability distribution.
numtimes = 500
timstep = 0.1
starttime = 0.
times = (timstep * np.arange(0, numtimes)) + starttime
clickstream = {}
outcomes = ['0','1','2']
for o in outcomes:
clickstream[o] = []
for i in range(len(times)):
click = np.random.randint(0,3)
for o in outcomes:
if int(o) == click:
clickstream[o].append(1)
else:
clickstream[o].append(0)
# Test construction of a constant probability trajectory model
pt = drift.probtrajectory.ConstantProbTrajectory(['0','1','2'],{'0':[0.5],'1':[0.2],})
# Test MLE runs (this calls pretty much everything in the probability trajectory code)
ptmax = drift.probtrajectory.maxlikelihood(pt, clickstream, times, verbosity=1)
parameters = ptmax.parameters_copy()
# The exact MLE is the data mean, so check the returned MLE is close to that.
for o in outcomes[:-1]:
assert(abs(parameters[o][0] - np.mean(clickstream[o])) < 1e-3)
# Check the minimization has actually increased the likelihood from the seed.
assert(drift.probtrajectory.negloglikelihood(ptmax, clickstream, times) <= drift.probtrajectory.negloglikelihood(pt, clickstream, times))
# Test construction of a DCT probability trajectory model
ptdct = drift.probtrajectory.CosineProbTrajectory(['0','1','2'], [0,2], {'0':[0.5,0.02],'1':[0.2,0.03],}, 0, 0.1, 1000)
# Test set parameters from list is working correctly.
ptdct.set_parameters_from_list([0.5,0.02,0.2,0.03])
assert(ptdct.parameters_copy() == {'0':[0.5, 0.02], '1':[0.2, 0.03], })
# Test set parameters from list is working correctly.
assert(ptdct.parameters_as_list() == [0.5, 0.02, 0.2, 0.03])
# Run MLE.
ptdctmax = drift.probtrajectory.maxlikelihood(ptdct, clickstream, times, verbosity=2)
probsmax = ptdctmax.probabilities(times)
# Check the minimization has actually increased the likelihood from the seed.
assert(drift.probtrajectory.negloglikelihood(ptdctmax, clickstream, times) <= drift.probtrajectory.negloglikelihood(ptdct, clickstream, times))
ptdct_invalid = drift.probtrajectory.CosineProbTrajectory(['0','1','2'], [0,2], {'0':[0.5, 0.5],'1':[0.2, 1.2],}, 0, 0.1, 1000)
pt, check = drift.probtrajectory.amplitude_compression(ptdct_invalid, np.linspace(0,1000,2000))
assert(check)
params = pt.parameters_copy()
def test_timeresolvemodel(self):
# A trmodel is a baseclass, and it's pretty trivial.
trmodel = drift.trmodel.TimeResolvedModel([0,],[0.4])
trmodel.parameters_copy()
trmodel.set_parameters([0.2])
trmodel.hyperparameters
@unittest.skip("Need to update this test - do_stability_analysis no longer exists")
def test_core_and_stabilityanalyzer(self):
ds = pygsti.io.read_time_dependent_dataset("cmp_chk_files/timeseries_data_trunc.txt")
fiducial_strs = ['{}', 'Gx', 'Gy', 'GxGx', 'GxGxGx', 'GyGyGy']
germ_strs = ['Gi']
fiducials = [pygsti.objects.Circuit(None, stringrep=fs) for fs in fiducial_strs]
germs = [pygsti.objects.Circuit(None, stringrep=mdl) for mdl in germ_strs]
max_lengths = [256, ]
gssList = pygsti.circuits.make_lsgst_structs(std1Q_XYI.gates, fiducials, fiducials, germs, max_lengths)
gss = gssList[-1]
# Test the integrated routine
results = drift.do_stability_analysis(ds, significance=0.01, estimator='mle')
# Redo but testing direct access of the stability analyzer.
results = drift.StabilityAnalyzer(ds, ids=True)
print(results)
results.compute_spectra()
print(results)
results.run_instability_detection(verbosity=0)
print(results)
results.run_instability_characterization(estimator='filter', modelselector=('default',()),verbosity=0)
print(results)
# Test adding more detectors.
inclass_correction={'dataset': 'Bonferroni', 'circuit': 'Bonferroni', 'spectrum': 'Benjamini-Hochberg'}
results.run_instability_detection(inclass_correction=inclass_correction, tests=(('circuit',),), saveas='fdr-1',
default=False, verbosity=0)
inclass_correction={'dataset': 'Bonferroni', 'circuit': 'Benjamini-Hochberg', 'spectrum': 'Benjamini-Hochberg',}
results.run_instability_detection(inclass_correction=inclass_correction, tests=((),('circuit',),), saveas='fdr-2',
default=False)
results.run_instability_detection(inclass_correction=inclass_correction, tests=(('circuit',),), saveas='fdr-3',
default=False)
results.unstable_circuits(fromtests=[(), ('circuit',)])
freq, s = results.power_spectrum()
circuit = ds.keys()[0]
#Create a workspace to show plots
w = pygsti.report.Workspace()
#w.init_notebook_mode(connected=False, autodisplay=True)
w.ProbTrajectoriesPlot(results, circuit, outcome=('0',))
w.PowerSpectraPlot(results, detectorkey='fdr-2')
w.PowerSpectraPlot(results, spectrumlabel={'circuit': circuit})
w.PowerSpectraPlot(results, spectrumlabel={'dataset': '0', 'circuit': [c for c in ds.keys()[:10]], 'outcome': ('1',)},
showlegend=True)
w.ProbTrajectoriesPlot(results, [c for c in ds.keys()[:5]], outcome=('1', ))
w.GermFiducialProbTrajectoriesPlot(results, gss, 'Gx', 'Gi', 'Gx', outcome=('1', ))
w.GermFiducialPowerSpectraPlot(results, gss, 'Gx', 'Gi', 'Gx')
drift.driftreport.create_drift_report(results, gss, "../temp_test_files/DriftReport", title="Test Drift Report",
verbosity=10)
if __name__ == '__main__':
unittest.main(verbosity=2)