def test_mapcalc_hessian(self):
chi2, chi2Hessian = pygsti.chi2(self.gateset, self.ds,
returnHessian=True)
gs_mapcalc = self.gateset.copy()
gs_mapcalc._calcClass = GateMapCalc
chi2, chi2Hessian_mapcalc = pygsti.chi2(self.gateset, self.ds,
returnHessian=True)
self.assertArraysAlmostEqual(chi2Hessian, chi2Hessian_mapcalc)
def test_mapcalc_hessian(self):
chi2, chi2Hessian = pygsti.chi2(self.model, self.ds,
returnHessian=True)
mdl_mapcalc = self.model.copy()
mdl_mapcalc._calcClass = MapForwardSimulator
chi2, chi2Hessian_mapcalc = pygsti.chi2(self.model, self.ds,
returnHessian=True)
self.assertArraysAlmostEqual(chi2Hessian, chi2Hessian_mapcalc)
def test_chi2_fn(self):
ds = pygsti.objects.DataSet(fileToLoadFrom="cmp_chk_files/analysis.dataset")
chi2, grad = pygsti.chi2(ds, std.gs_target, returnGradient=True)
pygsti.gate_string_chi2( ('Gx',), ds, std.gs_target)
pygsti.chi2fn_2outcome( N=100, p=0.5, f=0.6)
pygsti.chi2fn_2outcome_wfreqs( N=100, p=0.5, f=0.6)
pygsti.chi2fn( N=100, p=0.5, f=0.6)
pygsti.chi2fn_wfreqs( N=100, p=0.5, f=0.6)
with self.assertRaises(ValueError):
pygsti.chi2(ds, std.gs_target, useFreqWeightedChiSq=True) #no impl yet
def test_hessian_projection(self):
chi2, chi2Grad, chi2Hessian = pygsti.chi2(self.ds, self.gateset,
returnGradient=True,
returnHessian=True)
proj_non_gauge = self.gateset.get_nongauge_projector()
projectedHessian = np.dot(proj_non_gauge,
np.dot(chi2Hessian, proj_non_gauge))
self.assertEqual( projectedHessian.shape, (56,56) )
self.assertEqual( np.linalg.matrix_rank(proj_non_gauge), 40)
self.assertEqual( np.linalg.matrix_rank(projectedHessian), 40)
eigvals = np.sort(abs(np.linalg.eigvals(projectedHessian)))
eigvals_chk = np.array(
[ 2.36932621e-11, 4.57739349e-11, 8.49888013e-11, 8.49888013e-11,
1.22859895e-10, 1.22859895e-10, 1.38705957e-10, 1.38705957e-10,
3.75441328e-10, 6.46644807e-10, 6.46644807e-10, 7.06181642e-10,
7.06181642e-10, 7.65472749e-10, 1.62672899e-09, 1.62672899e-09,
9.23280752e+04, 9.88622140e+04, 1.24577730e+05, 1.40500652e+05,
1.88461602e+05, 1.98466608e+05, 2.07084497e+05, 2.27489018e+05,
2.33403285e+05, 2.52573024e+05, 2.82350782e+05, 3.12087185e+05,
3.21855420e+05, 3.31659734e+05, 3.52649174e+05, 3.60682071e+05,
3.90777833e+05, 4.59913853e+05, 5.02652879e+05, 5.59311926e+05,
5.78891070e+05, 6.82325323e+05, 7.57318263e+05, 8.16739390e+05,
1.06466062e+06, 1.20075694e+06, 1.37368639e+06, 1.58356629e+06,
1.68898356e+06, 2.12277359e+06, 3.30650801e+06, 3.75869331e+06,
4.00195245e+06, 4.42427797e+06, 5.06956256e+06, 7.31166332e+06,
9.19432790e+06, 9.99944236e+06, 1.31027722e+07, 5.80310818e+07] )
for val,chk in zip(eigvals,eigvals_chk):
self.assertAlmostEqual(abs(val-chk)/(abs(chk)+1e-6), 0.0, places=3)
def test_model_selection(self):
ds = self.ds
#pygsti.construction.generate_fake_data(self.datagen_gateset, self.lsgstStrings[-1],
# nSamples=1000,sampleError='binomial', seed=100)
mdl_lgst4 = pygsti.do_lgst(ds, self.fiducials, self.fiducials, self.model, svdTruncateTo=4, verbosity=0)
mdl_lgst6 = pygsti.do_lgst(ds, self.fiducials, self.fiducials, self.model, svdTruncateTo=6, verbosity=0)
sys.stdout.flush()
self.runSilent(pygsti.do_lgst, ds, self.fiducials, self.fiducials, self.model, svdTruncateTo=6, verbosity=4) # test verbose prints
chiSq4 = pygsti.chi2(mdl_lgst4, ds, self.lgstStrings, minProbClipForWeighting=1e-4)
chiSq6 = pygsti.chi2(mdl_lgst6, ds, self.lgstStrings, minProbClipForWeighting=1e-4)
print("LGST dim=4 chiSq = ",chiSq4)
print("LGST dim=6 chiSq = ",chiSq6)
#self.assertAlmostEqual(chiSq4, 174.061524953) #429.271983052)
#self.assertAlmostEqual(chiSq6, 267012993.861, places=1) #1337.74222467) #Why is this so large??? -- DEBUG later
# Least squares GST with model selection
mdl_lsgst = self.runSilent(pygsti.do_iterative_mc2gst_with_model_selection, ds, mdl_lgst4, 1, self.lsgstStrings[0:3],
verbosity=10, minProbClipForWeighting=1e-3, probClipInterval=(-1e5,1e5))
# Run again with other parameters
tuple_strings = [ list(map(tuple, gsList)) for gsList in self.lsgstStrings[0:3] ] #to test tuple argument
errorVecs, mdl_lsgst_wts = self.runSilent(pygsti.do_iterative_mc2gst_with_model_selection, ds, mdl_lgst4,
1, tuple_strings, verbosity=10, minProbClipForWeighting=1e-3,
probClipInterval=(-1e5,1e5), circuitWeightsDict={ ('Gx',): 2.0 },
returnAll=True, returnErrorVec=True)
# Do non-iterative to cover Circuit->tuple conversion
mdl_non_iterative = self.runSilent( pygsti.do_mc2gst_with_model_selection, ds,
mdl_lgst4, 1, self.lsgstStrings[0],
verbosity=10, probClipInterval=(-1e5,1e5) )
# RUN BELOW LINES TO SEED SAVED GATESET FILES
if os.environ.get('PYGSTI_REGEN_REF_FILES','no').lower() in ("yes","1","true"):
pygsti.io.write_model(mdl_lsgst,compare_files + "/lsgstMS.model", "Saved LSGST Model with model selection")
mdl_lsgst_compare = pygsti.io.load_model(compare_files + "/lsgstMS.model")
mdl_lsgst_go = pygsti.gaugeopt_to_target(mdl_lsgst, mdl_lsgst_compare, {'spam':1.0}, checkJac=True)
self.assertAlmostEqual( mdl_lsgst_go.frobeniusdist(mdl_lsgst_compare), 0, places=4)
def test_model_selection(self):
ds = self.ds
#pygsti.construction.generate_fake_data(self.datagen_gateset, self.lsgstStrings[-1],
# nSamples=1000,sampleError='binomial', seed=100)
gs_lgst4 = pygsti.do_lgst(ds, self.specs, self.gateset, svdTruncateTo=4, verbosity=0)
gs_lgst6 = pygsti.do_lgst(ds, self.specs, self.gateset, svdTruncateTo=6, verbosity=0)
sys.stdout.flush()
self.runSilent(pygsti.do_lgst, ds, self.specs, self.gateset, svdTruncateTo=6, verbosity=4) # test verbose prints
chiSq4 = pygsti.chi2(ds, gs_lgst4, self.lgstStrings, minProbClipForWeighting=1e-4)
chiSq6 = pygsti.chi2(ds, gs_lgst6, self.lgstStrings, minProbClipForWeighting=1e-4)
print("LGST dim=4 chiSq = ",chiSq4)
print("LGST dim=6 chiSq = ",chiSq6)
#self.assertAlmostEqual(chiSq4, 174.061524953) #429.271983052)
#self.assertAlmostEqual(chiSq6, 267012993.861, places=1) #1337.74222467) #Why is this so large??? -- DEBUG later
# Least squares GST with model selection
gs_lsgst = self.runSilent(pygsti.do_iterative_mc2gst_with_model_selection, ds, gs_lgst4, 1, self.lsgstStrings[0:3],
verbosity=10, minProbClipForWeighting=1e-3, probClipInterval=(-1e5,1e5))
# Run again with other parameters
tuple_strings = [ list(map(tuple, gsList)) for gsList in self.lsgstStrings[0:3] ] #to test tuple argument
errorVecs, gs_lsgst_wts = self.runSilent(pygsti.do_iterative_mc2gst_with_model_selection, ds, gs_lgst4,
1, tuple_strings, verbosity=10, minProbClipForWeighting=1e-3,
probClipInterval=(-1e5,1e5), gatestringWeightsDict={ ('Gx',): 2.0 },
returnAll=True, returnErrorVec=True)
# Do non-iterative to cover GateString->tuple conversion
gs_non_iterative = self.runSilent( pygsti.do_mc2gst_with_model_selection, ds,
gs_lgst4, 1, self.lsgstStrings[0],
verbosity=10, probClipInterval=(-1e5,1e5) )
# RUN BELOW LINES TO SEED SAVED GATESET FILES
#pygsti.io.write_gateset(gs_lsgst,compare_files + "/lsgstMS.gateset", "Saved LSGST Gateset with model selection")
gs_lsgst_compare = pygsti.io.load_gateset(compare_files + "/lsgstMS.gateset")
gs_lsgst_go = pygsti.optimize_gauge(gs_lsgst, 'target', targetGateset=gs_lsgst_compare, spamWeight=1.0)
self.assertAlmostEqual( gs_lsgst_go.frobeniusdist(gs_lsgst_compare), 0, places=5)
def test_chi2_fn(self):
ds = pygsti.objects.DataSet(fileToLoadFrom=compare_files + "/analysis.dataset")
chi2, grad = pygsti.chi2(ds, std.gs_target, returnGradient=True)
pygsti.chi2(ds, std.gs_target, returnHessian=True)
pygsti.gate_string_chi2( ('Gx',), ds, std.gs_target)
pygsti.chi2fn_2outcome( N=100, p=0.5, f=0.6)
pygsti.chi2fn_2outcome_wfreqs( N=100, p=0.5, f=0.6)
pygsti.chi2fn( N=100, p=0.5, f=0.6)
pygsti.chi2fn_wfreqs( N=100, p=0.5, f=0.6)
with self.assertRaises(ValueError):
pygsti.chi2(ds, std.gs_target, useFreqWeightedChiSq=True) #no impl yet
# Memory tests
with self.assertRaises(MemoryError):
pygsti.chi2(ds, std.gs_target, memLimit=0) # No memory for you
pygsti.chi2(ds, std.gs_target, memLimit=100000)
def test_chi2_fn(self):
ds = pygsti.objects.DataSet(fileToLoadFrom=compare_files +
"/analysis.dataset%s" % self.versionsuffix)
chi2, grad = pygsti.chi2(std.gs_target, ds, returnGradient=True)
pygsti.chi2(std.gs_target, ds, returnHessian=True)
#pygsti.gate_string_chi2( ('Gx',), ds, std.gs_target)
pygsti.chi2fn_2outcome(N=100, p=0.5, f=0.6)
pygsti.chi2fn_2outcome_wfreqs(N=100, p=0.5, f=0.6)
pygsti.chi2fn(N=100, p=0.5, f=0.6)
pygsti.chi2fn_wfreqs(N=100, p=0.5, f=0.6)
with self.assertRaises(ValueError):
pygsti.chi2(std.gs_target, ds,
useFreqWeightedChiSq=True) #no impl yet
# Memory tests
with self.assertRaises(MemoryError):
pygsti.chi2(std.gs_target, ds, memLimit=0) # No memory for you
pygsti.chi2(std.gs_target, ds, memLimit=100000)
def test_hessian_projection(self):
chi2, chi2Grad, chi2Hessian = pygsti.chi2(self.model, self.ds,
returnGradient=True,
returnHessian=True)
proj_non_gauge = self.model.get_nongauge_projector()
projectedHessian = np.dot(proj_non_gauge,
np.dot(chi2Hessian, proj_non_gauge))
print(self.model.num_params())
print(proj_non_gauge.shape)
self.assertEqual( projectedHessian.shape, (60,60) )
#print("Evals = ")
#print("\n".join( [ "%d: %g" % (i,ev) for i,ev in enumerate(np.linalg.eigvals(projectedHessian))] ))
self.assertEqual( np.linalg.matrix_rank(proj_non_gauge), 44)
self.assertEqual( np.linalg.matrix_rank(projectedHessian), 44)
eigvals = np.sort(abs(np.linalg.eigvals(projectedHessian)))
print("eigvals = ",eigvals)
eigvals_chk = np.array([2.51663034e-10, 2.51663034e-10, 6.81452335e-10, 7.72039792e-10,
8.76915081e-10, 8.76915081e-10, 1.31455011e-09, 3.03808236e-09,
3.03808236e-09, 3.13457752e-09, 3.21805358e-09, 3.21805358e-09,
4.78549720e-09, 7.83389490e-09, 1.82493106e-08, 1.82493106e-08,
9.23087831e+05, 1.05783101e+06, 1.16457705e+06, 1.39492929e+06,
1.84015484e+06, 2.10613947e+06, 2.37963392e+06, 2.47192689e+06,
2.64566761e+06, 2.68722871e+06, 2.82383377e+06, 2.86584033e+06,
2.94590436e+06, 2.96180212e+06, 3.08322015e+06, 3.29389050e+06,
3.66581786e+06, 3.76266448e+06, 3.81921738e+06, 3.86624688e+06,
3.89045873e+06, 4.72831630e+06, 4.96416855e+06, 6.53286834e+06,
1.01424911e+07, 1.11347312e+07, 1.26152967e+07, 1.30081040e+07,
1.36647082e+07, 1.49293583e+07, 1.58234599e+07, 1.80999182e+07,
2.09155048e+07, 2.17444267e+07, 2.46870311e+07, 2.64427393e+07,
2.72410297e+07, 3.34988002e+07, 3.45005948e+07, 3.69040745e+07,
5.08647137e+07, 9.43153151e+07, 1.36088308e+08, 6.30304807e+08])
TOL = 1e-7
for val,chk in zip(eigvals,eigvals_chk):
if abs(val) > TOL or abs(chk) > TOL:
self.assertAlmostEqual(abs(val-chk)/(abs(chk)+TOL), 0.0, places=3)
def test_confidenceRegion(self):
chi2, chi2Hessian = pygsti.chi2(self.ds, self.gateset, returnHessian=True)
ci_std = pygsti.obj.ConfidenceRegion(self.gateset, chi2Hessian, 95.0, hessianProjection="std")
ci_noproj = pygsti.obj.ConfidenceRegion(self.gateset, chi2Hessian, 95.0, hessianProjection="none")
ci_opt = pygsti.obj.ConfidenceRegion(
self.gateset, chi2Hessian, 95.0, hessianProjection="optimal gate CIs", tol=0.1
) # very low tol so doesn't take long
with self.assertRaises(ValueError):
pygsti.obj.ConfidenceRegion(
self.gateset, chi2Hessian, 95.0, hessianProjection="FooBar"
) # bad hessianProjection
self.assertWarns(
pygsti.obj.ConfidenceRegion, self.gateset, chi2Hessian, 0.95, hessianProjection="none"
) # percentage < 1.0
ar_of_intervals_Gx = ci_std.get_profile_likelihood_confidence_intervals("Gx")
ar_of_intervals_rho0 = ci_std.get_profile_likelihood_confidence_intervals("rho0")
ar_of_intervals_E0 = ci_std.get_profile_likelihood_confidence_intervals("E0")
ar_of_intervals = ci_std.get_profile_likelihood_confidence_intervals()
def fnOfGate_float(mx):
return float(mx[0, 0])
def fnOfGate_0D(mx):
return np.array(float(mx[0, 0]))
def fnOfGate_1D(mx):
return mx[0, :]
def fnOfGate_2D(mx):
return mx[:, :]
def fnOfGate_3D(mx):
return np.zeros((2, 2, 2), "d") # just to test for error
df = ci_std.get_gate_fn_confidence_interval(fnOfGate_float, "Gx", verbosity=0)
df = ci_std.get_gate_fn_confidence_interval(fnOfGate_0D, "Gx", verbosity=0)
df = ci_std.get_gate_fn_confidence_interval(fnOfGate_1D, "Gx", verbosity=0)
df = ci_std.get_gate_fn_confidence_interval(fnOfGate_2D, "Gx", verbosity=0)
df, f0 = self.runSilent(
ci_std.get_gate_fn_confidence_interval, fnOfGate_float, "Gx", returnFnVal=True, verbosity=4
)
with self.assertRaises(ValueError):
ci_std.get_gate_fn_confidence_interval(fnOfGate_3D, "Gx", verbosity=0)
def fnOfSpam_float(rhoVecs, EVecs):
return float(np.dot(rhoVecs[0].T, EVecs[0]))
def fnOfSpam_0D(rhoVecs, EVecs):
return np.array(float(np.dot(rhoVecs[0].T, EVecs[0])))
def fnOfSpam_1D(rhoVecs, EVecs):
return np.array([np.dot(rhoVecs[0].T, EVecs[0]), 0])
def fnOfSpam_2D(rhoVecs, EVecs):
return np.array([[np.dot(rhoVecs[0].T, EVecs[0]), 0], [0, 0]])
def fnOfSpam_3D(rhoVecs, EVecs):
return np.zeros((2, 2, 2), "d") # just to test for error
df = ci_std.get_spam_fn_confidence_interval(fnOfSpam_float, verbosity=0)
df = ci_std.get_spam_fn_confidence_interval(fnOfSpam_0D, verbosity=0)
df = ci_std.get_spam_fn_confidence_interval(fnOfSpam_1D, verbosity=0)
df = ci_std.get_spam_fn_confidence_interval(fnOfSpam_2D, verbosity=0)
df, f0 = self.runSilent(ci_std.get_spam_fn_confidence_interval, fnOfSpam_float, returnFnVal=True, verbosity=4)
with self.assertRaises(ValueError):
ci_std.get_spam_fn_confidence_interval(fnOfSpam_3D, verbosity=0)
def test_table_generation(self):
import pygsti.report.generation as gen
formats = ['latex','html','test','ppt'] #all the formats we know
tableclass = "cssClass"
longtable = False
confidenceRegionInfo = None
gateset = pygsti.io.load_gateset(compare_files + "/analysis.gateset")
ds = pygsti.objects.DataSet(fileToLoadFrom=compare_files + "/analysis.dataset")
chi2, chi2Hessian = pygsti.chi2(ds, gateset, returnHessian=True)
ci = pygsti.obj.ConfidenceRegion(gateset, chi2Hessian, 95.0,
hessianProjection="std")
gateset_tp = pygsti.contract(gateset,"TP"); gateset_tp.set_all_parameterizations("TP")
chi2, chi2Hessian_TP = pygsti.chi2(ds, gateset_tp, returnHessian=True)
ci_TP = pygsti.obj.ConfidenceRegion(gateset_tp, chi2Hessian_TP, 95.0,
hessianProjection="std")
chi2, chi2Hessian_tgt = pygsti.chi2(ds, std.gs_target, returnHessian=True)
ci_tgt = pygsti.obj.ConfidenceRegion(std.gs_target, chi2Hessian_tgt, 95.0,
hessianProjection="std")
target_tp = std.gs_target.copy(); target_tp.set_all_parameterizations("TP")
chi2, chi2Hessian_tgt_TP = pygsti.chi2(ds, target_tp, returnHessian=True)
ci_TP_tgt = pygsti.obj.ConfidenceRegion(target_tp, chi2Hessian_tgt_TP, 95.0,
hessianProjection="std")
gateset_2q = pygsti.construction.build_gateset(
[4], [('Q0','Q1')],['GIX','GIY','GXI','GYI','GCNOT'],
[ "I(Q0):X(pi/2,Q1)", "I(Q0):Y(pi/2,Q1)", "X(pi/2,Q0):I(Q1)", "Y(pi/2,Q0):I(Q1)", "CX(pi,Q0,Q1)" ],
prepLabels=['rho0'], prepExpressions=["0"], effectLabels=['E0','E1','E2'], effectExpressions=["0","1","2"],
spamdefs={'upup': ('rho0','E0'), 'updn': ('rho0','E1'), 'dnup': ('rho0','E2'),
'dndn': ('rho0','remainder') }, basis="pp" )
#tests which fill in the cracks of the full-report tests
tab = gen.get_gateset_spam_table(gateset, None)
tab_wCI = gen.get_gateset_spam_table(gateset, ci)
table_wCI_as_str = str(tab_wCI)
gen.get_gateset_spam_table(gateset, None)
gen.get_gateset_gates_table(gateset_tp, ci_TP) #test zero-padding
gen.get_unitary_gateset_gates_table(std.gs_target, ci_tgt) #unitary gates w/CIs
gen.get_unitary_gateset_gates_table(target_tp, ci_TP_tgt) #unitary gates w/CIs
gen.get_gateset_closest_unitary_table(gateset_2q, None) #test higher-dim gateset
gen.get_gateset_closest_unitary_table(gateset, ci) #test with CIs (long...)
gen.get_gateset_rotn_axis_table(std.gs_target, None) #try to get "--"s and "X"s to display
gen.get_chi2_progress_table([0], [gateset_tp], [ [('Gx',)],], ds) #TP case
gen.get_chi2_confidence_region(gateset_tp, ds, 95) #TP case
gen.get_gatestring_multi_table([ [('Gx',),('Gz',)], [('Gy',)] ],
["list1","list2"], commonTitle=None) #commonTitle == None case w/diff length lists
with self.assertRaises(ValueError):
gen.get_unitary_gateset_gates_table(std.gs_target, ci) #gateset-CI mismatch
with self.assertRaises(ValueError):
gen.get_gateset_spam_parameters_table(std.gs_target, ci) #gateset-CI mismatch
#LogL case tests
gen.get_logl_progress_table([0], [gateset_tp], [ [('Gx',)],], ds) # logL case
gen.get_logl_progress_table([0], [gateset], [ [('Gx',)],], ds) # logL case
gen.get_logl_confidence_region(gateset_tp, ds, 95,
gatestring_list=None, probClipInterval=(-1e6,1e6),
minProbClip=1e-4, radius=1e-4, hessianProjection="std")
gen.get_logl_confidence_region(gateset, ds, 95,
gatestring_list=None, probClipInterval=(-1e6,1e6),
minProbClip=1e-4, radius=1e-4, hessianProjection="std")
def test_total_chi2(self):
chi2 = pygsti.chi2(self.ds, self.lsgst_gateset)
self.assertAlmostEqual(chi2, 729.182015795)
def test_table_generation(self):
import pygsti.report.generation as gen
formats = ['latex','html','py','ppt'] #all the formats we know
tableclass = "cssClass"
longtable = False
confidenceRegionInfo = None
gateset = pygsti.io.load_gateset("cmp_chk_files/analysis.gateset")
ds = pygsti.objects.DataSet(fileToLoadFrom="cmp_chk_files/analysis.dataset")
chi2, chi2Hessian = pygsti.chi2(ds, gateset, returnHessian=True)
ci = pygsti.obj.ConfidenceRegion(gateset, chi2Hessian, 95.0,
hessianProjection="std")
gateset_tp = pygsti.contract(gateset,"TP"); gateset_tp.set_all_parameterizations("TP")
chi2, chi2Hessian_TP = pygsti.chi2(ds, gateset_tp, returnHessian=True)
ci_TP = pygsti.obj.ConfidenceRegion(gateset_tp, chi2Hessian_TP, 95.0,
hessianProjection="std")
chi2, chi2Hessian_tgt = pygsti.chi2(ds, std.gs_target, returnHessian=True)
ci_tgt = pygsti.obj.ConfidenceRegion(std.gs_target, chi2Hessian_tgt, 95.0,
hessianProjection="std")
target_tp = std.gs_target.copy(); target_tp.set_all_parameterizations("TP")
chi2, chi2Hessian_tgt_TP = pygsti.chi2(ds, target_tp, returnHessian=True)
ci_TP_tgt = pygsti.obj.ConfidenceRegion(target_tp, chi2Hessian_tgt_TP, 95.0,
hessianProjection="std")
gateset_2q = pygsti.construction.build_gateset(
[4], [('Q0','Q1')],['GIX','GIY','GXI','GYI','GCNOT'],
[ "I(Q0):X(pi/2,Q1)", "I(Q0):Y(pi/2,Q1)", "X(pi/2,Q0):I(Q1)", "Y(pi/2,Q0):I(Q1)", "CX(pi,Q0,Q1)" ],
prepLabels=['rho0'], prepExpressions=["0"], effectLabels=['E0','E1','E2'], effectExpressions=["0","1","2"],
spamdefs={'upup': ('rho0','E0'), 'updn': ('rho0','E1'), 'dnup': ('rho0','E2'),
'dndn': ('rho0','remainder') }, basis="pp" )
#tests which fill in the cracks of the full-report tests
with self.assertRaises(ValueError):
gen.get_gateset_spam_table(gateset, formats, tableclass,
longtable, None, mxBasis="fooBar")
tab = gen.get_gateset_spam_table(gateset, formats, tableclass, longtable, None, mxBasis="gm")
tab_wCI = gen.get_gateset_spam_table(gateset, formats, tableclass, longtable, ci, mxBasis="gm")
gen.get_gateset_spam_table(gateset, formats, tableclass, longtable, None, mxBasis="std")
gen.get_gateset_spam_table(gateset, formats, tableclass, longtable, None, mxBasis="pp")
gen.get_gateset_gates_table(gateset_tp, formats, tableclass, longtable, ci_TP) #test zero-padding
gen.get_unitary_gateset_gates_table(std.gs_target, formats, tableclass, longtable, ci_tgt) #unitary gates w/CIs
gen.get_unitary_gateset_gates_table(target_tp, formats, tableclass, longtable, ci_TP_tgt) #unitary gates w/CIs
gen.get_gateset_closest_unitary_table(gateset_2q, formats, tableclass, longtable, None) #test higher-dim gateset
gen.get_gateset_closest_unitary_table(gateset, formats, tableclass, longtable, ci) #test with CIs (long...)
gen.get_gateset_rotn_axis_table(std.gs_target, formats, tableclass, longtable, None) #try to get "--"s and "X"s to display
gen.get_chi2_progress_table([0], [gateset_tp], [ [('Gx',)],], ds, formats, tableclass, longtable) #TP case
gen.get_chi2_confidence_region(gateset_tp, ds, 95) #TP case
gen.get_gatestring_multi_table([ [('Gx',),('Gz',)], [('Gy',)] ], ["list1","list2"], formats,
tableclass, longtable, commonTitle=None) #commonTitle == None case w/diff length lists
with self.assertRaises(ValueError):
gen.get_unitary_gateset_gates_table(std.gs_target, formats, tableclass, longtable, ci) #gateset-CI mismatch
with self.assertRaises(ValueError):
gen.get_gateset_spam_parameters_table(std.gs_target, formats, tableclass, longtable, ci) #gateset-CI mismatch
#Test ReportTable object
rowLabels = tab.keys()
row1Data = tab[rowLabels[0]]
colLabels = row1Data.keys()
self.assertTrue(rowLabels, tab.row_names)
self.assertTrue(colLabels, tab.col_names)
self.assertTrue(len(rowLabels), tab.num_rows)
self.assertTrue(len(colLabels), tab.num_cols)
el00 = tab[rowLabels[0]][colLabels[0]]
self.assertTrue( rowLabels[0] in tab )
self.assertTrue( tab.has_key(rowLabels[0]) )
table_len = len(tab)
self.assertEqual(table_len, tab.num_rows)
table_as_str = str(tab)
table_wCI_as_str = str(tab_wCI)
row1a = tab.row(key=rowLabels[0])
col1a = tab.col(key=colLabels[0])
row1b = tab.row(index=0)
col1b = tab.col(index=0)
self.assertEqual(row1a,row1b)
self.assertEqual(col1a,col1b)
with self.assertRaises(KeyError):
tab['foobar']
with self.assertRaises(KeyError):
tab.row(key='foobar') #invalid key
with self.assertRaises(ValueError):
tab.row(index=100000) #out of bounds
with self.assertRaises(ValueError):
tab.row() #must specify key or index
with self.assertRaises(ValueError):
tab.row(key='foobar',index=1) #cannot specify key and index
with self.assertRaises(KeyError):
tab.col(key='foobar') #invalid key
with self.assertRaises(ValueError):
tab.col(index=100000) #out of bounds
with self.assertRaises(ValueError):
tab.col() #must specify key or index
with self.assertRaises(ValueError):
tab.col(key='foobar',index=1) #cannot specify key and index
with self.assertRaises(ValueError):
tab.render(fmt="foobar") #invalid format
latexOnlyTab = pygsti.report.table.ReportTable(['latex'], ['Col1','Col2'], (None,None),
"tableClass",False)
with self.assertRaises(KeyError):
x = str(latexOnlyTab) #must have 'py' format
#LogL case tests
gen.get_logl_progress_table([0], [gateset_tp], [ [('Gx',)],], ds, formats, tableclass, longtable) # logL case
gen.get_logl_progress_table([0], [gateset], [ [('Gx',)],], ds, formats, tableclass, longtable) # logL case
gen.get_logl_confidence_region(gateset_tp, ds, 95,
gatestring_list=None, probClipInterval=(-1e6,1e6),
minProbClip=1e-4, radius=1e-4, hessianProjection="std")
gen.get_logl_confidence_region(gateset, ds, 95,
gatestring_list=None, probClipInterval=(-1e6,1e6),
minProbClip=1e-4, radius=1e-4, hessianProjection="std")
def test_hessian_projection(self):
chi2, chi2Grad, chi2Hessian = pygsti.chi2(self.gateset, self.ds,
returnGradient=True,
returnHessian=True)
proj_non_gauge = self.gateset.get_nongauge_projector()
projectedHessian = np.dot(proj_non_gauge,
np.dot(chi2Hessian, proj_non_gauge))
print(self.gateset.num_params())
print(proj_non_gauge.shape)
self.assertEqual( projectedHessian.shape, (60,60) )
#print("Evals = ")
#print("\n".join( [ "%d: %g" % (i,ev) for i,ev in enumerate(np.linalg.eigvals(projectedHessian))] ))
self.assertEqual( np.linalg.matrix_rank(proj_non_gauge), 44)
self.assertEqual( np.linalg.matrix_rank(projectedHessian), 44)
eigvals = np.sort(abs(np.linalg.eigvals(projectedHessian)))
print("eigvals = ",eigvals)
eigvals_chk = np.array(
[6.11142452e-11, 5.84760068e-10, 6.73066231e-10, 6.89070172e-10,
6.89070172e-10, 1.16844369e-09, 1.99649746e-09, 1.99649746e-09,
2.41697718e-09, 2.41697718e-09, 2.80832759e-09, 4.29148412e-09,
4.29148412e-09, 6.59655396e-09, 6.59655396e-09, 6.76915973e-09,
9.23087831e+05, 1.05783101e+06, 1.16457705e+06, 1.39492929e+06,
1.84015484e+06, 2.10613947e+06, 2.37963392e+06, 2.47192689e+06,
2.64566761e+06, 2.68722871e+06, 2.82383377e+06, 2.86584033e+06,
2.94590436e+06, 2.96180212e+06, 3.08322015e+06, 3.29389050e+06,
3.66581786e+06, 3.76266448e+06, 3.81921738e+06, 3.86624688e+06,
3.89045873e+06, 4.72831630e+06, 4.96416855e+06, 6.53286834e+06,
1.01424911e+07, 1.11347312e+07, 1.26152967e+07, 1.30081040e+07,
1.36647082e+07, 1.49293583e+07, 1.58234599e+07, 1.80999182e+07,
2.09155048e+07, 2.17444267e+07, 2.46870311e+07, 2.64427393e+07,
2.72410297e+07, 3.34988002e+07, 3.45005948e+07, 3.69040745e+07,
5.08647137e+07, 9.43153151e+07, 1.36088308e+08, 6.30304807e+08] )
#OLD
#[ 1.06310603e-10, 2.65635664e-10, 6.50301005e-10, 7.97927477e-10,
# 7.97927477e-10, 1.00960481e-09, 1.20668922e-09, 1.60395671e-09,
# 1.60395671e-09, 2.10414039e-09, 2.40422523e-09, 2.66328705e-09,
# 2.66328705e-09, 3.78157418e-09, 4.97929640e-09, 1.69073771e-08,
# 9.22542920e+05, 1.05940963e+06, 1.16240250e+06, 1.39506940e+06,
# 1.83925308e+06, 2.11037916e+06, 2.39385711e+06, 2.47432236e+06,
# 2.63561759e+06, 2.68157105e+06, 2.81568070e+06, 2.86569765e+06,
# 2.94543146e+06, 2.95039566e+06, 3.08684833e+06, 3.28869042e+06,
# 3.66558726e+06, 3.76232707e+06, 3.82389736e+06, 3.86638791e+06,
# 3.88874028e+06, 4.73808468e+06, 4.96550964e+06, 6.53177028e+06,
# 1.01544928e+07, 1.11525472e+07, 1.25572253e+07, 1.30411488e+07,
# 1.36881398e+07, 1.49309288e+07, 1.57790471e+07, 1.81263731e+07,
# 2.08276405e+07, 2.18675949e+07, 2.46968548e+07, 2.64099665e+07,
# 2.72117335e+07, 3.35172152e+07, 3.45138716e+07, 3.68918207e+07,
# 5.09742531e+07, 9.43260992e+07, 1.36044734e+08, 6.30355637e+08])
TOL = 1e-7
for val,chk in zip(eigvals,eigvals_chk):
if abs(val) > TOL or abs(chk) > TOL:
self.assertAlmostEqual(abs(val-chk)/(abs(chk)+TOL), 0.0, places=3)
def test_hessian_projection(self):
chi2, chi2Grad, chi2Hessian = pygsti.chi2(self.gateset,
self.ds,
returnGradient=True,
returnHessian=True)
proj_non_gauge = self.gateset.get_nongauge_projector()
projectedHessian = np.dot(proj_non_gauge,
np.dot(chi2Hessian, proj_non_gauge))
print(self.gateset.num_params())
print(proj_non_gauge.shape)
self.assertEqual(projectedHessian.shape, (60, 60))
#print("Evals = ")
#print("\n".join( [ "%d: %g" % (i,ev) for i,ev in enumerate(np.linalg.eigvals(projectedHessian))] ))
self.assertEqual(np.linalg.matrix_rank(proj_non_gauge), 44)
self.assertEqual(np.linalg.matrix_rank(projectedHessian), 44)
eigvals = np.sort(abs(np.linalg.eigvals(projectedHessian)))
print("eigvals = ", eigvals)
eigvals_chk = np.array([
1.06310603e-10, 2.65635664e-10, 6.50301005e-10, 7.97927477e-10,
7.97927477e-10, 1.00960481e-09, 1.20668922e-09, 1.60395671e-09,
1.60395671e-09, 2.10414039e-09, 2.40422523e-09, 2.66328705e-09,
2.66328705e-09, 3.78157418e-09, 4.97929640e-09, 1.69073771e-08,
9.22542920e+05, 1.05940963e+06, 1.16240250e+06, 1.39506940e+06,
1.83925308e+06, 2.11037916e+06, 2.39385711e+06, 2.47432236e+06,
2.63561759e+06, 2.68157105e+06, 2.81568070e+06, 2.86569765e+06,
2.94543146e+06, 2.95039566e+06, 3.08684833e+06, 3.28869042e+06,
3.66558726e+06, 3.76232707e+06, 3.82389736e+06, 3.86638791e+06,
3.88874028e+06, 4.73808468e+06, 4.96550964e+06, 6.53177028e+06,
1.01544928e+07, 1.11525472e+07, 1.25572253e+07, 1.30411488e+07,
1.36881398e+07, 1.49309288e+07, 1.57790471e+07, 1.81263731e+07,
2.08276405e+07, 2.18675949e+07, 2.46968548e+07, 2.64099665e+07,
2.72117335e+07, 3.35172152e+07, 3.45138716e+07, 3.68918207e+07,
5.09742531e+07, 9.43260992e+07, 1.36044734e+08, 6.30355637e+08
])
#OLD
#[ 2.53636344e-10, 3.87263955e-10, 4.49523968e-10,
# 8.17955744e-10, 8.17955744e-10, 1.22910388e-09,
# 1.23803907e-09, 1.67547571e-09, 1.67547571e-09,
# 1.75147770e-09, 2.20582127e-09, 2.20582127e-09,
# 2.84333714e-09, 4.43169431e-09, 4.43169431e-09,
# 1.75164250e-08, 9.38919779e+05, 9.57169426e+05,
# 9.69271265e+05, 1.41963844e+06, 1.52443387e+06,
# 1.89627852e+06, 1.97543819e+06, 2.05177386e+06,
# 2.18333142e+06, 2.30078215e+06, 2.31036461e+06,
# 2.40108194e+06, 2.63301339e+06, 2.72062783e+06,
# 2.73970548e+06, 2.90332118e+06, 3.15705184e+06,
# 3.86079309e+06, 3.87209620e+06, 4.70586582e+06,
# 8.76738379e+06, 9.73067464e+06, 1.04062266e+07,
# 1.15408214e+07, 1.21868610e+07, 1.33524791e+07,
# 1.34516720e+07, 1.50757108e+07, 1.74290255e+07,
# 1.83023511e+07, 2.15141700e+07, 2.22614418e+07,
# 2.32690752e+07, 2.88149432e+07, 3.04306844e+07,
# 3.10300863e+07, 4.25290585e+07, 8.95794195e+07,
# 1.29723323e+08, 5.69560469e+08])
TOL = 1e-7
for val, chk in zip(eigvals, eigvals_chk):
if abs(val) > TOL or abs(chk) > TOL:
self.assertAlmostEqual(abs(val - chk) / (abs(chk) + TOL),
0.0,
places=3)
def test_table_generation(self):
import pygsti.report.generation as gen
formats = ['latex', 'html', 'py', 'ppt'] #all the formats we know
tableclass = "cssClass"
longtable = False
confidenceRegionInfo = None
gateset = pygsti.io.load_gateset("cmp_chk_files/analysis.gateset")
ds = pygsti.objects.DataSet(
fileToLoadFrom="cmp_chk_files/analysis.dataset")
chi2, chi2Hessian = pygsti.chi2(ds, gateset, returnHessian=True)
ci = pygsti.obj.ConfidenceRegion(gateset,
chi2Hessian,
95.0,
hessianProjection="std")
gateset_tp = pygsti.contract(gateset, "TP")
gateset_tp.set_all_parameterizations("TP")
chi2, chi2Hessian_TP = pygsti.chi2(ds, gateset_tp, returnHessian=True)
ci_TP = pygsti.obj.ConfidenceRegion(gateset_tp,
chi2Hessian_TP,
95.0,
hessianProjection="std")
chi2, chi2Hessian_tgt = pygsti.chi2(ds,
std.gs_target,
returnHessian=True)
ci_tgt = pygsti.obj.ConfidenceRegion(std.gs_target,
chi2Hessian_tgt,
95.0,
hessianProjection="std")
target_tp = std.gs_target.copy()
target_tp.set_all_parameterizations("TP")
chi2, chi2Hessian_tgt_TP = pygsti.chi2(ds,
target_tp,
returnHessian=True)
ci_TP_tgt = pygsti.obj.ConfidenceRegion(target_tp,
chi2Hessian_tgt_TP,
95.0,
hessianProjection="std")
gateset_2q = pygsti.construction.build_gateset(
[4], [('Q0', 'Q1')], ['GIX', 'GIY', 'GXI', 'GYI', 'GCNOT'], [
"I(Q0):X(pi/2,Q1)", "I(Q0):Y(pi/2,Q1)", "X(pi/2,Q0):I(Q1)",
"Y(pi/2,Q0):I(Q1)", "CX(pi,Q0,Q1)"
],
prepLabels=['rho0'],
prepExpressions=["0"],
effectLabels=['E0', 'E1', 'E2'],
effectExpressions=["0", "1", "2"],
spamdefs={
'upup': ('rho0', 'E0'),
'updn': ('rho0', 'E1'),
'dnup': ('rho0', 'E2'),
'dndn': ('rho0', 'remainder')
},
basis="pp")
#tests which fill in the cracks of the full-report tests
tab = gen.get_gateset_spam_table(gateset, None)
tab_wCI = gen.get_gateset_spam_table(gateset, ci)
gen.get_gateset_spam_table(gateset, None)
gen.get_gateset_gates_table(gateset_tp, ci_TP) #test zero-padding
gen.get_unitary_gateset_gates_table(std.gs_target,
ci_tgt) #unitary gates w/CIs
gen.get_unitary_gateset_gates_table(target_tp,
ci_TP_tgt) #unitary gates w/CIs
gen.get_gateset_closest_unitary_table(gateset_2q,
None) #test higher-dim gateset
gen.get_gateset_closest_unitary_table(gateset,
ci) #test with CIs (long...)
gen.get_gateset_rotn_axis_table(
std.gs_target, None) #try to get "--"s and "X"s to display
gen.get_chi2_progress_table([0], [gateset_tp], [
[('Gx', )],
], ds) #TP case
gen.get_chi2_confidence_region(gateset_tp, ds, 95) #TP case
gen.get_gatestring_multi_table(
[[('Gx', ), ('Gz', )], [('Gy', )]], ["list1", "list2"],
commonTitle=None) #commonTitle == None case w/diff length lists
with self.assertRaises(ValueError):
gen.get_unitary_gateset_gates_table(std.gs_target,
ci) #gateset-CI mismatch
with self.assertRaises(ValueError):
gen.get_gateset_spam_parameters_table(std.gs_target,
ci) #gateset-CI mismatch
#Test ReportTable object
rowLabels = tab.keys()
row1Data = tab[rowLabels[0]]
colLabels = row1Data.keys()
self.assertTrue(rowLabels, tab.row_names)
self.assertTrue(colLabels, tab.col_names)
self.assertTrue(len(rowLabels), tab.num_rows)
self.assertTrue(len(colLabels), tab.num_cols)
el00 = tab[rowLabels[0]][colLabels[0]]
self.assertTrue(rowLabels[0] in tab)
self.assertTrue(tab.has_key(rowLabels[0]))
table_len = len(tab)
self.assertEqual(table_len, tab.num_rows)
table_as_str = str(tab)
table_wCI_as_str = str(tab_wCI)
row1a = tab.row(key=rowLabels[0])
col1a = tab.col(key=colLabels[0])
row1b = tab.row(index=0)
col1b = tab.col(index=0)
self.assertEqual(row1a, row1b)
self.assertEqual(col1a, col1b)
with self.assertRaises(KeyError):
tab['foobar']
with self.assertRaises(KeyError):
tab.row(key='foobar') #invalid key
with self.assertRaises(ValueError):
tab.row(index=100000) #out of bounds
with self.assertRaises(ValueError):
tab.row() #must specify key or index
with self.assertRaises(ValueError):
tab.row(key='foobar', index=1) #cannot specify key and index
with self.assertRaises(KeyError):
tab.col(key='foobar') #invalid key
with self.assertRaises(ValueError):
tab.col(index=100000) #out of bounds
with self.assertRaises(ValueError):
tab.col() #must specify key or index
with self.assertRaises(ValueError):
tab.col(key='foobar', index=1) #cannot specify key and index
with self.assertRaises(ValueError):
tab.render(fmt="foobar") #invalid format
#LogL case tests
gen.get_logl_progress_table([0], [gateset_tp], [
[('Gx', )],
], ds) # logL case
gen.get_logl_progress_table([0], [gateset], [
[('Gx', )],
], ds) # logL case
gen.get_logl_confidence_region(gateset_tp,
ds,
95,
gatestring_list=None,
probClipInterval=(-1e6, 1e6),
minProbClip=1e-4,
radius=1e-4,
hessianProjection="std")
gen.get_logl_confidence_region(gateset,
ds,
95,
gatestring_list=None,
probClipInterval=(-1e6, 1e6),
minProbClip=1e-4,
radius=1e-4,
hessianProjection="std")
def test_hessian_projection(self):
chi2, chi2Grad, chi2Hessian = pygsti.chi2(self.model,
self.ds,
returnGradient=True,
returnHessian=True)
proj_non_gauge = self.model.get_nongauge_projector()
projectedHessian = np.dot(proj_non_gauge,
np.dot(chi2Hessian, proj_non_gauge))
print(self.model.num_params())
print(proj_non_gauge.shape)
self.assertEqual(projectedHessian.shape, (60, 60))
#print("Evals = ")
#print("\n".join( [ "%d: %g" % (i,ev) for i,ev in enumerate(np.linalg.eigvals(projectedHessian))] ))
self.assertEqual(np.linalg.matrix_rank(proj_non_gauge), 44)
self.assertEqual(np.linalg.matrix_rank(projectedHessian), 44)
eigvals = np.sort(abs(np.linalg.eigvals(projectedHessian)))
print("eigvals = ", eigvals)
eigvals_chk = np.array([
1.22771144e-10, 5.09703519e-10, 7.77414421e-10, 7.77414421e-10,
9.54941949e-10, 9.54941949e-10, 1.31838597e-09, 1.68301591e-09,
1.68301591e-09, 1.90734690e-09, 1.90734690e-09, 2.49398698e-09,
2.63214649e-09, 5.55011910e-09, 8.25271609e-09, 1.79101190e-08,
9.23263305e+05, 1.05796436e+06, 1.16544156e+06, 1.39483644e+06,
1.83988792e+06, 2.10385396e+06, 2.37875879e+06, 2.47138866e+06,
2.64517713e+06, 2.68781067e+06, 2.82400957e+06, 2.86645483e+06,
2.94682035e+06, 2.96132874e+06, 3.08410148e+06, 3.29632639e+06,
3.66669038e+06, 3.76286040e+06, 3.81791306e+06, 3.86566785e+06,
3.88861934e+06, 4.72670353e+06, 4.96458456e+06, 6.53314879e+06,
1.01442686e+07, 1.11366689e+07, 1.26150705e+07, 1.30030291e+07,
1.36655922e+07, 1.49308924e+07, 1.58259068e+07, 1.81000865e+07,
2.09105990e+07, 2.17439110e+07, 2.46822193e+07, 2.64512967e+07,
2.72379830e+07, 3.34861817e+07, 3.44988014e+07, 3.68957880e+07,
5.08236925e+07, 9.43118724e+07, 1.36087264e+08, 6.30301385e+08
])
#OLD
#[6.11142452e-11, 5.84760068e-10, 6.73066231e-10, 6.89070172e-10,
# 6.89070172e-10, 1.16844369e-09, 1.99649746e-09, 1.99649746e-09,
# 2.41697718e-09, 2.41697718e-09, 2.80832759e-09, 4.29148412e-09,
# 4.29148412e-09, 6.59655396e-09, 6.59655396e-09, 6.76915973e-09,
# 9.23087831e+05, 1.05783101e+06, 1.16457705e+06, 1.39492929e+06,
# 1.84015484e+06, 2.10613947e+06, 2.37963392e+06, 2.47192689e+06,
# 2.64566761e+06, 2.68722871e+06, 2.82383377e+06, 2.86584033e+06,
# 2.94590436e+06, 2.96180212e+06, 3.08322015e+06, 3.29389050e+06,
# 3.66581786e+06, 3.76266448e+06, 3.81921738e+06, 3.86624688e+06,
# 3.89045873e+06, 4.72831630e+06, 4.96416855e+06, 6.53286834e+06,
# 1.01424911e+07, 1.11347312e+07, 1.26152967e+07, 1.30081040e+07,
# 1.36647082e+07, 1.49293583e+07, 1.58234599e+07, 1.80999182e+07,
# 2.09155048e+07, 2.17444267e+07, 2.46870311e+07, 2.64427393e+07,
# 2.72410297e+07, 3.34988002e+07, 3.45005948e+07, 3.69040745e+07,
# 5.08647137e+07, 9.43153151e+07, 1.36088308e+08, 6.30304807e+08] )
TOL = 1e-7
for val, chk in zip(eigvals, eigvals_chk):
if abs(val) > TOL or abs(chk) > TOL:
self.assertAlmostEqual(abs(val - chk) / (abs(chk) + TOL),
0.0,
places=3)
def test_hessian_projection(self):
chi2, chi2Grad, chi2Hessian = pygsti.chi2(self.ds, self.gateset, returnGradient=True, returnHessian=True)
proj_non_gauge = self.gateset.get_nongauge_projector()
projectedHessian = np.dot(proj_non_gauge, np.dot(chi2Hessian, proj_non_gauge))
self.assertEqual(projectedHessian.shape, (56, 56))
self.assertEqual(np.linalg.matrix_rank(proj_non_gauge), 40)
self.assertEqual(np.linalg.matrix_rank(projectedHessian), 40)
eigvals = np.sort(abs(np.linalg.eigvals(projectedHessian)))
eigvals_chk = np.array(
[
2.36932621e-11,
4.57739349e-11,
8.49888013e-11,
8.49888013e-11,
1.22859895e-10,
1.22859895e-10,
1.38705957e-10,
1.38705957e-10,
3.75441328e-10,
6.46644807e-10,
6.46644807e-10,
7.06181642e-10,
7.06181642e-10,
7.65472749e-10,
1.62672899e-09,
1.62672899e-09,
9.23280752e04,
9.88622140e04,
1.24577730e05,
1.40500652e05,
1.88461602e05,
1.98466608e05,
2.07084497e05,
2.27489018e05,
2.33403285e05,
2.52573024e05,
2.82350782e05,
3.12087185e05,
3.21855420e05,
3.31659734e05,
3.52649174e05,
3.60682071e05,
3.90777833e05,
4.59913853e05,
5.02652879e05,
5.59311926e05,
5.78891070e05,
6.82325323e05,
7.57318263e05,
8.16739390e05,
1.06466062e06,
1.20075694e06,
1.37368639e06,
1.58356629e06,
1.68898356e06,
2.12277359e06,
3.30650801e06,
3.75869331e06,
4.00195245e06,
4.42427797e06,
5.06956256e06,
7.31166332e06,
9.19432790e06,
9.99944236e06,
1.31027722e07,
5.80310818e07,
]
)
TOL = 1e-7
for val, chk in zip(eigvals, eigvals_chk):
if abs(val) > TOL or abs(chk) > TOL:
self.assertAlmostEqual(abs(val - chk) / (abs(chk) + TOL), 0.0, places=3)
def test_model_selection(self):
ds = self.ds
#pygsti.construction.generate_fake_data(self.datagen_gateset, self.lsgstStrings[-1],
# nSamples=1000,sampleError='binomial', seed=100)
gs_lgst4 = pygsti.do_lgst(ds,
self.specs,
self.gateset,
svdTruncateTo=4,
verbosity=0)
gs_lgst6 = pygsti.do_lgst(ds,
self.specs,
self.gateset,
svdTruncateTo=6,
verbosity=0)
sys.stdout.flush()
self.runSilent(pygsti.do_lgst,
ds,
self.specs,
self.gateset,
svdTruncateTo=6,
verbosity=4) # test verbose prints
chiSq4 = pygsti.chi2(ds,
gs_lgst4,
self.lgstStrings,
minProbClipForWeighting=1e-4)
chiSq6 = pygsti.chi2(ds,
gs_lgst6,
self.lgstStrings,
minProbClipForWeighting=1e-4)
print("LGST dim=4 chiSq = ", chiSq4)
print("LGST dim=6 chiSq = ", chiSq6)
#self.assertAlmostEqual(chiSq4, 174.061524953) #429.271983052)
#self.assertAlmostEqual(chiSq6, 267012993.861, places=1) #1337.74222467) #Why is this so large??? -- DEBUG later
# Least squares GST with model selection
gs_lsgst = self.runSilent(
pygsti.do_iterative_mc2gst_with_model_selection,
ds,
gs_lgst4,
1,
self.lsgstStrings[0:3],
verbosity=10,
minProbClipForWeighting=1e-3,
probClipInterval=(-1e5, 1e5))
# Run again with other parameters
tuple_strings = [
list(map(tuple, gsList)) for gsList in self.lsgstStrings[0:3]
] #to test tuple argument
errorVecs, gs_lsgst_wts = self.runSilent(
pygsti.do_iterative_mc2gst_with_model_selection,
ds,
gs_lgst4,
1,
tuple_strings,
verbosity=10,
minProbClipForWeighting=1e-3,
probClipInterval=(-1e5, 1e5),
gatestringWeightsDict={('Gx', ): 2.0},
returnAll=True,
returnErrorVec=True)
# Do non-iterative to cover GateString->tuple conversion
gs_non_iterative = self.runSilent(
pygsti.do_mc2gst_with_model_selection,
ds,
gs_lgst4,
1,
self.lsgstStrings[0],
verbosity=10,
probClipInterval=(-1e5, 1e5))
# RUN BELOW LINES TO SEED SAVED GATESET FILES
#pygsti.io.write_gateset(gs_lsgst,compare_files + "/lsgstMS.gateset", "Saved LSGST Gateset with model selection")
gs_lsgst_compare = pygsti.io.load_gateset(compare_files +
"/lsgstMS.gateset")
gs_lsgst_go = pygsti.optimize_gauge(gs_lsgst,
'target',
targetGateset=gs_lsgst_compare,
spamWeight=1.0)
self.assertAlmostEqual(gs_lsgst_go.frobeniusdist(gs_lsgst_compare),
0,
places=5)
def test_confidenceRegion(self):
chi2, chi2Hessian = pygsti.chi2(self.ds, self.gateset,
returnHessian=True)
ci_std = pygsti.obj.ConfidenceRegion(self.gateset, chi2Hessian, 95.0,
hessianProjection="std")
ci_noproj = pygsti.obj.ConfidenceRegion(self.gateset, chi2Hessian, 95.0,
hessianProjection="none")
ci_opt = pygsti.obj.ConfidenceRegion(self.gateset, chi2Hessian, 95.0,
hessianProjection="optimal gate CIs",
tol=0.1) #very low tol so doesn't take long
with self.assertRaises(ValueError):
pygsti.obj.ConfidenceRegion(self.gateset, chi2Hessian, 95.0,
hessianProjection="FooBar") #bad hessianProjection
self.assertWarns(pygsti.obj.ConfidenceRegion, self.gateset,
chi2Hessian, 0.95, hessianProjection="none") # percentage < 1.0
ar_of_intervals_Gx = ci_std.get_profile_likelihood_confidence_intervals("Gx")
ar_of_intervals_rho0 = ci_std.get_profile_likelihood_confidence_intervals("rho0")
ar_of_intervals_E0 = ci_std.get_profile_likelihood_confidence_intervals("E0")
ar_of_intervals = ci_std.get_profile_likelihood_confidence_intervals()
def fnOfGate_float(mx):
return float(mx[0,0])
def fnOfGate_0D(mx):
return np.array( float(mx[0,0]) )
def fnOfGate_1D(mx):
return mx[0,:]
def fnOfGate_2D(mx):
return mx[:,:]
def fnOfGate_3D(mx):
return np.zeros( (2,2,2), 'd') #just to test for error
df = ci_std.get_gate_fn_confidence_interval(fnOfGate_float, 'Gx', verbosity=0)
df = ci_std.get_gate_fn_confidence_interval(fnOfGate_0D, 'Gx', verbosity=0)
df = ci_std.get_gate_fn_confidence_interval(fnOfGate_1D, 'Gx', verbosity=0)
df = ci_std.get_gate_fn_confidence_interval(fnOfGate_2D, 'Gx', verbosity=0)
df, f0 = self.runSilent(ci_std.get_gate_fn_confidence_interval,
fnOfGate_float, 'Gx', returnFnVal=True, verbosity=4)
with self.assertRaises(ValueError):
ci_std.get_gate_fn_confidence_interval(fnOfGate_3D, 'Gx', verbosity=0)
def fnOfSpam_float(rhoVecs, EVecs):
return float( np.dot( rhoVecs[0].T, EVecs[0] ) )
def fnOfSpam_0D(rhoVecs, EVecs):
return np.array( float( np.dot( rhoVecs[0].T, EVecs[0] ) ) )
def fnOfSpam_1D(rhoVecs, EVecs):
return np.array( [ np.dot( rhoVecs[0].T, EVecs[0] ), 0] )
def fnOfSpam_2D(rhoVecs, EVecs):
return np.array( [[ np.dot( rhoVecs[0].T, EVecs[0] ), 0],[0,0]] )
def fnOfSpam_3D(rhoVecs, EVecs):
return np.zeros( (2,2,2), 'd') #just to test for error
df = ci_std.get_spam_fn_confidence_interval(fnOfSpam_float, verbosity=0)
df = ci_std.get_spam_fn_confidence_interval(fnOfSpam_0D, verbosity=0)
df = ci_std.get_spam_fn_confidence_interval(fnOfSpam_1D, verbosity=0)
df = ci_std.get_spam_fn_confidence_interval(fnOfSpam_2D, verbosity=0)
df, f0 = self.runSilent(ci_std.get_spam_fn_confidence_interval,
fnOfSpam_float, returnFnVal=True, verbosity=4)
with self.assertRaises(ValueError):
ci_std.get_spam_fn_confidence_interval(fnOfSpam_3D, verbosity=0)
def test_total_chi2(self):
chi2 = pygsti.chi2( self.ds, self.lsgst_gateset )
self.assertAlmostEqual(chi2, 729.182015795)
