def test_group():
conv_ensemble_sparse = Ensemble.fromfilelist(map_list)
conv_ensemble_sparse.load(callback_loader=default_callback_loader,pool=pool,l_edges=l_edges)
conv_ensemble_sparse.group(group_size=2,kind="sparse")
assert conv_ensemble_sparse.num_realizations==2
conv_ensemble_contiguous = Ensemble.fromfilelist(map_list)
conv_ensemble_contiguous.load(callback_loader=default_callback_loader,pool=pool,l_edges=l_edges)
conv_ensemble_contiguous.group(group_size=2,kind="contiguous")
assert conv_ensemble_contiguous.num_realizations==2
fig,ax = plt.subplots()
for n in range(conv_ensemble.num_realizations):
ax.plot(l,l*(l+1)*conv_ensemble.data[n]/(2.0*np.pi),label="Original {0}".format(n+1),linestyle="-")
for n in range(conv_ensemble_sparse.num_realizations):
ax.plot(l,l*(l+1)*conv_ensemble_sparse.data[n]/(2.0*np.pi),label="Sparse {0}".format(n+1),linestyle="--")
for n in range(conv_ensemble_contiguous.num_realizations):
ax.plot(l,l*(l+1)*conv_ensemble_contiguous.data[n]/(2.0*np.pi),label="Contiguous {0}".format(n+1),linestyle="-.")
ax.set_xscale("log")
ax.set_yscale("log")
ax.set_xlabel(r"$l$")
ax.set_ylabel(r"$l(l+1)P_l/2\pi$")
ax.legend(loc="upper left",prop={"size":7})
plt.savefig("power_ensemble_grouped.png")
plt.clf()
return conv_ensemble_sparse._scheme,conv_ensemble_contiguous._scheme
def test_add(): conv_ensemble1 = Ensemble.fromfilelist(map_list[0:2]) conv_ensemble2 = Ensemble.fromfilelist(map_list[2:]) conv_ensemble1.load(callback_loader=default_callback_loader,pool=None,l_edges=l_edges) conv_ensemble2.load(callback_loader=default_callback_loader,pool=None,l_edges=l_edges) conv_ensemble_union = conv_ensemble1 + conv_ensemble2 assert conv_ensemble_union.num_realizations == 4 assert len(conv_ensemble_union.file_list) == 4 assert conv_ensemble_union.data.shape[0] == 4 assert conv_ensemble_union.data.shape[1] == conv_ensemble1.data.shape[1]
def test_selfChi2():
ens = Ensemble.read(
"Data/all/Om0.295_Ol0.705_w-1.878_ns0.960_si0.100/subfield1/sigma05/power_spectrum.npy"
)
chi2 = ens.selfChi2()
assert chi2.shape[0] == ens.data.shape[0]
#Plot histogram
fig, ax = plt.subplots()
n, bins, patch = ax.hist(chi2,
bins=50,
normed=True,
histtype="stepfilled",
alpha=0.5)
#Compare to chi2 distribution
ax.plot(stats.chi2.pdf(bins, ens.data.shape[1]))
#Labels
ax.set_xlabel(r"$\chi^2$")
ax.set_ylabel(r"$P(\chi^2)$")
#Save figure
fig.savefig("self_chi2.png")
def test_group():
conv_ensemble_sparse = Ensemble.fromfilelist(map_list)
conv_ensemble_sparse.load(callback_loader=default_callback_loader,
pool=pool,
l_edges=l_edges)
conv_ensemble_sparse.group(group_size=2, kind="sparse")
assert conv_ensemble_sparse.num_realizations == 2
conv_ensemble_contiguous = Ensemble.fromfilelist(map_list)
conv_ensemble_contiguous.load(callback_loader=default_callback_loader,
pool=pool,
l_edges=l_edges)
conv_ensemble_contiguous.group(group_size=2, kind="contiguous")
assert conv_ensemble_contiguous.num_realizations == 2
fig, ax = plt.subplots()
for n in range(conv_ensemble.num_realizations):
ax.plot(l,
l * (l + 1) * conv_ensemble.data[n] / (2.0 * np.pi),
label="Original {0}".format(n + 1),
linestyle="-")
for n in range(conv_ensemble_sparse.num_realizations):
ax.plot(l,
l * (l + 1) * conv_ensemble_sparse.data[n] / (2.0 * np.pi),
label="Sparse {0}".format(n + 1),
linestyle="--")
for n in range(conv_ensemble_contiguous.num_realizations):
ax.plot(l,
l * (l + 1) * conv_ensemble_contiguous.data[n] / (2.0 * np.pi),
label="Contiguous {0}".format(n + 1),
linestyle="-.")
ax.set_xscale("log")
ax.set_yscale("log")
ax.set_xlabel(r"$l$")
ax.set_ylabel(r"$l(l+1)P_l/2\pi$")
ax.legend(loc="upper left", prop={"size": 7})
plt.savefig("power_ensemble_grouped.png")
plt.clf()
return conv_ensemble_sparse._scheme, conv_ensemble_contiguous._scheme
def test_index():
#Decide the statistical descriptors to measure, and build an index
idx = Indexer.stack([
PowerSpectrum(l_edges),
Peaks(thresholds_pk, norm=True),
MinkowskiAll(thresholds_mf, norm=True),
PDF(thresholds_mf, norm=True),
Moments(connected=True)
])
l = idx[0].l
v = idx[1].midpoints
v_mf = idx[2].midpoints
#Initiate the statistical ensemble
ens = Ensemble.fromfilelist(map_list)
#Load measurements into the ensemble (this is the expensive part!!!)
ens.load(callback_loader=convergence_measure_all, pool=None, index=idx)
#Split the ensemble in power_spectrum,peaks, and the second and third minkowski functional
mink_idx = idx[2].separate()
subset_idx = Indexer([idx[0], idx[1], idx[3], mink_idx[2], idx[-1]])
ens_pow, ens_pk, ens_pdf, ens_mink2, ens_mom = ens.split(subset_idx)
#####################################################################
#Plot to check
fig, ax = plt.subplots(2, 2, figsize=(16, 16))
for i in range(ens.num_realizations):
ax[0, 0].plot(l, l * (l + 1) * ens_pow.data[i] / (2.0 * np.pi))
ax[0, 1].plot(v, ens_pk.data[i])
ax[1, 0].plot(v_mf, ens_pdf.data[i])
ax[1, 1].plot(v_mf, ens_mink2.data[i])
ax[0, 0].set_xscale("log")
ax[0, 0].set_yscale("log")
ax[0, 0].set_xlabel(r"$l$")
ax[0, 0].set_ylabel(r"$l(l+1)P_l/2\pi$")
ax[0, 1].set_xlabel(r"$\nu$")
ax[0, 1].set_ylabel(r"$dN/d\nu$")
ax[1, 0].set_xlabel(r"$\nu$")
ax[1, 0].set_ylabel(r"$P(\nu)$")
ax[1, 1].set_xlabel(r"$\nu$")
ax[1, 1].set_ylabel(r"$V_2(\nu)$")
fig.tight_layout()
plt.savefig("conv_all.png")
plt.clf()
#Save moments to check
np.savetxt("moments.txt", ens_mom.mean())
def test_save_and_load():
conv_ensemble.save("ensemble_saved.npy")
conv_ensemble.save("ensemble_saved", format="matlab", appendmat=True)
conv_ensemble_new = Ensemble.read("ensemble_saved.npy")
assert conv_ensemble_new.num_realizations == conv_ensemble.num_realizations
assert conv_ensemble_new.data.shape == conv_ensemble.data.shape
def test_save_and_load():
conv_ensemble.save("ensemble_saved.npy")
conv_ensemble.save("ensemble_saved",format="matlab",appendmat=True)
conv_ensemble_new = Ensemble.read("ensemble_saved.npy")
assert conv_ensemble_new.num_realizations == conv_ensemble.num_realizations
assert conv_ensemble_new.data.shape == conv_ensemble.data.shape
def test_add():
conv_ensemble1 = Ensemble.fromfilelist(map_list[0:2])
conv_ensemble2 = Ensemble.fromfilelist(map_list[2:])
conv_ensemble1.load(callback_loader=default_callback_loader,
pool=None,
l_edges=l_edges)
conv_ensemble2.load(callback_loader=default_callback_loader,
pool=None,
l_edges=l_edges)
conv_ensemble_union = conv_ensemble1 + conv_ensemble2
assert conv_ensemble_union.num_realizations == 4
assert len(conv_ensemble_union.file_list) == 4
assert conv_ensemble_union.data.shape[0] == 4
assert conv_ensemble_union.data.shape[1] == conv_ensemble1.data.shape[1]
def test_chi2():
conv_ensemble1 = Ensemble.fromfilelist(map_list[0:2])
conv_ensemble1.load(callback_loader=default_callback_loader,
pool=None,
l_edges=l_edges)
print("chi2 difference = {0}".format(
conv_ensemble.compare(conv_ensemble1)))
def test_multiply(): conv_ensemble_peaks = Ensemble.fromfilelist(map_list) conv_ensemble_peaks.load(callback_loader=peaks_loader,pool=None,thresholds=thresholds_pk) conv_ensemble_both = conv_ensemble * conv_ensemble_peaks assert conv_ensemble_both.num_realizations == 4 assert conv_ensemble_both.data.shape[0] == 4 assert conv_ensemble_both.data.shape[1] == len(l_edges) + len(thresholds_pk) - 2
def confusionMatrix(descriptor,measurement_list,measurement_covariance):
#Instantiate a FisherAnalysis instance
analysis = FisherAnalysis()
#Populate with the models
for measurement in measurement_list:
ens = Ensemble.read(measurement.savename(descriptor))
analysis.add_model(measurement.model.squeeze(with_ns=True),ens.mean())
#Compute the covariance matrix
covariance = Ensemble.read(measurement_covariance.savename(descriptor)).covariance()
#################################################################################
#####Now we are ready to compute the confusion matrix, for each parameter########
#################################################################################
#Allocate space for confusion matrix
confusion_matrix = np.zeros((4,3,3))
#Find where are the variations for each model parameter
locations = analysis.where()
#Cycle over parameters
for n in range(4):
l0 = analysis._fiducial
l1,l2 = locations[n]
print("[+] n={0}, fiducial: {1}, variation 1: {2}, variation 2:{3}".format(n,analysis.parameter_set[l0],analysis.parameter_set[l1],analysis.parameter_set[l2]))
for i,m in enumerate([l0,l1,l2]):
#Load the ensemble
ens = Ensemble.read(measurement_list[m].savename(descriptor))
#Compute the confusion matrix
confusion_matrix[n,i] = analysis.classify(ens.data,covariance,labels=[l0,l1,l2],confusion=True)
#Return the confusion matrix for the selected descriptor
return confusion_matrix
def test_index():
#Decide the statistical descriptors to measure, and build an index
idx = Indexer.stack([PowerSpectrum(l_edges),Peaks(thresholds_pk,norm=True),MinkowskiAll(thresholds_mf,norm=True),PDF(thresholds_mf,norm=True),Moments(connected=True)])
l = idx[0].l
v = idx[1].midpoints
v_mf = idx[2].midpoints
#Initiate the statistical ensemble
ens = Ensemble.fromfilelist(map_list)
#Load measurements into the ensemble (this is the expensive part!!!)
ens.load(callback_loader=convergence_measure_all,pool=None,index=idx)
#Split the ensemble in power_spectrum,peaks, and the second and third minkowski functional
mink_idx = idx[2].separate()
subset_idx = Indexer([idx[0],idx[1],idx[3],mink_idx[2],idx[-1]])
ens_pow,ens_pk,ens_pdf,ens_mink2,ens_mom = ens.split(subset_idx)
#####################################################################
#Plot to check
fig,ax = plt.subplots(2,2,figsize=(16,16))
for i in range(ens.num_realizations):
ax[0,0].plot(l,l*(l+1)*ens_pow.data[i]/(2.0*np.pi))
ax[0,1].plot(v,ens_pk.data[i])
ax[1,0].plot(v_mf,ens_pdf.data[i])
ax[1,1].plot(v_mf,ens_mink2.data[i])
ax[0,0].set_xscale("log")
ax[0,0].set_yscale("log")
ax[0,0].set_xlabel(r"$l$")
ax[0,0].set_ylabel(r"$l(l+1)P_l/2\pi$")
ax[0,1].set_xlabel(r"$\nu$")
ax[0,1].set_ylabel(r"$dN/d\nu$")
ax[1,0].set_xlabel(r"$\nu$")
ax[1,0].set_ylabel(r"$P(\nu)$")
ax[1,1].set_xlabel(r"$\nu$")
ax[1,1].set_ylabel(r"$V_2(\nu)$")
fig.tight_layout()
plt.savefig("conv_all.png")
plt.clf()
#Save moments to check
np.savetxt("moments.txt",ens_mom.mean())
def test_multiply():
conv_ensemble_peaks = Ensemble.fromfilelist(map_list)
conv_ensemble_peaks.load(callback_loader=peaks_loader,
pool=None,
thresholds=thresholds_pk)
conv_ensemble_both = conv_ensemble * conv_ensemble_peaks
assert conv_ensemble_both.num_realizations == 4
assert conv_ensemble_both.data.shape[0] == 4
assert conv_ensemble_both.data.shape[1] == len(l_edges) + len(
thresholds_pk) - 2
def test_pca():
pca_ensemble = Ensemble.read("Data/ensemble_pca.npy")
pca = pca_ensemble.principalComponents()
assert len(pca.explained_variance_) == pca_ensemble.data.shape[1]
fig, ax = plt.subplots(1, 2, figsize=(16, 8))
ax[0].plot(pca.explained_variance_)
ax[1].plot(pca.explained_variance_.cumsum())
ax[0].set_xlabel(r"$n$")
ax[1].set_xlabel(r"$n$")
ax[0].set_ylabel(r"$\lambda_n$")
ax[1].set_ylabel(r"$\sum^n\lambda_n$")
fig.savefig("pca.png")
def test_pca():
pca_ensemble = Ensemble.read("Data/ensemble_pca.npy")
pca = pca_ensemble.principalComponents()
assert len(pca.explained_variance_)==pca_ensemble.data.shape[1]
fig,ax = plt.subplots(1,2,figsize=(16,8))
ax[0].plot(pca.explained_variance_)
ax[1].plot(pca.explained_variance_.cumsum())
ax[0].set_xlabel(r"$n$")
ax[1].set_xlabel(r"$n$")
ax[0].set_ylabel(r"$\lambda_n$")
ax[1].set_ylabel(r"$\sum^n\lambda_n$")
fig.savefig("pca.png")
def measure_hough(self,pool=None,threshold=0.1,bins=np.linspace(0.0,0.0014,50),save_type="npy"):
assert self.measurer==igs1_measure_hough
realizations = range(1,self.nrealizations+1)
#Build the ensemble
ens = Ensemble.fromfilelist(realizations)
#Load the data into the ensemble by calling the measurer on each map
ens.load(callback_loader=self.measurer,pool=pool,model=self.model,redshift=self.redshift,big_fiducial_set=self.big_fiducial_set,threshold=threshold,bins=bins)
#Save
savename = self.savename(descriptor="hough")
logging.info("Saving hough histograms to {0}".format(savename))
ens.save(savename)
def measure_all_histograms(models,options,pool):
#Look at a sample map
sample_map = ConvergenceMap.fromfilename(models[0].getNames(z=1.0,realizations=[1])[0],loader=load_fits_default_convergence)
#Initialize Gaussian shape noise generator for the sample map shape and angle
generator = GaussianNoiseGenerator.forMap(sample_map)
#Parsed from options
num_realizations = options.getint("analysis","num_realizations")
smoothing_scales = [float(scale) for scale in options.get("analysis","smoothing_scales").split(",")]
bin_edges = np.ogrid[options.getfloat("analysis","bin_edge_low"):options.getfloat("analysis","bin_edge_high"):(options.getint("analysis","num_bins") - 2)*1j]
bin_edges = np.hstack((-10.0,bin_edges,10.0))
z = options.getfloat("analysis","redshift")
bin_midpoints = 0.5*(bin_edges[1:] + bin_edges[:-1])
#Create smoothing scale index for the histograms
idx = Indexer.stack([PDF(bin_edges) for scale in smoothing_scales])
#Build the data type of the structure array in output
data_type = [(model.name,Ensemble) for model in models]
#Append info about the smoothing scale
data_type = [("Smooth",np.float),] + data_type
#Create output struct array
ensemble_array = np.zeros(len(smoothing_scales),dtype=data_type)
#Write smoothing scale information
ensemble_array["Smooth"] = np.array(smoothing_scales)
#The for loop runs the distributed computations
for model in models:
#Build Ensemble instance with the maps to analyze
map_ensemble = Ensemble.fromfilelist(range(1,num_realizations+1))
#Measure the histograms and load the data in the ensemble
map_ensemble.load(callback_loader=compute_map_histograms,pool=pool,simulation_set=model,smoothing_scales=smoothing_scales,index=idx,generator=generator,bin_edges=bin_edges,redshift=z)
#Split the ensemble between different smoothing scales
map_ensemble_list = map_ensemble.split(idx)
#Add to output struct array
ensemble_array[model.name] = np.array(map_ensemble_list)
return ensemble_array
def measure(self,pool=None): """ Measures the features specified in the Indexer for all the maps whose names are calculated by get_all_map_names; saves the ensemble results in numpy array format """ #Build the ensemble ens = Ensemble.fromfilelist(self.map_names) #Load the data into the ensemble by calling the measurer on each map ens.load(callback_loader=self.measurer,pool=pool,**self.kwargs) #Break the ensemble into sub-ensemble, one for each feature single_feature_ensembles = ens.split(self.kwargs["index"]) #For each of the sub_ensembles, save it in the appropriate directory for n,ensemble in enumerate(single_feature_ensembles): ensemble.save(os.path.join(self.full_save_path,self.kwargs["index"][n].name) + ".npy")
def test_selfChi2():
ens = Ensemble.read("Data/all/Om0.295_Ol0.705_w-1.878_ns0.960_si0.100/subfield1/sigma05/power_spectrum.npy")
chi2 = ens.selfChi2()
assert chi2.shape[0]==ens.data.shape[0]
#Plot histogram
fig,ax = plt.subplots()
n,bins,patch = ax.hist(chi2,bins=50,normed=True,histtype="stepfilled",alpha=0.5)
#Compare to chi2 distribution
ax.plot(stats.chi2.pdf(bins,ens.data.shape[1]))
#Labels
ax.set_xlabel(r"$\chi^2$")
ax.set_ylabel(r"$P(\chi^2)$")
#Save figure
fig.savefig("self_chi2.png")
def test_differentiate():
thresholds = np.arange(-0.04, 0.12, 0.001)
midpoints = 0.5 * (thresholds[:-1] + thresholds[1:])
index = Indexer.stack([MinkowskiAll(thresholds)])
index_separate = Indexer(MinkowskiAll(thresholds).separate())
diff_ensemble = Ensemble.fromfilelist(map_list)
diff_ensemble.load(callback_loader=convergence_measure_all, index=index)
ensemble_0 = diff_ensemble.split(index_separate)[0]
ensemble_pdf = ensemble_0.differentiate(step=thresholds[0] - thresholds[1])
fig, ax = plt.subplots()
for i in range(ensemble_0.num_realizations):
ax.plot(0.5 * (midpoints[:-1] + midpoints[1:]), ensemble_pdf[i])
ax.set_xlabel(r"$\kappa$")
ax.set_ylabel(r"$P(\kappa)$")
fig.savefig("ensemble_differentiate.png")
def test_differentiate():
thresholds = np.arange(-0.04,0.12,0.001)
midpoints = 0.5*(thresholds[:-1] + thresholds[1:])
index = Indexer.stack([MinkowskiAll(thresholds)])
index_separate = Indexer(MinkowskiAll(thresholds).separate())
diff_ensemble = Ensemble.fromfilelist(map_list)
diff_ensemble.load(callback_loader=convergence_measure_all,index=index)
ensemble_0 = diff_ensemble.split(index_separate)[0]
ensemble_pdf = ensemble_0.differentiate(step=thresholds[0]-thresholds[1])
fig,ax = plt.subplots()
for i in range(ensemble_0.num_realizations):
ax.plot(0.5*(midpoints[:-1]+midpoints[1:]),ensemble_pdf[i])
ax.set_xlabel(r"$\kappa$")
ax.set_ylabel(r"$P(\kappa)$")
fig.savefig("ensemble_differentiate.png")
def measure(self,pool=None,save_type="npy"):
"""
Measures the features specified in the Indexer for all the maps whose names are calculated by get_all_map_names; saves the ensemble results in numpy array format
"""
realizations = range(1,self.nrealizations+1)
#Build the ensemble
ens = Ensemble.fromfilelist(realizations)
#Load the data into the ensemble by calling the measurer on each map
ens.load(callback_loader=self.measurer,pool=pool,model=self.model,index=self.index,mask_filename=None,redshift=self.redshift,big_fiducial_set=self.big_fiducial_set)
#Break the ensemble into sub-ensemble, one for each feature
single_feature_ensembles = ens.split(self.index)
#For each of the sub_ensembles, save it in the appropriate directory
for n,ensemble in enumerate(single_feature_ensembles):
savename = self.savename(descriptor=self.index[n])
logging.debug("Saving features to {0}".format(savename))
ensemble.save(savename)
map_mock_ids = range(int(sys.argv[1]))
igs1_set = IGS1(
root_path=
"/Users/andreapetri/Documents/Columbia/spurious_shear/convergence_maps")
map_igs1_ids = igs1_set.getNames(z=1.0,
realizations=range(1,
int(sys.argv[1]) + 1))
gen = GaussianNoiseGenerator(shape=(2048, 2048),
side_angle=3.41 * deg,
label="convergence")
power_func = np.loadtxt("Data/ee4e-7.txt", unpack=True)
ens_mock = Ensemble.fromfilelist(map_mock_ids)
ens_igs1 = Ensemble.fromfilelist(map_igs1_ids)
ens_mock.load(callback_loader=generate_and_measure,
pool=pool,
generator=gen,
power_func=power_func)
ens_igs1.load(callback_loader=measure_from_IGS1, pool=pool)
if pool is not None:
pool.close()
np.savetxt(
"moments_mock.txt",
np.array([ens_mock.mean(),
np.sqrt(ens_mock.covariance().diagonal())]))
def test_interpolation():
root_path = "Data/all"
analysis = LikelihoodAnalysis()
#Read in model names
models = CFHTemu1.getModels()[:17]
assert len(models) == 17
#Shuffle the models
np.random.seed(1)
np.random.shuffle(models)
#Divide into training and testing
training_models = models[:-1]
testing_model = models[-1]
#Read multipoles
ell = np.load(os.path.join(root_path, "ell.npy"))
#Load in the means of the power spectra of the 17 models, and populate the analysis instance
for model in training_models:
ens = Ensemble.fromfilelist([
os.path.join(root_path, model._cosmo_id_string, "subfield1",
"sigma05", "power_spectrum.npy")
])
ens.load(from_old=True)
analysis.add_model(parameters=model.squeeze(with_ns=True),
feature=ens.mean())
#Add the multipoles to the analysis
analysis.add_feature_label(ell)
l = analysis.feature_label
ens = Ensemble.fromfilelist([
os.path.join(root_path, testing_model._cosmo_id_string, "subfield1",
"sigma05", "power_spectrum.npy")
])
ens.load(from_old=True)
testing_Pl = ens.mean()
#Load in also the observed power spectrum
ens = Ensemble.fromfilelist([
os.path.join(root_path, "observations", "subfield1", "sigma05",
"power_spectrum.npy")
])
ens.load(from_old=True)
observed_Pl = ens.mean()
#Output the analysis stats
np.savetxt("16_parameter_points.txt", analysis.parameter_set)
for n in range(len(training_models)):
plt.plot(l, l * (l + 1) * analysis.training_set[n] / (2 * np.pi))
plt.plot(l,
l * (l + 1) * observed_Pl / (2 * np.pi),
linestyle="--",
label="Observation")
plt.xlabel(r"$l$")
plt.ylabel(r"$l(l+1)P_l/2\pi$")
plt.yscale("log")
plt.legend(loc="upper left")
plt.savefig("16_power_spectra.png")
plt.clf()
#Train the interpolators
analysis.train(use_parameters=range(3))
assert hasattr(analysis, "_interpolator")
assert hasattr(analysis, "_num_bins")
#Emulator portability test with pickle/unpickle
analysis.save("analysis.p")
emulator = LikelihoodAnalysis.load("analysis.p")
#Predict the power spectrum at the remaining point
predicted_Pl = emulator.predict(testing_model.squeeze())
#Plot it against the measured one
fig, ax = plt.subplots(2, 1, figsize=(16, 8))
#Measured
ax[0].plot(l, l * (l + 1) * testing_Pl / (2 * np.pi), label="measured")
#Predicted
ax[0].plot(l,
l * (l + 1) * predicted_Pl / (2 * np.pi),
label="interpolated")
#Fractional difference
ax[1].plot(l, (predicted_Pl - testing_Pl) / testing_Pl)
ax[1].set_xlabel(r"$l$")
ax[0].set_ylabel(r"$l(l+1)P_l/2\pi$")
ax[1].set_ylabel(r"$P_l^I-P_l^M/P_l^M$")
ax[0].set_yscale("log")
ax[0].legend(loc="upper left")
plt.savefig("power_interpolator_test.png")
plt.clf()
#Give it a shot with two points in parameter space to test vectorization
two_parameter_points = np.array(
(training_models[0].squeeze(), testing_model.squeeze()))
two_predicted_Pl = emulator.predict(two_parameter_points)
fig, ax = plt.subplots(2, 1, figsize=(16, 8))
#Predicted
ax[0].plot(l,
l * (l + 1) * two_predicted_Pl[0] / (2 * np.pi),
color="red",
linestyle="--")
ax[0].plot(l,
l * (l + 1) * two_predicted_Pl[1] / (2 * np.pi),
color="green",
linestyle="--")
#Measured
ax[0].plot(l,
l * (l + 1) * emulator.training_set[0] / (2 * np.pi),
color="red",
linestyle="-")
ax[0].plot(l,
l * (l + 1) * testing_Pl / (2 * np.pi),
color="green",
linestyle="-")
#Fractional difference
ax[1].plot(l, (two_predicted_Pl[0] - emulator.training_set[0]) /
emulator.training_set[0],
color="red")
ax[1].plot(l, (two_predicted_Pl[1] - testing_Pl) / testing_Pl,
color="green")
ax[1].set_xlabel(r"$l$")
ax[0].set_ylabel(r"$l(l+1)P_l/2\pi$")
ax[1].set_ylabel(r"$P_l^I-P_l^M/P_l^M$")
ax[0].set_yscale("log")
plt.savefig("power_interpolator_test_2.png")
plt.clf()
#Generate a fudge power spectrum covariance matrix
covariance = np.diag(testing_Pl**2 / (0.5 + l))
#Generate a fudge observation by wiggling the testing power spectrum
observation = testing_Pl + np.random.uniform(low=-testing_Pl * 0.1,
high=testing_Pl * 0.1)
#Choose a bunch of points in parameter space
points = emulator.parameter_set[:, :-1]
#Compute the chi2
chi2_values_1 = emulator.chi2(points, observation, covariance)
chi2_values_2 = emulator.chi2(points,
observation,
covariance,
split_chunks=4)
assert chi2_values_1.shape == chi2_values_2.shape
#Compute the individual contributions
chi2_contributions = emulator.chi2Contributions(points[0], observation,
covariance)
#Plot
plt.imshow(chi2_contributions, interpolation="nearest")
plt.colorbar()
plt.xlabel(r"$j$")
plt.ylabel(r"$i$")
plt.savefig("chi2_contributions.png")
plt.clf()
return chi2_values_1, chi2_values_2
for subfield in subfields: fig_power,ax_power = plt.subplots() fig_peaks,ax_peaks = plt.subplots() fig_minkowski,ax_minkowski = plt.subplots(1,3,figsize=(24,8)) #Plot for the sumulations for model in training_models + [observed_model]: m = Measurement(model=model,options=options,subfield=subfield,smoothing_scale=smoothing_scale,measurer=None) m.get_all_map_names() #Load the features and plot #Power spectrum ensemble_power = Ensemble.fromfilelist([os.path.join(m.full_save_path,"power_spectrum.npy")]) ensemble_power.load(from_old=True) P = ensemble_power.mean() if type(model) == CFHTLens: ax_power.plot(l,l*(l+1)*P/(2*np.pi),linestyle="--",color="black") elif type(model) == CFHTemu1: ax_power.plot(l,l*(l+1)*P/(2*np.pi)) #Peaks ensemble_peaks = Ensemble.fromfilelist([os.path.join(m.full_save_path,"peaks.npy")]) ensemble_peaks.load(from_old=True) pk = ensemble_peaks.mean() if type(model) == CFHTLens: ax_peaks.plot(v_pk,pk,linestyle="--",color="black")
#Create smoothing scale index for the histogram
idx = Indexer.stack([PDF(bin_edges) for scale in smoothing_scales])
#Create IGS1 simulation set object to look for the right simulations
simulation_set = IGS1(root_path=root_path)
#Look at a sample map
sample_map = ConvergenceMap.load(
simulation_set.getNames(z=1.0, realizations=[1])[0])
#Initialize Gaussian shape noise generator
generator = GaussianNoiseGenerator.forMap(sample_map)
#Build Ensemble instance with the maps to analyze
map_ensemble = Ensemble.fromfilelist(range(1, num_realizations + 1))
#Measure the histograms and load the data in the ensemble
map_ensemble.load(callback_loader=compute_histograms,
pool=pool,
simulation_set=simulation_set,
smoothing_scales=smoothing_scales,
index=idx,
generator=generator,
bin_edges=bin_edges)
if pool is not None:
pool.close()
##########################################################################################################################################
###############################Ensemble data available at this point for covariance, PCA, etc...##########################################
def pbBias(cmd_args,feature_name="convergence_power_s0_nb100",title="Power spectrum",kappa_models=("Born",),callback=None,variation_idx=(0,),bootstrap_size=1000,resample=1000,return_results=False,fontsize=22):
#Initialize plot
fig,ax = plt.subplots(len(variation_idx),3,figsize=(24,8*len(variation_idx)))
ax = np.atleast_2d(ax)
##################
#Load in the data#
##################
#Observation
bootstrap_mean = lambda e: e.values.mean(0)
feature_ray = Ensemble.read(os.path.join(fiducial["c0"].getMapSet("kappa").home,feature_name+".npy"),callback_loader=callback).bootstrap(bootstrap_mean,bootstrap_size=bootstrap_size,resample=resample,seed=0)
#Containers for cosmological model
modelFeatures = dict()
for mf in kappa_models:
modelFeatures[mf] = dict()
parameters = dict()
for model in models:
parameters[model.cosmo_id] = np.array([model.cosmology.Om0,model.cosmology.w0,model.cosmology.sigma8])
for mf in kappa_models:
try:
modelFeatures[mf][model.cosmo_id] = Ensemble.read(os.path.join(model["c0"].getMapSet("kappa"+mf).home,feature_name+".npy"),callback_loader=callback)
except IOError:
pass
#Fit each model
for mf in kappa_models:
#Select correct
features = modelFeatures[mf]
###############################
#Compute the covariance matrix#
###############################
features_covariance = features[fiducial.cosmo_id].cov()
################################################
#Load in the feature to fit, bootstrap the mean#
################################################
feature_born = features[fiducial.cosmo_id].bootstrap(bootstrap_mean,bootstrap_size=bootstrap_size,resample=resample,seed=0)
for nv,v in enumerate(variation_idx):
###############################
#Initialize the FisherAnalysis#
###############################
ftr = np.array([features[m.cosmo_id].values.mean(0) for m in [fiducial] + variations[v]])
par = np.array([parameters[m.cosmo_id] for m in [fiducial] + variations[v]])
fisher = FisherAnalysis.from_features(ftr,par,parameter_index=["Om","w0","si8"])
#############
####Fit######
#############
fitted_parameters_born = fisher.fit(feature_born,features_covariance)
fitted_parameters_ray = fisher.fit(feature_ray,features_covariance)
if return_results:
assert len(kappa_models)==1
assert len(variation_idx)==1
return fitted_parameters_born,fitted_parameters_ray
##########
#Plotting#
##########
for n,p in enumerate(fisher.parameter_names):
fitted_parameters_born[p].plot.hist(bins=50,ax=ax[nv,n],edgecolor="none",label=r"${\rm Control}$")
fitted_parameters_ray[p].plot.hist(bins=50,ax=ax[nv,n],edgecolor="none",label=r"${\rm Observation}$")
ax[nv,n].set_xlabel(plab[p],fontsize=fontsize)
ax[nv,n].set_title(title)
ax[nv,n].legend(loc="upper right",mode="expand",ncol=2,prop={"size":20})
#Labels
for a in ax.flatten():
plt.setp(a.get_xticklabels(),rotation=30)
#Save
fig.tight_layout()
fig.savefig("{0}/bornBias_{1}.{0}".format(cmd_args.type,feature_name))
bin_midpoints = 0.5*(bin_edges[1:] + bin_edges[:-1]) #Create smoothing scale index for the histogram idx = Indexer.stack([PDF(bin_edges) for scale in smoothing_scales]) #Create IGS1 simulation set object to look for the right simulations simulation_set = IGS1(root_path=root_path) #Look at a sample map sample_map = ConvergenceMap.load(simulation_set.getNames(z=1.0,realizations=[1])[0]) #Initialize Gaussian shape noise generator generator = GaussianNoiseGenerator.forMap(sample_map) #Build Ensemble instance with the maps to analyze map_ensemble = Ensemble.fromfilelist(range(1,num_realizations+1)) #Measure the histograms and load the data in the ensemble map_ensemble.load(callback_loader=compute_histograms,pool=pool,simulation_set=simulation_set,smoothing_scales=smoothing_scales,index=idx,generator=generator,bin_edges=bin_edges) if pool is not None: pool.close() ########################################################################################################################################## ###############################Ensemble data available at this point for covariance, PCA, etc...########################################## ########################################################################################################################################## #Plot results to check fig,ax = plt.subplots(len(smoothing_scales),1) for i in range(len(smoothing_scales)):
pool = None
if (pool is not None) and not(pool.is_master()):
pool.wait()
sys.exit(0)
map_mock_ids = range(int(sys.argv[1]))
igs1_set = IGS1(root_path="/Users/andreapetri/Documents/Columbia/spurious_shear/convergence_maps")
map_igs1_ids = igs1_set.getNames(z=1.0,realizations=range(1,int(sys.argv[1])+1))
gen = GaussianNoiseGenerator(shape=(2048,2048),side_angle=3.41*deg,label="convergence")
power_func = np.loadtxt("Data/ee4e-7.txt",unpack=True)
ens_mock = Ensemble.fromfilelist(map_mock_ids)
ens_igs1 = Ensemble.fromfilelist(map_igs1_ids)
ens_mock.load(callback_loader=generate_and_measure,pool=pool,generator=gen,power_func=power_func)
ens_igs1.load(callback_loader=measure_from_IGS1,pool=pool)
if pool is not None:
pool.close()
np.savetxt("moments_mock.txt",np.array([ens_mock.mean(),np.sqrt(ens_mock.covariance().diagonal())]))
np.savetxt("moments_igs1.txt",np.array([ens_igs1.mean(),np.sqrt(ens_igs1.covariance().diagonal())]))
logging.info("Done!")
pool = None
#The only parallelized part is the loading of the ensemble (that's the computationally expensive part)
if (pool is not None) and not(pool.is_master()):
pool.wait()
sys.exit(0)
map_list = ["Data/conv1.fit","Data/conv2.fit","Data/conv3.fit","Data/conv4.fit"]
l_edges = np.arange(200.0,50000.0,200.0)
thresholds_pk = np.arange(-1.0,5.0,0.2)
l = 0.5*(l_edges[:-1] + l_edges[1:])
conv_ensemble = Ensemble.fromfilelist(map_list)
conv_ensemble.load(callback_loader=default_callback_loader,pool=pool,l_edges=l_edges)
if pool is not None:
pool.close()
def test_shape():
assert conv_ensemble.num_realizations==len(map_list)
assert conv_ensemble.data.shape==(len(map_list),len(l_edges)-1)
def test_power_plot():
fig,ax = plt.subplots()
for n in range(conv_ensemble.num_realizations):
ax.plot(l,l*(l+1)*conv_ensemble[n]/(2.0*np.pi),label="Map {0}".format(n+1),linestyle="--")
def test_chi2():
conv_ensemble1 = Ensemble.fromfilelist(map_list[0:2])
conv_ensemble1.load(callback_loader=default_callback_loader,pool=None,l_edges=l_edges)
print("chi2 difference = {0}".format(conv_ensemble.compare(conv_ensemble1)))
def test_interpolation():
root_path = "Data/all"
analysis = LikelihoodAnalysis()
#Read in model names
models = CFHTemu1.getModels()[:17]
assert len(models) == 17
#Shuffle the models
np.random.seed(1)
np.random.shuffle(models)
#Divide into training and testing
training_models = models[:-1]
testing_model = models[-1]
#Read multipoles
ell = np.load(os.path.join(root_path,"ell.npy"))
#Load in the means of the power spectra of the 17 models, and populate the analysis instance
for model in training_models:
ens = Ensemble.fromfilelist([os.path.join(root_path,model._cosmo_id_string,"subfield1","sigma05","power_spectrum.npy")])
ens.load(from_old=True)
analysis.add_model(parameters=model.squeeze(with_ns=True),feature=ens.mean())
#Add the multipoles to the analysis
analysis.add_feature_label(ell)
l = analysis.feature_label
ens = Ensemble.fromfilelist([os.path.join(root_path,testing_model._cosmo_id_string,"subfield1","sigma05","power_spectrum.npy")])
ens.load(from_old=True)
testing_Pl = ens.mean()
#Load in also the observed power spectrum
ens = Ensemble.fromfilelist([os.path.join(root_path,"observations","subfield1","sigma05","power_spectrum.npy")])
ens.load(from_old=True)
observed_Pl = ens.mean()
#Output the analysis stats
np.savetxt("16_parameter_points.txt",analysis.parameter_set)
for n in range(len(training_models)):
plt.plot(l,l*(l+1)*analysis.training_set[n]/(2*np.pi))
plt.plot(l,l*(l+1)*observed_Pl/(2*np.pi),linestyle="--",label="Observation")
plt.xlabel(r"$l$")
plt.ylabel(r"$l(l+1)P_l/2\pi$")
plt.yscale("log")
plt.legend(loc="upper left")
plt.savefig("16_power_spectra.png")
plt.clf()
#Train the interpolators
analysis.train(use_parameters=range(3))
assert hasattr(analysis,"_interpolator")
assert hasattr(analysis,"_num_bins")
#Emulator portability test with pickle/unpickle
analysis.save("analysis.p")
emulator = LikelihoodAnalysis.load("analysis.p")
#Predict the power spectrum at the remaining point
predicted_Pl = emulator.predict(testing_model.squeeze())
#Plot it against the measured one
fig,ax = plt.subplots(2,1,figsize=(16,8))
#Measured
ax[0].plot(l,l*(l+1)*testing_Pl/(2*np.pi),label="measured")
#Predicted
ax[0].plot(l,l*(l+1)*predicted_Pl/(2*np.pi),label="interpolated")
#Fractional difference
ax[1].plot(l,(predicted_Pl - testing_Pl)/testing_Pl)
ax[1].set_xlabel(r"$l$")
ax[0].set_ylabel(r"$l(l+1)P_l/2\pi$")
ax[1].set_ylabel(r"$P_l^I-P_l^M/P_l^M$")
ax[0].set_yscale("log")
ax[0].legend(loc="upper left")
plt.savefig("power_interpolator_test.png")
plt.clf()
#Give it a shot with two points in parameter space to test vectorization
two_parameter_points = np.array((training_models[0].squeeze(),testing_model.squeeze()))
two_predicted_Pl = emulator.predict(two_parameter_points)
fig,ax = plt.subplots(2,1,figsize=(16,8))
#Predicted
ax[0].plot(l,l*(l+1)*two_predicted_Pl[0]/(2*np.pi),color="red",linestyle="--")
ax[0].plot(l,l*(l+1)*two_predicted_Pl[1]/(2*np.pi),color="green",linestyle="--")
#Measured
ax[0].plot(l,l*(l+1)*emulator.training_set[0]/(2*np.pi),color="red",linestyle="-")
ax[0].plot(l,l*(l+1)*testing_Pl/(2*np.pi),color="green",linestyle="-")
#Fractional difference
ax[1].plot(l,(two_predicted_Pl[0] - emulator.training_set[0])/emulator.training_set[0],color="red")
ax[1].plot(l,(two_predicted_Pl[1] - testing_Pl)/testing_Pl,color="green")
ax[1].set_xlabel(r"$l$")
ax[0].set_ylabel(r"$l(l+1)P_l/2\pi$")
ax[1].set_ylabel(r"$P_l^I-P_l^M/P_l^M$")
ax[0].set_yscale("log")
plt.savefig("power_interpolator_test_2.png")
plt.clf()
#Generate a fudge power spectrum covariance matrix
covariance = np.diag(testing_Pl**2/(0.5 + l))
#Generate a fudge observation by wiggling the testing power spectrum
observation = testing_Pl + np.random.uniform(low=-testing_Pl*0.1,high=testing_Pl*0.1)
#Choose a bunch of points in parameter space
points = emulator.parameter_set[:,:-1]
#Compute the chi2
chi2_values_1 = emulator.chi2(points,observation,covariance)
chi2_values_2 = emulator.chi2(points,observation,covariance,split_chunks=4)
assert chi2_values_1.shape == chi2_values_2.shape
#Compute the individual contributions
chi2_contributions = emulator.chi2Contributions(points[0],observation,covariance)
#Plot
plt.imshow(chi2_contributions,interpolation="nearest")
plt.colorbar()
plt.xlabel(r"$j$")
plt.ylabel(r"$i$")
plt.savefig("chi2_contributions.png")
plt.clf()
return chi2_values_1,chi2_values_2
def load_features(self,model,save_new=False):
#First create an empty ensemble
ensemble_all_subfields = Ensemble()
#Then cycle through all the subfields and gather the features for each one
for subfield in self.subfields:
#Dictionary that holds all the measurements
m = dict()
for smoothing_scale in self.smoothing_scales:
m[smoothing_scale] = Measurement(model=model,options=self.options,subfield=subfield,smoothing_scale=smoothing_scale,measurer=None,mean_subtract=self.cmd_args.mean_subtract)
m[smoothing_scale].get_all_map_names()
#Construct one ensemble for each feature (with included smoothing scales) and load in the data
ensemble_subfield = list()
#Prevent randomness in dictionary keys
features_to_measure = self.features_to_measure.keys()
features_to_measure.sort()
for feature_type in features_to_measure:
for smoothing_scale in self.features_to_measure[feature_type]:
#Construct the subfield/smoothing scale/feature specific ensemble
ens_filename = os.path.join(m[smoothing_scale].full_save_path,npy_filename(feature_type))
logging.info("Reading ensemble from {0}".format(ens_filename))
ens = Ensemble.fromfilelist([ens_filename])
ens.load(from_old=True)
#Check if we want to cut out some of the peaks
if self.cmd_args.cut_convergence and feature_type=="peaks":
new_thresholds = ens.cut(self.kappa_min,self.kappa_max,feature_label=self.kappa_peaks)
logging.log(DEBUG_PLUS,"Performed cut on the peaks convergence, new limits are {0},{1}".format(new_thresholds[0],new_thresholds[-1]))
if save_new:
logging.info("Saving new kappa values to {0}...".format(os.path.join(self.save_path,"th_new_peaks.npy")))
np.save(os.path.join(self.save_path,"th_new_peaks.npy"),new_thresholds)
#Check the masked fraction of the field of view
masked_fraction = self.masked_fraction[smoothing_scale][subfield]
###########################################################################################################################################################################################################
#Scale to the non masked area: if we treat each subfield independently (i.e. group_subfields is False) then we need to scale each subfield to the same area when considering the power spectrum and peaks##
#if on the other hand we group subfields together, then the power spectrum and peaks are simply added between subfields, but the MFs and the moments need to be scaled#####################################
###########################################################################################################################################################################################################
if (self.cmd_args.mask_scale) and not(self.cmd_args.group_subfields):
if feature_type=="power_spectrum":
logging.log(DEBUG_PLUS,"Scaling power spectrum of subfield {0}, masked fraction {1}, multiplying by {2}".format(subfield,masked_fraction,1.0/(1.0 - masked_fraction)**2))
ens.scale(1.0/(1.0 - masked_fraction)**2)
elif feature_type=="peaks":
logging.log(DEBUG_PLUS,"Scaling peak counts of subfield {0}, masked fraction {1}, multiplying by {2}".format(subfield,masked_fraction,1.0/(1.0 - masked_fraction)))
ens.scale(1.0/(1.0 - masked_fraction))
elif (self.cmd_args.mask_scale) and (self.cmd_args.group_subfields):
if "minkowski" in feature_type or "moments" in feature_type:
logging.log(DEBUG_PLUS,"Scaling {0} of subfield {1}, masked fraction {2}, multiplying by {3}".format(feature_type,subfield,masked_fraction,(1.0 - masked_fraction)/self.total_non_masked_fraction[smoothing_scale]))
ens.scale((1.0 - masked_fraction)/self.total_non_masked_fraction[smoothing_scale])
#Regular expressions to parse the feature string
num = re.match(r"minkowski_([0-2]+)",feature_type)
momParse = re.match(r"moments(_[qsk][1-4]+)?(_[qsk][1-4]+)?(_[qsk][1-4]+)?",feature_type)
assert (num is None) or (momParse is None)
if num is not None:
###############################################################################
##MFs only: check if we want to discard some of the Minkowski functionals######
###############################################################################
mink_to_measure = [ int(n_mf) for n_mf in list(num.group(1)) ]
ens_split = ens.split(self.mink_idx)
#Differentiate Minkowski 0 to find the PDF?
if self.cmd_args.differentiate:
logging.log(DEBUG_PLUS,"Differentiating Minkowski 0 to get the PDF")
ens_split[0] = ens_split[0].differentiate(step=self.kappa_minkowski[0]-self.kappa_minkowski[1])
#Perform the convergence cut if option is enabled
if self.cmd_args.cut_convergence:
new_thresholds = [ ens_split[n_mf].cut(self.kappa_min,self.kappa_max,feature_label=self.kappa_minkowski) for n_mf in mink_to_measure ]
logging.log(DEBUG_PLUS,"Performed cut on the minkowski convergence, new limits are {0},{1}".format(new_thresholds[0][0],new_thresholds[0][-1]))
if save_new:
logging.info("Saving new kappa values to {0}...".format(os.path.join(self.save_path,"th_new_minkowski.npy")))
np.save(os.path.join(self.save_path,"th_new_minkowski.npy"),new_thresholds[0])
[ ensemble_subfield.append(ens_split[n_mf]) for n_mf in mink_to_measure ]
elif momParse is not None:
###############################################################################
##Moments only: check if we want to keep only a subset of the moments##########
###############################################################################
momGroups = momParse.groups()
mom_indices = list()
#Check, one by one, the moment indices
for gr in momGroups:
if gr is not None:
moment_type = gr[1]
moment_numbers = [ int(n_mom)-1 for n_mom in gr[2:] ]
#Compute offset
if moment_type=="q":
mom_offset=0
elif moment_type=="s":
mom_offset=2
elif moment_type=="k":
mom_offset=5
else:
raise ValueError("Only quadratic, cubic, quartic moments implemented!")
#Append indices to list
for mom_num in moment_numbers:
mom_indices.append(mom_offset+mom_num)
#Slice the ensemble accordingly
if len(mom_indices)>0:
logging.info("Measuring moments {0}".format(mom_indices))
ens.cut(mom_indices)
#Append to subfield
ensemble_subfield.append(ens)
else:
ensemble_subfield.append(ens)
if self.cmd_args.cut_convergence:
logging.log(DEBUG_PLUS,"Convergence cut on MFs not performed, select minkowski_012 instead of minkowski_all")
#############################################################################################
#Add the features to the cumulative subfield ensemble
ensemble_all_subfields += reduce(mul,ensemble_subfield)
#If option is specified, group all the subfields together, for each realization
if self.cmd_args.group_subfields:
logging.log(DEBUG_PLUS,"Taking means over the {0} subfields...".format(len(self.subfields)))
ensemble_all_subfields.group(group_size=len(self.subfields),kind="sparse")
#Return the created ensemble
return ensemble_all_subfields
#The only parallelized part is the loading of the ensemble (that's the computationally expensive part)
if (pool is not None) and not (pool.is_master()):
pool.wait()
sys.exit(0)
map_list = [
"Data/conv1.fit", "Data/conv2.fit", "Data/conv3.fit", "Data/conv4.fit"
]
l_edges = np.arange(200.0, 50000.0, 200.0)
thresholds_pk = np.arange(-1.0, 5.0, 0.2)
l = 0.5 * (l_edges[:-1] + l_edges[1:])
conv_ensemble = Ensemble.fromfilelist(map_list)
conv_ensemble.load(callback_loader=default_callback_loader,
pool=pool,
l_edges=l_edges)
if pool is not None:
pool.close()
def test_shape():
assert conv_ensemble.num_realizations == len(map_list)
assert conv_ensemble.data.shape == (len(map_list), len(l_edges) - 1)
def test_power_plot():
