def getMCMCchainForSamplesSize( iSampleSize, nIterations,\
hubbleInterpolaterClass):
samples = None
pklFile = 'picklesDualMass/multiFitSamples_dualMass_%i.pkl' \
% (iSampleSize)
if os.path.isfile(pklFile):
return pkl.load(open(pklFile, 'rb'))
logProb = np.array([])
for iIteration in range(nIterations):
print("Iteration: %i/%i" % (iIteration + 1, nIterations))
selectedTimeDelays = \
selectRandomSampleOfTimeDelays( iSampleSize)
fitHubbleClass = \
fitHubble.fitHubbleParameterClass( selectedTimeDelays, \
hubbleInterpolaterClass)
y, x = np.histogram(fitHubbleClass.samples[:, 0])
xc = (x[1:] + x[:-1]) / 2.
if samples is None:
samples = fitHubbleClass.samples
else:
samples = np.vstack((samples, fitHubbleClass.samples))
pdb.set_trace()
pkl.dump(samples, open(pklFile, 'wb'))
return samples
def main(pklFile='fitDataToModel_withMass_LCDMk_noZ0.6.pkl', \
monteCarlo=False):
if not os.path.isfile('pickles/monteCarlosOfData/%s' % pklFile):
selectedTimeDelays = getObservations(monteCarlo=monteCarlo)
dataSelectionFunction = '../output/CDM/selectionFunction/SF_data.pkl'
hubbleInterpolaterClass = \
hubbleInterpolator(allDistributionsPklFile=dataSelectionFunction)
hubbleInterpolaterClass.getTrainingData(
'pickles/trainingDataForObsWithMass.pkl')
hubbleInterpolaterClass.getTimeDelayModel()
fitHubbleClass = \
fitHubble.fitHubbleParameterClass( selectedTimeDelays, \
hubbleInterpolaterClass)
pkl.dump(fitHubbleClass.samples,
open('pickles/monteCarlosOfData/%s' % pklFile, 'wb'))
samples = fitHubbleClass.samples
else:
samples = pkl.load(open('pickles/monteCarlosOfData/%s' % pklFile,
'rb'))
if monteCarlo:
return samples
ndim = samples.shape[1]
figcorner, axarr = plt.subplots(ndim, ndim, figsize=(8, 8))
parRange = [[0.55, 0.8], [0.45, 0.6], [-2., -1.6]]
labels = \
[r'$H_0/ (100 $km s$^{-1}$Mpc$^{-1}$)',r'$z_{lens}$',r'$\alpha$', r'$log(M(<5kpc)/M_\odot)$'\
r'$\Omega_M$',r'$\Omega_\Lambda$',r'$\Omega_K$', r'$Sigma' ]
nsamples = samples.shape[0]
#plotMockedData( figcorner)
corner.corner(samples, levels=[0.68], bins=20,\
plot_datapoints=False, labels=labels, fig=figcorner,\
weights=np.ones(nsamples)/nsamples, color='black', \
hist_kwargs={'linewidth':3.},\
contour_kwargs={'linewidths':3.}, smooth=True)
for i in range(samples.shape[1]):
lo, med, hi = np.percentile(samples[:, i], [16, 50, 84])
print(med, med - lo, hi - med)
plt.savefig('../plots/fitDataToModel.pdf')
plt.show()
def getMCMCchainForSamplesSize( iSampleSize, nIterations,\
hubbleInterpolaterClass,\
minimumTimeDelay=0.):
samples = None
if minimumTimeDelay > 0:
pklFile = 'picklesMinimumDelay/multiFitSamples_withMass_%i.pkl' \
% (iSampleSize)
else:
pklFile = 'exactPDFpickles/multiFitSamples_withMass_%i.pkl' \
% iSampleSize
if os.path.isfile(pklFile):
return pkl.load(open(pklFile, 'rb'))
logProb = np.array([])
for iIteration in range(nIterations):
print("Iteration: %i/%i" % (iIteration+1, nIterations))
selectedTimeDelays = \
selectRandomSampleOfTimeDelays( iSampleSize, \
minimumTimeDelay=minimumTimeDelay)
fitHubbleClass = \
fitHubble.fitHubbleParameterClass( selectedTimeDelays, \
hubbleInterpolaterClass)
if samples is None:
samples = fitHubbleClass.samples
else:
samples = np.vstack( (samples, fitHubbleClass.samples))
pkl.dump(samples, open(pklFile, 'wb'))
return samples
def getMCMCchainForSamplesSize( iSampleSize, nIterations,\
hubbleInterpolaterClass, massCut=0.):
samples = None
if massCut > 0:
pklFile = 'picklesMassCut/multiFitSamples_%i_%0.2f.pkl' \
% (iSampleSize, massCut)
else:
pklFile = 'exactPDFpickles/multiFitSamples_%i.pkl' \
% iSampleSize
if os.path.isfile(pklFile):
return pkl.load(open(pklFile, 'rb'))
for iIteration in range(nIterations):
print("Iteration: %i/%i" % (iIteration+1, nIterations))
selectedTimeDelays = \
selectRandomSampleOfTimeDelays( iSampleSize, massCut=massCut)
fitHubbleClass = \
fitHubble.fitHubbleParameterClass( selectedTimeDelays, \
hubbleInterpolaterClass)
y, x = np.histogram(fitHubbleClass.samples[:,0])
xc = (x[1:] + x[:-1])/2.
if samples is None:
samples = fitHubbleClass.samples
else:
samples = np.vstack( (samples, fitHubbleClass.samples))
pkl.dump(samples, open(pklFile, 'wb'))
return samples