def __init__(self, config, r, Alens, radec):
self.config = config
self.nus = self.config['frequency']
self.depth_p = self.config['depth_p']
#print(self.depth_p)
#stop
self.nside = 64
self.npix = 12 * self.nside**2
self.nfreqs = len(self.nus)
self.w = None
self.lmin = 20
self.lmax = 2 * self.nside - 1
#self.lmax=355
self.dl = 10
self.r = r
self.Alens = Alens
self.cc = 0
self.fsky = self.config['fsky']
self.radec = radec
self.covmap = get_coverage(self.fsky,
self.nside,
center_radec=self.radec)
maskpix = np.zeros(12 * self.nside**2)
self.pixok = self.covmap > 0
maskpix[self.pixok] = 1
self.Namaster = nam.Namaster(maskpix,
lmin=self.lmin,
lmax=self.lmax,
delta_ell=self.dl)
self.Namaster.fsky = self.fsky
self.Namaster.ell_binned, _ = self.Namaster.get_binning(self.nside)
#self.Namaster.ell_binned=self.Namaster.ell_binned[:-1]
#print(self.Namaster.ell_binned.shape)
print('*********************************')
print('********** Parameters ***********')
print('*********************************')
print()
print(' Frequency [GHz] : {}'.format(self.nus))
print(' Depth [muK.arcmin] : {}'.format(self.depth_p))
print(' Nside : {}'.format(self.nside))
print(' Fsky : {}'.format(self.fsky))
print(' Patch sky : {}'.format(self.radec))
print(' lmin : {}'.format(self.lmin))
print(' lmax : {}'.format(self.lmax))
print(' dl : {}'.format(self.dl))
print(' r : {}'.format(self.r))
print(' Alens : {}'.format(self.Alens))
print(' ell : {}'.format(self.Namaster.ell_binned))
print()
print('*********************************')
print('*********************************')
print('*********************************\n')
def get_binned_spectra(nside, lmin, lmax, delta_ell, input_spectra):
### Making mask - it will be automaticall apodized when instanciating the object with default (tunable) parameters
import NamasterLib as nam
mask = np.zeros(12 * nside**2)
mask[np.arange(10)] = 1
Namaster = nam.Namaster(mask, lmin=lmin, lmax=lmax, delta_ell=delta_ell)
ell_binned, b = Namaster.get_binning(nside)
#print(input_spectra.shape)
binned_spectra = Namaster.bin_spectra(
input_spectra, nside) #Namaster.bin_spectra(input_Dl[0, :], nside)
return binned_spectra
# print(residuals.shape)
#
# # There is only the patch so you need to put them in a full map to plot with healpy
# noisemaps = np.zeros_like(qubicmaps)
# noisemaps[:, :, seenmap, :] = residuals
# ================== Power spectrum =============================
print('\n =============== Starting power spectrum ================')
# Create a Namaster object
lmin = 40
lmax = 2 * d['nside'] - 1
delta_ell = 30
mask = np.zeros(12 * d['nside']**2)
mask[seenmap] = 1
Namaster = nam.Namaster(mask, lmin=lmin, lmax=lmax, delta_ell=delta_ell)
mask_apo = Namaster.get_apodized_mask()
ell_binned, b = Namaster.get_binning(d['nside'])
nbins = len(ell_binned)
print('lmin:', lmin)
print('lmax:', lmax)
print('delta_ell:', delta_ell)
print('nbins:', nbins)
print('ell binned:', ell_binned)
# Possible combinations between bands
combi = list(combinations_with_replacement(np.arange(nbands), 2))
ncombi = len(combi)
print('combi:', combi)
print('ncombi:', ncombi)
def apply_fgb(inmaps,
freqs,
fwhms,
verbose=True,
apodize=0,
plot_apo=False,
apocut=False,
apotype='C1',
coverage_recut=None,
coverage=None,
resol_correction=True,
ref_fwhm=0.5,
alm_space=False,
plot_separated=False,
center=None,
add_title='',
plot_residuals=False,
truth=None,
alm_maps=False,
apply_to_unconvolved=False):
### FGB Configuration
instrument = fgb.get_instrument('Qubic')
instrument.frequency = freqs
instrument.fwhm = fwhms
components = [fgb.Dust(150., temp=20.), fgb.CMB()]
### Check good pixels
pixok = inmaps[0, 0, :] != hp.UNSEEN
nside = hp.npix2nside(len(inmaps[0, 0, :]))
if apodize != 0:
mymask = pixok.astype(float)
nmt = nam.Namaster(mymask,
40,
400,
30,
aposize=apodize,
apotype=apotype)
apodized_mask = nmt.get_apodized_mask()
if plot_apo:
hp.gnomview(apodized_mask,
title='Apodized Mask {} deg.'.format(apodize),
reso=15,
rot=center)
maps = inmaps * apodized_mask
maps[:, :, ~pixok] = hp.UNSEEN
else:
maps = inmaps.copy()
apodized_mask = np.ones(12 * nside**2)
### Data to feed FGB:
if alm_space:
if verbose:
print(
'\nFBG in alm-space with resol_correction={} and ref_resol={}'.
format(resol_correction, ref_fwhm))
mydata = get_alm_maps(maps,
fwhms,
ref_fwhm=ref_fwhm,
resol_correction=resol_correction,
verbose=verbose)
if (truth is not None) & ~apply_to_unconvolved:
if verbose:
print('\nNow reconvolving truth to ref_fwhm = {}'.format(
ref_fwhm))
mytruth = [
hp.smoothing(truth[0],
fwhm=np.deg2rad(ref_fwhm),
pol=True,
verbose=False),
hp.smoothing(truth[1],
fwhm=np.deg2rad(ref_fwhm),
pol=True,
verbose=False)
]
space = ' (alm based)'
else:
if verbose:
print(
'\nFBG in pixel-space with resol_correction={} and ref_resol={}'
.format(resol_correction, ref_fwhm))
if resol_correction:
if verbose:
print('\nNow reconvolving input maps to ref_fwhm = {}'.format(
ref_fwhm))
mydata = reconvolve(maps, fwhms, ref_fwhm, verbose=verbose)
if (truth is not None):
if verbose:
print(
'\nNow reconvolving truth (dust and CMB) to ref_fwhm = {}'
.format(ref_fwhm))
mytruth = [
hp.smoothing(truth[0],
fwhm=np.deg2rad(ref_fwhm),
pol=True,
verbose=False),
hp.smoothing(truth[1],
fwhm=np.deg2rad(ref_fwhm),
pol=True,
verbose=False)
]
space = ' (Pixel based - Reconv.)'
else:
mydata = maps
if (truth is not None):
mytruth = truth.copy()
space = ' (Pixel based - No Reconv.)'
if coverage_recut is not None:
if verbose:
print('Applying coverage recut to {}'.format(coverage_recut))
fidregion = (coverage > (coverage_recut * np.max(coverage)))
mydata[..., ~fidregion] = hp.UNSEEN
mapregions = np.zeros(12 * nside**2) + hp.UNSEEN
mapregions[pixok] = 1
mapregions[fidregion] = 2
if verbose:
hp.gnomview(mapregions,
rot=center,
reso=15,
title='Fiducial region: {}'.format(coverage_recut))
show()
if (apodize != 0) & (apocut == True):
fidregion = apodized_mask == 1
mydata[..., ~fidregion] = hp.UNSEEN
### FGB itself
if verbose: print('Starting FGBuster in s')
r = separate(components, instrument, mydata, print_option=verbose)
if verbose: print('Resulting beta: {}'.format(r.x[0]))
### Resulting separated maps
if alm_space:
if alm_maps:
### Directly use the output from FGB => not very accurate because of alm-transform near the edges of the patch
almdustrec = r.s[0, :, :]
print('ALM', np.shape(almdustrec))
dustrec = hp.alm2map(
almdustrec[..., ::2] + almdustrec[..., 1::2] * 1j, nside)
dustrec[:, ~pixok] = hp.UNSEEN
almcmbrec = r.s[1, :, :]
cmbrec = hp.alm2map(
almcmbrec[..., ::2] + almcmbrec[..., 1::2] * 1j, nside)
cmbrec[:, ~pixok] = hp.UNSEEN
else:
### Instead we use the fitted beta and recalculate the component maps in pixel space
### This avoids inaccuracy of the alm transformation near the edges of the patch
A = fgb.MixingMatrix(*components)
print(' >>> building s = Wd in pixel space')
A_ev = A.evaluator(instrument.frequency)
A_maxL = A_ev(r.x)
print('A_maxl', np.shape(A_maxL))
if apply_to_unconvolved:
# We apply the mixing matrix to maps each at its own resolution...
# this allows to keep the effecgive resolution as good as possible, but surely the bell is no
# longer Gaussian. It is the combnation of various resolutions with weird weights.
themaps = maps.copy()
# Now we calculate the effective Bell
Bl_each = []
for fw in fwhms:
Bl_gauss_fwhm = hp.gauss_beam(np.radians(fw),
lmax=2 * nside + 1)
Bl_each.append(Bl_gauss_fwhm)
plot(Bl_gauss_fwhm, label='FWHM {:4.2}'.format(fw))
Bl_each = np.array(Bl_each)
### Compute W matrix (from Josquin)
invN = np.diag(np.ones(len(fwhms)))
inv_AtNA = np.linalg.inv(A_maxL.T.dot(invN).dot(A_maxL))
W = inv_AtNA.dot(A_maxL.T).dot(invN)
Bl_eff_Dust = W.dot(Bl_each)[0] #1 is for the Dust
Bl_eff_CMB = W.dot(Bl_each)[1] #1 is for the CMB component
# print('W matrix')
# print(W)
# print('A_maxL matrix')
# print(A_maxL)
# print('Bl_each:',np.shape(Bl_each))
Bl_eff_Dust_new = np.zeros(2 * nside + 1)
Bl_eff_CMB_new = np.zeros(2 * nside + 1)
ones_comp = np.ones(2) #number of components
for i in range(2 * nside + 1):
blunknown = W.dot(Bl_each[:, i] *
np.dot(A_maxL, ones_comp))
Bl_eff_Dust_new[i] = blunknown[0]
Bl_eff_CMB_new[i] = blunknown[1]
plot(Bl_eff_CMB, ':', label='Effective Bl CMB: W.Bl', lw=3)
plot(Bl_eff_Dust, ':', label='Effective Bl Dust: W.Bl', lw=3)
plot(Bl_eff_CMB_new,
'--',
label='Effective Bl CMB NEW: W.B.A.1',
lw=3)
plot(Bl_eff_Dust_new,
'--',
label='Effective Bl Dust NEW: W.B.A.1',
lw=3)
legend()
# We need to smooth the truth with this Bell
if (truth is not None):
if verbose:
print(
'\nNow reconvolving truth (dust and CMB) with effective Bell'
)
mytruth = [
hp.smoothing(truth[0],
beam_window=Bl_eff_Dust_new,
pol=True,
verbose=False),
hp.smoothing(truth[1],
beam_window=Bl_eff_CMB_new,
pol=True,
verbose=False)
]
else:
# We will apply the mixing matrix to maps at a chosen reeference resolution
# this might not bee the most optimal although it is simple in the sens that the components maps resolution
# is what we have chosen and is gaussian.
themaps = reconvolve(maps, fwhms, ref_fwhm, verbose=verbose)
components_bell = hp.gauss_beam(np.radians(ref_fwhm),
lmax=2 * nside + 1)
themaps[themaps == hp.UNSEEN] = 0
### Needed to comment the two lines below as prewhiten_factors was None
#prewhiten_factors = fgb.separation_recipes._get_prewhiten_factors(instrument, themaps.shape, nside)
#invN = np.zeros(prewhiten_factors.shape+prewhiten_factors.shape[-1:])
r.s = fgb.algebra.Wd(A_maxL, themaps.T) #, invN=invN)
r.s = np.swapaxes(r.s, -1, 0)
dustrec = r.s[0, :, :]
dustrec[:, ~pixok] = hp.UNSEEN
cmbrec = r.s[1, :, :]
cmbrec[:, ~pixok] = hp.UNSEEN
else:
dustrec = r.s[0, :, :]
cmbrec = r.s[1, :, :]
if plot_separated:
display_maps(
dustrec,
bigtitle=r'$\beta=${0:7.6f} - Reconstructed Dust'.format(r.x[0]) +
space,
rot=center,
figsize=(16, 7))
display_maps(
cmbrec,
bigtitle=r'$\beta=${0:7.6f} - Reconstructed CMB'.format(r.x[0]) +
space,
rot=center,
figsize=(16, 7))
if truth:
resid_dust = dustrec - mytruth[0] * apodized_mask
resid_dust[:, ~pixok] = hp.UNSEEN
resid_cmb = cmbrec - mytruth[1] * apodized_mask
resid_cmb[:, ~pixok] = hp.UNSEEN
if coverage_recut:
resid_cmb[:, ~fidregion] = hp.UNSEEN
resid_dust[:, ~fidregion] = hp.UNSEEN
pixok = fidregion.copy()
if (apodize != 0) & (apocut == True):
resid_cmb[:, ~fidregion] = hp.UNSEEN
resid_dust[:, ~fidregion] = hp.UNSEEN
pixok = fidregion.copy()
sigs_dust = np.std(resid_dust[:, pixok], axis=1)
sigs_cmb = np.std(resid_cmb[:, pixok], axis=1)
if plot_residuals:
# display_maps(mytruth[0], bigtitle=r'$\beta=${0:7.6f} Input Dust'.format(r.x[0])+space,
# rot=center, figsize=(16, 7), add_rms=True)
# display_maps(mytruth[1], bigtitle=r'$\beta=${0:7.6f} - Input CMB'.format(r.x[0])+space,
# rot=center, figsize=(16, 7), add_rms=True)
display_maps(mytruth[0],
bigtitle='Truth Dust Reconvolved',
rot=center,
figsize=(16, 7),
add_rms=True,
unseen=~pixok)
display_maps(mytruth[1],
bigtitle='Truth CMB Reconvolved',
rot=center,
figsize=(16, 7),
add_rms=True,
unseen=~pixok)
display_maps(
resid_dust,
bigtitle=r'$\beta=${0:7.6f} Residuals Dust'.format(r.x[0]) +
space,
rot=center,
figsize=(16, 7),
add_rms=True)
display_maps(
resid_cmb,
bigtitle=r'$\beta=${0:7.6f} - Residuals CMB'.format(r.x[0]) +
space,
rot=center,
figsize=(16, 7),
add_rms=True)
figure()
suptitle(r'$\beta=${0:7.6f} - Residuals:'.format(r.x[0]) + space,
fontsize=30,
y=1.05)
for i in range(3):
subplot(1, 3, i + 1)
hist(resid_dust[i, pixok],
range=[-5 * sigs_dust[i], 5 * sigs_dust[i]],
bins=100,
alpha=0.5,
color='b',
label='Dust: RMS={:4.2g}'.format(sigs_dust[i]),
density=True)
hist(resid_cmb[i, pixok],
range=[-5 * sigs_cmb[i], 5 * sigs_cmb[i]],
bins=100,
alpha=0.5,
color='r',
label='CMB: RMS={:4.2g}'.format(sigs_cmb[i]),
density=True)
title('Residuals Stokes {}'.format(stk[i]))
legend()
tight_layout()
if truth:
return r.x[
0], dustrec, cmbrec, sigs_dust, sigs_cmb, resid_dust, resid_cmb, mytruth[
0], mytruth[1]
else:
return r.x[0], dustrec, cmbrec
def _RUN_MC(self,
N,
fit,
x0,
covmap,
beta=[],
nside_param=0,
fixcmb=True,
noiseless=False):
maskpix = np.zeros(12 * 256**2)
pixok = covmap > 0
maskpix[pixok] = 1
Namaster = nam.Namaster(maskpix, lmin=21, lmax=355, delta_ell=35)
comp, _ = self._get_comp_instr(nu0=145, fit=fit, x0=x0)
instr = self._get_instr_nsamples(N_SAMPLE_BAND=10)
inputs = self.newfg.copy()
clBB = np.zeros((N, 9))
clres = np.zeros((N, 9))
components = np.zeros((N, 3, 3, np.sum(pixok)))
if nside_param != 0:
beta = np.zeros(
(N, fgbuster.MixingMatrix(*comp).n_param, 12 * 256**2))
else:
beta = np.zeros((N, fgbuster.MixingMatrix(*comp).n_param))
print('********** Separation ************')
for i in range(N):
if fixcmb:
seed = 42
else:
seed = np.random.randint(1000000)
print(seed)
cmb = self._get_cmb(self.covmap, seed)
if self.nside != 256:
newcmb = np.zeros((self.newfg.shape[0], self.newfg.shape[0],
12 * self.nside**2))
for i in range(self.fg.shape[0]):
newcmb[i] = hp.pixelfunc.ud_grade(cmb[i], self.nside)
else:
newcmb = cmb.copy()
data = inputs + newcmb.copy()
print(data.shape)
r1, components1 = self._separation(comp, instr, data, self.covmap,
beta, nside_param, fit, x0,
noiseless)
r2, components2 = self._separation(comp, instr, data, self.covmap,
beta, nside_param, fit, x0,
noiseless)
beta[i] = r1.copy()
components1[:, :, ~pixok] = 0
components2[:, :, ~pixok] = 0
components[i] = components1[:, :, pixok].copy()
leff, clBB[i] = self._get_clBB(components1[0], components2[0],
Namaster)
#res1=cmb[0]-components1[0]
#res2=cmb[0]-components2[0]
#res1[:, ~pixok]=0
#res2[:, ~pixok]=0
#leff, clres[i]=self._get_clBB(res1, res2, Namaster)
#print(cmb[0, 1, pixok]-components1[0, 1, pixok])
return leff, clBB, beta, components
def explore_like(self,
leff,
cl,
errors,
lmin,
dl,
cc,
rv,
otherp=None,
cov=None,
sample_variance=False):
# print(lmin, dl, cc)
# print(leff)
# print(scl_noise[:,2])
### Create Namaster Object
# Unfortunately we need to recalculate fsky for calculating sample variance
nside = 256
lmax = 355
if cov is None:
Namaster = nam.Namaster(None, lmin=lmin, lmax=lmax, delta_ell=dl)
Namaster.fsky = 0.03
else:
okpix = cov > (np.max(cov) * float(cc))
maskpix = np.zeros(12 * nside**2)
maskpix[okpix] = 1
Namaster = nam.Namaster(maskpix,
lmin=lmin,
lmax=lmax,
delta_ell=dl)
Namaster.fsky = 0.03
# print('Fsky: {}'.format(Namaster.fsky))
lbinned, b = Namaster.get_binning(nside)
### Bibnning CambLib
# binned_camblib = qc.bin_camblib(Namaster, '../../scripts/QubicGeneralPaper2020/camblib.pickle',
# nside, verbose=False)
global_dir = '/pbs/home/m/mregnier/sps1/QUBIC+/d0/cls'
binned_camblib = qc.bin_camblib(Namaster,
global_dir + '/camblib.pkl',
nside,
verbose=False)
### Redefine the function for getting binned Cls
def myclth(ell, r):
clth = qc.get_Dl_fromlib(ell, r, lib=binned_camblib,
unlensed=True)[1]
return clth
allfakedata = myclth(leff, 0.)
#lll, totDL, unlensedCL = qc.get_camb_Dl(lmax=3*256, r=0)
### And we need a fast one for BB only as well
def myBBth(ell, r):
clBB = qc.get_Dl_fromlib(ell,
r,
lib=binned_camblib,
unlensed=True,
specindex=2)[1]
return clBB
### Fake data
fakedata = cl.copy() #myBBth(leff, 0.)
if sample_variance:
covariance_model_funct = Namaster.knox_covariance
else:
covariance_model_funct = None
if otherp is None:
like, cumint, allrlim, maxL = self.ana_likelihood(
rv,
leff,
fakedata,
errors,
myBBth, [[0, 1]],
covariance_model_funct=covariance_model_funct)
else:
like, cumint, allrlim, other, maxL = self.ana_likelihood(
rv,
leff,
fakedata,
errors,
myBBth, [[0, 1]],
covariance_model_funct=covariance_model_funct,
otherp=otherp)
if otherp is None:
return like, cumint, allrlim, maxL
else:
return like, cumint, allrlim, other, maxL
def CreateAnafastPartialSky(cl, nside, lmin, lmax, delta_ell, f_sky = 1., plot_results = False, noise_rms = 0):
import qubic.NamasterLib as nam
npix = 12 * nside**2
# Determine SEEN pixels from f_sky using query_disc
vec = hp.pixelfunc.ang2vec(np.pi/2, np.pi*3/4)
radius = f_sky*np.pi
ipix_disc = hp.query_disc(nside=nside, vec=vec, radius=radius, nest=False)
while len(ipix_disc) < f_sky*npix:
radius += 0.01*np.pi
ipix_disc = hp.query_disc(nside=nside, vec=vec, radius=radius, nest=False)
#print("npix_partial_sky: ", len(ipix_disc))
m = np.arange(npix)
m = np.delete(m, ipix_disc, axis=None)
# Define the seen pixels
seenpix = ipix_disc
### Making mask - it will be automaticall apodized when instanciating the object with default (tunable) parameters
mask = np.zeros(npix)
mask[seenpix] = 1
Namaster = nam.Namaster(mask, lmin=lmin, lmax=lmax, delta_ell=delta_ell)
if plot_results:
if not os.path.isdir(path_):
os.mkdir(path_)
plt.figure()
hp.mollview(mask)
figname = 'fig_map_partial_mask.png'
dest = os.path.join(path_, figname)
plt.savefig(dest) # write image to file
plt.clf()
plt.figure()
mask_apo = Namaster.mask_apo
hp.mollview(mask_apo)
figname = 'fig_map_partial_mask_apo.png'
dest = os.path.join(path_, figname)
plt.savefig(dest) # write image to file
plt.clf()
ell_binned, b = Namaster.get_binning(nside)
# Get binned input spectra
cl_theo_binned = np.zeros(shape=(4, ell_binned.shape[0]))
for i in range(4):
cl_theo_binned[i, :] = Namaster.bin_spectra(cl.T[i, :], nside)
print(cl_theo_binned.shape)
# Create map
map_ = hp.synfast(cl.T, nside, pixwin=False, verbose=False, new = True)
noise = np.random.randn(npix)*noise_rms
map_ = map_ + noise
# constructing partial map
map_partial = map_.copy()
# Set UNSEEN pixels to zero for NaMaster
map_partial[:, m] = 0
# Get spectra
leff, dells, w = Namaster.get_spectra(map_partial,
purify_e=True,
purify_b=False,
beam_correction=None,
pixwin_correction=None,
verbose=False)
dells = dells.T
if plot_results:
clnames = ['TT', 'EE', 'BB', 'TE']
ll = np.arange(cl.shape[0])
#rc('figure', figsize=(12, 8))
plt.figure(figsize=(12, 8))
for i in range(dells.shape[0]):
plt.subplot(2, 2, i+1)
plt.plot(ll, ll * (ll + 1) * cl[:, i] / (2*np.pi),label="Input spectra")
plt.plot(leff, leff * (leff + 1) * cl_theo_binned[i, :] / (2*np.pi), "o", label="Binned input spectra")
plt.plot(leff, dells[i], label="NaMaster")
plt.xlabel('$\\ell$')
plt.ylabel('$D_\\ell$')
plt.title(clnames[i])
plt.legend()
plt.tight_layout()
figname = 'fig_theor_namaster_all_Dl.png'
dest = os.path.join(path_, figname)
plt.savefig(dest) # write image to file
plt.clf()
# Anafast spectrum of this map
# Set UNSEEN pixels to hp.UNSEEN for Anafast
map_partial[:, m] = hp.UNSEEN
cl_ana, alm_ana = hp.anafast(map_partial, alm=True, lmax=lmax)
# Get binned anafast spectra
cl_ana_binned = np.zeros(shape=(4, ell_binned.shape[0]))
for i in range(4):
cl_ana_binned[i, :] = Namaster.bin_spectra(cl_ana[i, :], nside)
if plot_results:
mapnames = ['T', 'Q', 'U']
for i in range(3):
plt.figure()
hp.mollview(map_partial[i])
figname = 'fig_map_partial_{}.png'.format(mapnames[i])
dest = os.path.join(path_, figname)
plt.savefig(dest) # write image to file
plt.clf()
for i in range(4):
# Plot theoretical and anafast spectra
plt.figure()
plt.plot(ll, ll * (ll + 1) * cl_ana[i][:lmax+1], label="Anafast")
plt.plot(leff, leff * (leff + 1) * cl_ana_binned[i], "o", label="Anafast binned")
plt.plot(ll, ll * (ll + 1) * cl.T[i], 'r', label="Input spectra")
plt.xlim(0, max(ll))
plt.title(clnames[i])
plt.ylabel(r"$\ell (\ell + 1) C_{\ell}^{"+clnames[i]+r"}$")
plt.legend()
figname = 'fig_theor_anaf_{}.png'.format(clnames[i])
dest = os.path.join(path_, figname)
plt.savefig(dest) # write image to file
plt.clf()
return dells, alm_ana, cl_theo_binned, cl_ana_binned
def explore_like(leff,
mcl_noise,
errors,
lmin,
dl,
cc,
rv,
otherp=None,
cov=None,
plotlike=False,
plotcls=False,
verbose=False,
sample_variance=True,
mytitle='',
color=None,
mylabel='',
my_ylim=None):
# print(lmin, dl, cc)
# print(leff)
# print(scl_noise[:,2])
### Create Namaster Object
# Unfortunately we need to recalculate fsky for calculating sample variance
nside = 256
lmax = 2 * nside - 1
if cov is None:
Namaster = nam.Namaster(None, lmin=lmin, lmax=lmax, delta_ell=dl)
Namaster.fsky = 0.018
else:
okpix = cov > (np.max(cov) * float(cc))
maskpix = np.zeros(12 * nside**2)
maskpix[okpix] = 1
Namaster = nam.Namaster(maskpix, lmin=lmin, lmax=lmax, delta_ell=dl)
# print('Fsky: {}'.format(Namaster.fsky))
lbinned, b = Namaster.get_binning(nside)
### Bibnning CambLib
# binned_camblib = qc.bin_camblib(Namaster, '../../scripts/QubicGeneralPaper2020/camblib.pickle',
# nside, verbose=False)
binned_camblib = qc.bin_camblib(Namaster,
global_dir + '/doc/CAMB/camblib.pkl',
nside,
verbose=False)
### Redefine the function for getting binned Cls
def myclth(ell, r):
clth = qc.get_Dl_fromlib(ell, r, lib=binned_camblib, unlensed=False)[0]
return clth
allfakedata = myclth(leff, 0.)
### And we need a fast one for BB only as well
def myBBth(ell, r):
clBB = qc.get_Dl_fromlib(ell,
r,
lib=binned_camblib,
unlensed=False,
specindex=2)[0]
return clBB
### Fake data
fakedata = myBBth(leff, 0.)
if sample_variance:
covariance_model_funct = Namaster.knox_covariance
else:
covariance_model_funct = None
if otherp is None:
like, cumint, allrlim = ana_likelihood(
rv,
leff,
fakedata,
errors,
myBBth, [[0, 1]],
covariance_model_funct=covariance_model_funct)
else:
like, cumint, allrlim, other = ana_likelihood(
rv,
leff,
fakedata,
errors,
myBBth, [[0, 1]],
covariance_model_funct=covariance_model_funct,
otherp=otherp)
if plotcls:
if plotlike:
subplot(1, 2, 1)
if np.ndim(BBcov) == 2:
errorstoplot = np.sqrt(np.diag(errors))
else:
errorstoplot = errors
#plot(inputl, inputcl[:,2], 'k', label='r=0')
plot(leff, errorstoplot, label=mylabel + ' Errors', color=color)
xlim(0, lmax)
if my_ylim is None:
ylim(1e-4, 1e0)
else:
ylim(my_ylim[0], my_ylim[1])
yscale('log')
xlabel('$\\ell$')
ylabel('$D_\\ell$')
legend(loc='upper left')
if plotlike:
if plotcls:
subplot(1, 2, 2)
p = plot(rv,
like / np.max(like),
label=mylabel + ' $\sigma(r)={0:6.4f}$'.format(allrlim),
color=color)
plot(allrlim + np.zeros(2), [0, 1.2], ':', color=p[0].get_color())
xlabel('r')
ylabel('posterior')
legend(fontsize=8, loc='upper right')
xlim(0, 0.1)
ylim(0, 1.2)
title(mytitle)
if otherp is None:
return like, cumint, allrlim
else:
return like, cumint, allrlim, other