def plot_compare_optimalspectrum(fname):
spec_pick = "EE"
def _weightspec(icov_l, spectrum):
import copy
# retspec = copy.deepcopy(spectrum)
for specc, data in spectrum.items():
if specc == spec_pick:
icovsum = np.array(
[np.sum(icov_l[specc][l]) for l in range(lmax)])
# buff += spectrum[specc][freqc][:lmax] * spectrum[specc][freqc][:lmax] * weights[specc]["channel @{}GHz".format(freqs[0])].to_numpy()[:lmax] * weights[specc]["channel @{}GHz".format(freqs[0])].to_numpy()[:lmax]/(normaliser * normaliser)
retspec = {
specc: {
'optimal-optimal':
np.array([
1 / icovsum_l if icovsum_l is not None else 0
for icovsum_l in icovsum
])
}
}
return retspec
fname = io.make_filenamestring(cf)
lmax = cf['pa']['lmax']
freqcomb = [
"{}-{}".format(FREQ, FREQ2) for FREQ in PLANCKMAPFREQ
if FREQ not in freqfilter for FREQ2 in PLANCKMAPFREQ
if (FREQ2 not in freqfilter) and (int(FREQ2) >= int(FREQ))
]
dc = dcf["plot"]["spectrum"]
def _inpathname(freqc, spec):
return dc["indir_root"] + dc["indir_rel"] + spec + freqc + "-" + dc[
"in_desc"] + fname
speccs = [spec for spec in PLANCKSPECTRUM if spec not in specfilter]
spectrum = {
freqc: {
spec: np.array(io.load_cl(_inpathname(freqc, spec)))
for spec in speccs
}
for freqc in freqcomb
}
cov = pw.build_covmatrices(spectrum,
lmax=lmax,
freqfilter=freqfilter,
specfilter=specfilter)
icov_l = pw.invert_covmatrices(cov,
lmax=lmax,
freqfilter=freqfilter,
specfilter=specfilter)
spec_data = hpf.reorder_spectrum_dict(spectrum)
spec_data_wweighted = _weightspec(icov_l, spec_data)
spectrum_truth = io.load_truthspectrum(
abspath='/mnt/c/Users/sebas/OneDrive/Desktop/Uni/')
for specc, data in spec_data_wweighted.items():
title_string = "{} spectrum - {}".format(specc,
"optimal spectrum included")
if "Planck-" + specc in spectrum_truth.columns:
spectrum_trth = spectrum_truth["Planck-" + specc]
else:
spectrum_trth = None
mp = cplt.plot_powspec_binned(data,
lmax,
title_string=title_string,
truthfile=spectrum_trth,
truth_label="Planck-" + specc)
outpath_name = \
dc["outdir_root"] + \
dc["outdir_rel"] + \
specc+"_spectrum/" + \
specc+"_spectrum" + "-" + \
"withoptimal" + "-" + \
fname + ".jpg"
io.save_figure(mp=mp, path_name=outpath_name)
print("spectrum saved to {}".format(outpath_name))
def plot_variance():
ylim = {
"TT": (1e2, 1e5),
"EE": (1e-10, 1e-2),
"BB": (1e-5, 1e5),
"TE": (1e-2, 1e4),
"TB": (-1e-3, 1e3),
"EB": (-1e-3, 1e3),
"ET": (1e-2, 1e5),
"BT": (-1e-3, 1e3),
"BE": (-1e-3, 1e3)
}
freqcomb = [
"{}-{}".format(FREQ, FREQ2) for FREQ in PLANCKMAPFREQ
if FREQ not in freqfilter for FREQ2 in PLANCKMAPFREQ
if (FREQ2 not in freqfilter) and (int(FREQ2) >= int(FREQ))
]
dc = dcf["plot"]["spectrum"]
def _inpathname(freqc, spec):
return dc["indir_root"] + dc["indir_rel"] + spec + freqc + "-" + dc[
"in_desc"] + fname
speccs = [spec for spec in PLANCKSPECTRUM if spec not in specfilter]
# spec_data = {spec: {
# freqc: np.array(io.load_cl(_inpathname(freqc,spec)))
# for freqc in freqcomb} for
# spec in speccs}
spec_pick = "EE"
def _weightspec(icov_l, spectrum):
import copy
# retspec = copy.deepcopy(spectrum)
for specc, data in spectrum.items():
if specc == spec_pick:
icovsum = np.array(
[np.sum(icov_l[specc][l]) for l in range(lmax)])
# buff += spectrum[specc][freqc][:lmax] * spectrum[specc][freqc][:lmax] * weights[specc]["channel @{}GHz".format(freqs[0])].to_numpy()[:lmax] * weights[specc]["channel @{}GHz".format(freqs[0])].to_numpy()[:lmax]/(normaliser * normaliser)
retspec = {
specc: {
'optimal-optimal':
np.array([
1 / icovsum_l if icovsum_l is not None else 0
for icovsum_l in icovsum
])
}
}
return retspec
spectrum = {
freqc: {
spec: np.array(io.load_cl(_inpathname(freqc, spec)))
for spec in speccs
}
for freqc in freqcomb
}
lmax = cf['pa']['lmax']
npatch = 1
ll = np.arange(0, lmax, 1)
# fsky = np.zeros((npatch, npatch), float)
# np.fill_diagonal(fsky, 1/npatch*0.73)#*np.ones((npatch)))
cov = pw.build_covmatrices(spectrum,
lmax=lmax,
freqfilter=freqfilter,
specfilter=specfilter)
spec_data = hpf.reorder_spectrum_dict(spectrum)
spec_data_wweighted = _weightspec(icov_l, spec_data)
def dl2cl(dl):
return dl / (ll * (ll + 1)) * 2 * np.pi
def cl2var(cov_ltot_min, fsky):
return 2 * cov_ltot_min * cov_ltot_min / ((2 * ll + 1) * fsky)
var_data = {
spec: {
freqc: dl2cl(spec_data_wweighted[spec][freqc])
for freqc in ['optimal-optimal']
}
for spec in [spec_pick]
}
var_data[spec_pick].update({
'-2cl**2/nmode-':
cl2var(dl2cl(spec_data_wweighted[spec_pick]['optimal-optimal']), 0.96)
})
plotsubtitle = '{freqdset}"{split}" dataset - {mskset} masks'.format(
mskset=mskset, freqdset=freqdset, split=split)
spectrum_truth = io.load_truthspectrum(
abspath='/mnt/c/Users/sebas/OneDrive/Desktop/Uni/')
for specc, data in var_data.items():
title_string = "{} variance - {}".format(specc, plotsubtitle)
if "Planck-" + specc in spectrum_truth.columns:
pass
spectrum_trth = spectrum_truth["Planck-" + specc]
else:
spectrum_trth = None
mp = cplt.plot_variance_binned(data,
lmax,
title_string=title_string,
ylim=ylim[specc],
truthfile=dl2cl(spectrum_trth[:lmax]),
truth_label="Planck-" + specc)
outpath_name = \
dc["outdir_root"] + \
dc["outdir_rel"] + \
specc+"_spectrum/" + \
specc+"_variance" + "-" + \
dc["out_desc"] + "-" + \
fname + ".jpg"
io.save_figure(mp=mp, path_name=outpath_name)
def plot_spectrum_comparison(fname):
"""
Creates a plot with two diagrams, comparing two sets of spectra with one another.
This can be used to validate_spectra and validate_smica, and just compare two spectra, i.e. compare,
1. synspectra to spectra
2. smica spectra to input spectra
3. two different spectra (e.g. varying in masks)
"""
name1 = "SPEC" + io.make_filenamestring(cf)
path1 = io.out_spec_path
pathname1 = path1 + name1
cf2 = copy.deepcopy(cf)
cf2['pa']['smoothing_window'] = 0
cf2['pa']['max_polynom'] = 0
name2 = "SPEC" + io.make_filenamestring(cf2)
path2 = io.out_spec_path
pathname2 = path2 + name2
spectrum1 = io.load_data(path_name=pathname1)
spectrum2 = io.load_data(path_name=pathname2)
spectrum1_re = hpf.reorder_spectrum_dict(spectrum1)
spectrum2_re = hpf.reorder_spectrum_dict(spectrum2)
lmax = cf['pa']['lmax']
plotsubtitle = '{freqdset}"{split}" dataset - {mskset} masks'.format(
mskset=mskset,
freqdset=freqdset,
split="Full"
if cf['pa']["freqdatsplit"] == "" else cf['pa']["freqdatsplit"])
diff_spectrum = dict()
for specc, va in spectrum1_re.items():
for freqc, val in spectrum1_re[specc].items():
if specc not in diff_spectrum.keys():
diff_spectrum.update({specc: {}})
diff_spectrum[specc].update({
freqc:
(spectrum1_re[specc][freqc] - spectrum2_re[specc][freqc]) /
spectrum1_re[specc][freqc]
})
spectrum_truth = io.load_truthspectrum()
for specc, diff_data in diff_spectrum.items():
color = ['red', 'green', 'blue', 'yellow', 'black', 'orange', 'purple']
title_string = specc + "-spectrum - " + plotsubtitle
if "Planck-" + specc in spectrum_truth.columns:
spectrum_trth = spectrum_truth["Planck-" + specc][:lmax + 1] / (
hpf.llp1e12(np.array([range(lmax + 1)])))[0] * 1e12
plt.figure(figsize=(8, 6))
gs = gridspec.GridSpec(2, 1, height_ratios=[3, 1])
ax1 = plt.subplot(gs[0])
mp = cplt.plot_compare_powspec_binned(plt,
spectrum1_re[specc],
spectrum2_re[specc],
lmax,
title_string=title_string,
truthfile=spectrum_trth,
truth_label="Planck-" + specc)
ax2 = mp.subplot(gs[1])
mp = cplt.plot_powspec_diff_binned(mp,
diff_spectrum[specc],
lmax,
color=color)
outpath = "/global/cscratch1/sd/sebibel/vis/" + \
specc+"_spectrum/"
io.iff_make_dir(outpath)
outpath_name = outpath + \
specc+"_spectrumbias" + "-" + ".jpg"
io.save_figure(mp=mp, path_name=outpath_name)
def plot_tf(fname):
"""
quite similar to plot_spectrum_comparison()
"""
path1 = io.out_spec_path
name1 = "SPEC" + io.make_filenamestring(cf)
pathname1 = path1 + name1
dcf2 = copy.deepcopy(dcf)
dcf2['pa']['smoothing_window'] = 0
dcf2['pa']['max_polynom'] = 0
pathname1 = path1 + name1
spectrum1 = io.load_data(io.specsmica_sc_path_name)[0, 0, :]
spectrum2 = io.load_data(path_name=cf[mch]['outdir_misc_ap'] +
"cmb_spec_in.npy")
spectrum1_re = hpf.reorder_spectrum_dict(spectrum1)
spectrum2_re = hpf.reorder_spectrum_dict(spectrum2)
lmax = cf['pa']['lmax']
spectrum_truth = io.load_truthspectrum()
plotsubtitle = '{freqdset}"{split}" dataset - {mskset} masks'.format(
mskset=mskset,
freqdset=freqdset,
split="Full"
if cf['pa']["freqdatsplit"] == "" else cf['pa']["freqdatsplit"])
diff_spectrum = io.load_data(cf[mch]['outdir_misc_ap'] + "tf.npy")
for specc, diff_data in diff_spectrum.items():
color = ['red', 'green', 'blue', 'yellow', 'black', 'orange', 'purple']
title_string = specc + "-spectrum - " + plotsubtitle
if "Planck-" + specc in spectrum_truth.columns:
spectrum_trth = spectrum_truth["Planck-" + specc][:lmax + 1] / (
hpf.llp1e12(np.array([range(lmax + 1)])))[0] * 1e12
plt.figure(figsize=(8, 6))
gs = gridspec.GridSpec(2, 1, height_ratios=[3, 1])
ax1 = plt.subplot(gs[0])
mp = cplt.plot_compare_powspec_binned(plt,
spectrum1_re[specc],
spectrum2_re[specc],
lmax,
title_string=title_string,
truthfile=spectrum_trth,
truth_label="Planck-" + specc)
ax2 = mp.subplot(gs[1])
mp = cplt.plot_powspec_diff_binned(mp,
diff_spectrum[specc],
lmax,
color=color)
outpath = "/global/cscratch1/sd/sebibel/vis/" + \
specc+"_spectrum/"
io.iff_make_dir(outpath)
outpath_name = outpath + \
specc+"_transferfunction" + "-" + ".jpg"
io.save_figure(mp=mp, path_name=outpath_name)
def plot_spectrum(fname):
if cf['pa']["Spectrum_scale"] == "D_l":
ylim = {
"TT": (1e2, 1e5),
"EE": (1e-3, 1e6),
"BB": (1e-5, 1e5),
"TE": (1e-2, 1e4),
"TB": (-1e-3, 1e3),
"EB": (-1e-3, 1e3),
"ET": (1e-2, 1e5),
"BT": (-1e-3, 1e3),
"BE": (-1e-3, 1e3)
}
else:
ylim = {
"TT": (1e2, 1e5),
"EE": (1e-6, 1e1),
"BB": (1e-5, 1e5),
"TE": (1e-2, 1e4),
"TB": (-1e-3, 1e3),
"EB": (-1e-3, 1e3),
"ET": (1e-2, 1e5),
"BT": (-1e-3, 1e3),
"BE": (-1e-3, 1e3)
}
dc = dcf["plot"]["spectrum"]
lmax = cf['pa']['lmax']
plotsubtitle = '{freqdset}"{split}" dataset - {mskset} masks'.format(
mskset=mskset, freqdset=freqdset, split=split)
spec_data = io.load_data(io.spec_sc_path_name)
# print(spec_data)
spectrum_truth = io.load_truthspectrum()
spec_data = hpf.reorder_spectrum_dict(spec_data)
for specc, data in spec_data.items():
# for freq in PLANCKMAPFREQ:
title_string = "{} spectrum - {}".format(specc, plotsubtitle)
if "Planck-" + specc in spectrum_truth.columns:
spectrum_trth = spectrum_truth["Planck-" + specc]
if cf['pa']["Spectrum_scale"] == "D_l":
pass
else:
spectrum_trth = spectrum_trth / hpf.llp1e12(
np.array([l for l in range(len(spectrum_trth))])) * 1e12
else:
spectrum_trth = None
mp = cplt.plot_powspec_binned(
data,
lmax,
title_string=title_string,
ylim=ylim[specc],
truthfile=spectrum_trth,
truth_label="Planck-" + specc,
# filter = freq
)
outpath_name = \
dc["outdir_root"] + \
dc["outdir_rel"] + \
specc+"_spectrum/" + \
specc+"_spectrum" + "-" + \
dc["out_desc"] + "-" + \
cf['pa']["Spectrum_scale"] + "-" + \
fname + ".jpg"
io.save_figure(mp=mp, path_name=outpath_name)
print("spectrum saved to {}".format(outpath_name))