def nongaussianity(nside,
simset,
cosmology,
instrument,
fitting_model,
power,
lkl,
nmc=50):
""" Function calculating skewness and kurtosis of a given setup.
"""
nong = pebbles.Nongaussianity(fitting_model,
nside,
simset,
cosmology,
instrument,
nmc=nmc)
obs, var = nong.cmb_and_noise_simulations()
thresholds = np.linspace(0.01, 0.5, 10)
means = []
stds = []
means_res = []
stds_res = []
res = pebbles.Pebbles(nside, simset, cosmology, instrument, nmc=nmc)
for threshold in thresholds:
statistics_arr = []
statistics_arr_res = []
for i, (ob, noiseivar) in enumerate(zip(obs, var)):
q = res.load_cleaned_amp_maps(fitting_model, 'cmb', 'q', i)
u = res.load_cleaned_amp_maps(fitting_model, 'cmb', 'u', i)
cmb_noise_sim = nong.scaled_noise_maps((ob, noiseivar), [-3., 1.5])
arr_res = pebbles.nongaussianity.apply_mask(
np.array((q, u)), so_mask_fitting(nside, threshold))
arr = pebbles.nongaussianity.apply_mask(
cmb_noise_sim, so_mask_fitting(nside, threshold))
sk = pebbles.nongaussianity.stats(arr)
sk_res = pebbles.nongaussianity.stats(arr_res)
statistics_arr.append(sk)
statistics_arr_res.append(sk_res)
statistics = np.array(statistics_arr)
statistics_res = np.array(statistics_arr_res)
means.append(np.mean(statistics, axis=0))
stds.append(np.std(statistics, axis=0))
means_res.append(np.mean(statistics_res, axis=0))
stds_res.append(np.std(statistics_res, axis=0))
means_res = np.array(means_res)
stds_res = np.array(stds_res)
means = np.array(means)
stds = np.array(stds)
fig, (ax1, ax2) = plt.subplots(2, 1)
ax1.set_title('Skewness')
l1, = ax1.plot(thresholds, means[:, 0], label='CMB + Noise sims')
ax1.fill_between(thresholds,
means[:, 0] - stds[:, 0],
means[:, 0] + stds[:, 0],
color=l1.get_color(),
alpha=0.5)
l2, = ax1.plot(thresholds, means_res[:, 0], label='Cleaned full sky sims')
ax1.fill_between(thresholds,
means_res[:, 0] - stds_res[:, 0],
means_res[:, 0] + stds_res[:, 0],
color=l2.get_color(),
alpha=0.5)
ax1.legend()
ax2.set_title('Kurtosis')
l3, = ax2.plot(thresholds, means[:, 1])
ax2.fill_between(thresholds,
means[:, 1] - stds[:, 1],
means[:, 1] + stds[:, 1],
color=l3.get_color(),
alpha=0.5)
l4, = ax2.plot(thresholds, means_res[:, 1])
ax2.fill_between(thresholds,
means_res[:, 1] - stds_res[:, 1],
means_res[:, 1] + stds_res[:, 1],
color=l4.get_color(),
alpha=0.5)
ax2.legend()
fig.savefig("{:04d}_{:s}_{:s}_{:s}_{:s}_nmc{:04d}.png".format(
nside, simset, cosmology, instrument, fitting_model, nmc))
return
class ThetaFormatterShiftPi(GeoAxes.ThetaFormatter):
"""Shifts labelling by pi
Shifts labelling from -180,180 to 0-360"""
def __call__(self, x, pos=None):
if x != 0:
x *= -1
if x < 0:
x += 2*np.pi
return GeoAxes.ThetaFormatter.__call__(self, x, pos)
width = 17
cmap = plt.cm.RdYlBu
colormaptag = "colombi1_"
peb = pebbles.Pebbles(nside, 'simset1', 'planck2015_AL0p5', 'SO_21', nmc=2)
(data) = np.moveaxis(peb.compute_simulated_data()[0, 0, 0, ...], 0, 1)
fig = plt.figure(figsize=(cm2inch(width), cm2inch(width)*.6))
norm = SymLogNorm(linthresh, 1.)
figure_rows, figure_columns = 2, 3
subplot_titles = ["27 GHz", "39 GHz", "93 GHz", "145 GHz", "225 GHz", "280 GHz"]
for i, (submap, s_title) in enumerate(zip(data, subplot_titles)):
# matplotlib is doing the mollveide projection
submap = pebbles.plotting.apply_so_mask(submap)
ax = plt.subplot(figure_rows, figure_columns, i+1, projection='mollweide')
# rasterized makes the map bitmap while the labels remain vectorial
# flip longitude to the astro convention
image = plt.pcolormesh(longitude[::-1], latitude, submap[grid_pix], vmin=vmin, vmax=vmax, rasterized=True, cmap=cmap, norm=norm)
arp.add_argument('-fit', action='store_true')
arp.add_argument('-power', action='store_true')
arp.add_argument('-sample', action='store_true')
args = arp.parse_args()
jobs = pebbles.read_config_file(args.configuration)
# set number of threads for running on cori
os.environ['OMP_NUM_THREADS'] = str(multiprocessing.cpu_count())
return (args.nside, args.nmc, jobs, args)
if __name__ == '__main__':
(NSIDE, NMC, JOBS, STEPS) = setup_run()
if STEPS.sim:
for sim, cos, ins in JOBS('SIMSETS', 'COSMOLOGIES', 'INSTRUMENTS'):
peb = pebbles.Pebbles(NSIDE, sim, cos, ins, nmc=NMC)
peb.compute_simulated_data()
if STEPS.fit:
for sim, cos, ins, fit in JOBS('SIMSETS', 'COSMOLOGIES', 'INSTRUMENTS',
'FITTING_MODELS'):
peb = pebbles.Pebbles(NSIDE, sim, cos, ins, nmc=NMC)
data = peb.load_simulated_data()
with JoblibPool(multiprocessing.cpu_count()) as pool:
peb.clean_simulated_data(data, fit, pool)
if STEPS.power:
for sim, cos, ins, fit, pwr in JOBS('SIMSETS', 'COSMOLOGIES',
'INSTRUMENTS', 'FITTING_MODELS',
'POWERS'):
powerspectra = pebbles.PowerSpectra(pwr,
simset = 'simset0'
if simset == 'simset1':
s1 = pysm.nominal.models("s1", nside_spec)
d1 = pysm.nominal.models("d1", nside_spec)
beta_d = d1[0]['spectral_index'][ipix]
T_d = d1[0]['temp'][ipix]
beta_s = s1[0]['spectral_index'][ipix]
else:
beta_d = 1.54
T_d = 20.
beta_s = -3
if sys.argv[1] == 'sample':
peb = pebbles.Pebbles(nside=nside, nside_spec=nside_spec,
gal_tag=simset, ins_tag='sofidV3_wn_smo',
cos_tag='planck2015', mask=mask)
try:
data = peb.load_simulated_data()
except FileNotFoundError:
data = peb.compute_simulated_data()
tasks = peb._prepare_simulated_data(data, [beta_s, beta_d, T_d])
# get hm1 settings for sampling this pixel
_, _, _, maplike, pos0, mc_fpath, _ = tasks[0]
nwalkers = 20
ndim = 3
max_n = 7000
nburn = 2500
# initial positions of the walkers
pos = [pos0 + 0.01 * np.random.randn(ndim) for i in range(nwalkers)]
# initiate emcee sample