def drop_vs_dnu_nomu(fmax=61, fstep=0.1, drops=[0, 1, 2]):
dnu = np.arange(1, fmax, fstep) * 1.e9
fish = fisher.FisherEstimation()
fish.set_signals(fncs=[
sd.DeltaI_reltSZ_2param_yweight, sd.DeltaI_DeltaT, fg.thermal_dust_rad,
fg.cib_rad, fg.jens_freefree_rad, fg.jens_synch_rad, fg.spinning_dust,
fg.co_rad
])
args = fish.args
N = len(args)
data = {}
for drop in drops:
print "on drop ", drop
data[drop] = {}
for arg in args:
data[drop][arg] = []
for k, nu in enumerate(dnu):
if k % 100 == 0:
print "on k ", k
scale = 15.e9 / nu
fish = fisher.FisherEstimation(fstep=nu, mult=scale, drop=drop)
fish.set_signals(fncs=[
sd.DeltaI_reltSZ_2param_yweight, sd.DeltaI_DeltaT,
fg.thermal_dust_rad, fg.cib_rad, fg.jens_freefree_rad,
fg.jens_synch_rad, fg.spinning_dust, fg.co_rad
])
fish.run_fisher_calculation()
for arg in args:
data[drop][arg].append(fish.errors[arg])
x = {}
x['drops'] = drops
x['dnu'] = dnu
x['data'] = data
np.save('fullcalc_10p_drop012_nomu', x)
return
def drop_vs_dnu(fmax=61, fstep=0.1, drops=[0, 1, 2]):
dnu = np.arange(1, fmax, fstep) * 1.e9
fish = fisher.FisherEstimation()
args = fish.args
N = len(args)
data = {}
for drop in drops:
print "on drop ", drop
data[drop] = {}
for arg in args:
data[drop][arg] = []
for k, nu in enumerate(dnu):
if k % 100 == 0:
print "on k ", k
scale = 15.e9 / nu
ps = {'As': 0.01, 'alps': 0.01}
#ps = {'As':0.1, 'alps': 0.1}
fish = fisher.FisherEstimation(fstep=nu,
mult=scale,
drop=drop,
priors=ps)
fish.run_fisher_calculation()
for arg in args:
data[drop][arg].append(fish.errors[arg])
x = {}
x['drops'] = drops
x['dnu'] = dnu
x['data'] = data
np.save('fullcalc_1p_drop012_coarse', x)
return
def sens_vs_nbins(sens=[1., 0.1, 0.01, 0.001]):
nbins = np.arange(50, 601)[::-1]
dnu = 3.e12 / nbins
fish = fisher.FisherEstimation()
args = fish.args
N = len(args)
data = {}
for sen in sens:
print "on sens ", sen
data[sen] = {}
for arg in args:
data[sen][arg] = []
for nu in dnu:
scale = sen * (15.e9 / nu)
#ps = {'As':0.1, 'alps': 0.1}
#ps = {'As':0.01, 'alps': 0.01}
ps = {}
fish = fisher.FisherEstimation(fstep=nu, mult=scale, priors=ps)
fish.run_fisher_calculation()
for arg in args:
data[sen][arg].append(fish.errors[arg])
x = {}
x['sens'] = sens
x['dnu'] = dnu
x['data'] = data
np.save('senscalc_nbins_0p', x)
return
def drop_vs_nbin(drops=[0, 1, 2]):
nbins = np.arange(50, 601)[::-1]
dnu = 3.e12 / nbins
fish = fisher.FisherEstimation()
args = fish.args
N = len(args)
data = {}
for drop in drops:
print "on drop ", drop
data[drop] = {}
for arg in args:
data[drop][arg] = []
for k, nu in enumerate(dnu):
if k % 100 == 0:
print "on k ", k
scale = 15.e9 / nu
#ps = {'As':0.01, 'alps': 0.01}
ps = {'As': 0.1, 'alps': 0.1}
fish = fisher.FisherEstimation(fstep=nu,
mult=scale,
drop=drop,
priors=ps)
fish.run_fisher_calculation()
for arg in args:
data[drop][arg].append(fish.errors[arg])
x = {}
x['drops'] = drops
x['dnu'] = dnu
x['data'] = data
np.save('fullcalc_10p_drop012_nbins_onefifthsynch', x)
return
def run():
pool = mp.Pool()
nargs = ['mu', 'y', 'kT', 'DeltaT']
fish = fisher.FisherEstimation(duration=12., bandpass=False)
fncs = [sd.DeltaI_mu, sd.DeltaI_reltSZ_2param_yweight, sd.DeltaI_DeltaT]
fish.set_signals(fncs)
x = fish.center_frequencies
noise = fish.noise / 1000.
y = sd.DeltaI_mu(x) + sd.DeltaI_reltSZ_2param_yweight(
x) + sd.DeltaI_DeltaT(x)
yerr = noise * np.random.randn(len(x))
y += yerr
ndim, nwalkers = 4, 32
pos = [
fish.p0 * (1. + 1.e-1 * np.random.randn(ndim)) for i in range(nwalkers)
]
sampler = emcee.EnsembleSampler(nwalkers,
ndim,
lnprob,
args=(x, y, noise),
pool=pool)
sampler.run_mcmc(pos, 2000)
pool.close()
samples = sampler.chain[:, 500:, :].reshape((-1, ndim))
p_mc = map(lambda v: (v[1], v[2] - v[1], v[1] - v[0]),
zip(*np.percentile(samples, [16, 50, 84], axis=0)))
for k in range(4):
print nargs[k], (p_mc[k][1] + p_mc[k][2]) / 2.
return
def __init__(fncs=None, duration=86.4, nwalkers=0, gf=1.e-3, threads=16):
if fncs is None:
print "need funcs"
return
fish = fisher.FisherEstimation(duration=duration, bandpass=False)
self.freqs = fish.center_frequencies
self.noise = fish.noise
self.fncs = fncs
self.setup_fncs()
self.generate_data()
self.setup_mcmc(nwalkers, gf, threads)
def lnlike(theta):
fs, ns = theta[:200], theta[200:]
fish = fisher.FisherEstimation(bandpass=False, priors={})
fish.set_signals([sd.DeltaI_mu, sd.DeltaI_reltSZ_2param_yweight, sd.DeltaI_DeltaT, \
fg.thermal_dust_rad, fg.cib_rad, fg.jens_freefree_rad, fg.jens_synch_rad])
fish.center_frequencies = fs
fish.noise = ns
fish.run_fisher_calculation()
if np.isnan(fish.errors['mu_amp']):
return -np.inf
return -(fish.errors['mu_amp'] / 2.e-8 - 1. / 3.)**2
def sens_vs_dnu(fmax=61, fstep=0.1, sens=[1., 0.1, 0.01, 0.001]):
dnu = np.arange(1, fmax, fstep) * 1.e9
fish = fisher.FisherEstimation()
args = fish.args
N = len(args)
data = {}
for sen in sens:
print "on sens ", sen
data[sen] = {}
for arg in args:
data[sen][arg] = []
for nu in dnu:
scale = sen * (15.e9 / nu)
fish = fisher.FisherEstimation(fstep=nu, mult=scale)
fish.run_fisher_calculation()
for arg in args:
data[sen][arg].append(fish.errors[arg])
x = {}
x['sens'] = sens
x['dnu'] = dnu
x['data'] = data
np.save('datatest', x)
return
return np.inf
if np.any(fs > 3.01e12):
return np.inf
if np.any(np.diff(fs) < 1e9):
return np.inf
if np.any(ns < 1.e-2):
return np.inf
if np.any(ns > 1.e2):
return np.inf
return 0.
def lnprob(theta):
lp = lnprior(theta)
if not np.isfinite(lp):
return -np.inf
return lp + lnlike(theta)
fish = fisher.FisherEstimation(bandpass=False, priors={})
fish.set_signals([sd.DeltaI_mu, sd.DeltaI_reltSZ_2param_yweight, sd.DeltaI_DeltaT, \
fg.thermal_dust_rad, fg.cib_rad, fg.jens_freefree_rad, fg.jens_synch_rad])
fs0 = fish.center_frequencies
noise0 = fish.noise / 10.
p0 = np.concatenate([fs0, noise0])
ndim, nwalkers = 400, 1024
pos = [p0 + 1.e-3 * np.random.randn(ndim) for i in range(nwalkers)]
sampler = emcee.EnsembleSampler(nwalkers, ndim, lnprob)
sampler.run_mcmc(pos, 64)
np.savez('optimize1', sampler)
import fisher
import foregrounds as fg
import spectral_distortions as sd
# PIXIE with all signals is the default.
fish = fisher.FisherEstimation()
# args are stored in fish.args, values are stored in dictionary fish.argvals,
# and fisher uncertainties are in fish.errors
fish.run_fisher_calculation()
# print the errors in sigma
fish.print_errors()
# To set the signals by hand, just modify the fncs arg here:
fish = fisher.FisherEstimation()
fish.set_signals(fncs=[sd.DeltaI_reltSZ_2param_yweight, sd.DeltaI_DeltaT, sd.DeltaI_mu,
fg.thermal_dust_rad, fg.cib_rad, fg.jens_freefree_rad,
fg.jens_synch_rad, fg.spinning_dust, fg.co_rad])
fish.run_fisher_calculation()
fish.print_errors()
# To change the frequencies (Hz), duration (months), or scale the noise by mult,
# turn off the step function bandpass or change fsky,
# edit any of the following
fish = fisher.FisherEstimation(fmin=5.e9, fmax=1.e12, fstep=5.e9, duration=60, mult=0.1, bandpass=False, fsky=0.5)
# Lastly to put priors (in fractions of the parameter value), drop the first n bins or mask out Galactic CO lines do:
fish = fisher.FisherEstimation(priors={'Td':0.1, 'Asd':0.01}, drop=2, doCO=True)