def testnoisebias(noise=0.04, niters=10000):
r_mpc = np.linspace(0.25, 3.0, 20)
beta_s = 0.5
zcluster = 0.5
g = createPerfectProfile(1e15, 4.0, zcluster, r_mpc, beta_s)
gerr = noise / np.sqrt(r_mpc)
# m200scan = np.logspace(np.log10(1e13), np.log10(5e15), 200)
m200scan = np.arange(-1e15 + 5e12, 5e15, 1e13)
config = nfwfit.readConfiguration('testnoisebias.config')
likelihood = Likelihood(config)
likelihood.bindData(r_mpc, g, gerr, beta_s, zcluster)
perfectpdf = np.array([likelihood(m200=curm200) for curm200 in m200scan])
maxs = np.zeros(niters)
for i in range(niters):
gwnoise = g + gerr * np.random.standard_normal(size=len(g))
likelihood.bindData(r_mpc, gwnoise, gerr, beta_s, zcluster)
logprob = np.array([likelihood(m200=curm200) for curm200 in m200scan])
maxs[i] = m200scan[logprob == np.max(logprob)]
return r_mpc, g, gerr, m200scan, perfectpdf, maxs
def stackCats(stackfile, configname, answerfile, outfile):
filebase = os.path.basename(answerfile)
match = re.match('(.+)_answers.pkl', filebase)
simtype = match.group(1)
with open(answerfile, 'rb') as input:
answers = cPickle.load(input)
tostack = [x[0] for x in readtxtfile.readtxtfile(stackfile)]
config = nfwfit.readConfiguration(configname)
simreader = nfwfit.buildSimReader(config)
nfwutils.global_cosmology.set_cosmology(simreader.getCosmology())
fitter = nfwfit.buildFitter(config)
profile = fitter.profileBuilder
if profile.binspacing == 'linear':
binedges = np.linspace(profile.minradii, profile.maxradii,
profile.nbins + 1)
else:
binedges = np.logspace(np.log10(profile.minradii),
np.log10(profile.maxradii), profile.nbins + 1)
stackedprofile = OnlineStatistics(binedges)
for catalogname in tostack:
filebase = os.path.basename(catalogname)
match = idpatterns[simtype].match(filebase)
haloid = int(match.group(1))
try:
truth = answers[haloid]
except KeyError:
print 'Failure at {0}'.format(output)
raise
catalog = nfwfit.readSimCatalog(catalogname, simreader, config)
stackedprofile.accumulate(catalog, truth)
stackedprofile.writeSimCat(outfile)
return stackedprofile
def run(simtype,
chaindir,
outfile,
delta,
massbin=0,
samples=__samples__,
ngauss=__ngauss__,
singlecore=__singlecore__):
outfile = '{}.gauss{}'.format(outfile, ngauss)
config = nfwfit.readConfiguration('%s/config.sh' % chaindir)
simreader = nfwfit.buildSimReader(config)
nfwutils.global_cosmology.set_cosmology(simreader.getCosmology())
massedges = rundln.defineMassEdges(simtype, delta)
halos = dln.loadPDFs(chaindir, simtype, simreader, delta, massedges,
massbin)
if len(halos) < 10:
sys.exit(0)
success = False
for i in range(20):
try:
parts = dln.buildGaussMixture1DModel(halos,
ngauss,
modeltype='ratio')
success = True
break
except (AssertionError, pymc.ZeroProbability) as e:
continue
assert (success is True)
with open('%s.massrange' % outfile, 'w') as output:
output.write('%f\n%f\n' % (massedges[massbin], massedges[massbin + 1]))
dln.sample(parts, outfile, samples, singlecore=singlecore)
def doMassUncert(argdict):
return nfwnoise.calcMassUncert(**argdict)
#####
if __name__ == '__main__':
nthreads = 4
pool = ThreadPool(nthreads)
nmasses = 10000
nperthread = nmasses / nthreads
config = nfwfit.readConfiguration('mxxl_imperial/snap41/c4_r10/config.sh')
r_mpc = np.arange(0.75, 3.0, 0.2)
beta_s = 0.5
zcluster = 0.5
masses = 10**np.random.uniform(14, 16, nmasses)
concens = np.random.uniform(1.1, 19.9, nmasses)
testmasses = np.logspace(14., 16., 400)
log10testmasses = np.log10(testmasses)
shearprofiles_p, shearerr_p = nfwnoise.createClusterSet(
config, masses, zcluster, r_mpc, beta_s, 0.01)
profile_threadgroups = [
dict(config=config,
r_mpc=r_mpc,
def plotLogProbSurface(catalogname,
configfile,
fig=None,
noiselevels=[0., 0.03, 0.07, 0.15]):
if fig is None:
fig = pylab.figure()
config = nfwfit.readConfiguration(configfile)
simreader = nfwfit.buildSimReader(config)
nfwutils.global_cosmology.set_cosmology(simreader.getCosmology())
catalog = nfwfit.readSimCatalog(catalogname, simreader, config)
fitter = nfwfit.buildFitter(config)
r_mpc, ghat, sigma_ghat, beta_s, beta_s2, zlens = fitter.prepData(
catalog, config)
print sigma_ghat
fitter.model.setData(beta_s, beta_s2, zlens)
guess = fitter.model.guess()
massgrid = np.arange(1e14, 1e15, 2.5e13) / fitter.model.massScale
masscenters = (massgrid[1:] + massgrid[:-1]) / 2.
for noise in noiselevels:
noisy_g = ghat + noise * np.random.standard_normal(len(ghat))
noisy_sigma = np.sqrt(sigma_ghat**2 + noise**2)
if len(guess) == 2:
concengrid = np.arange(1., 15., 0.25)
concencenters = (concengrid[1:] + concengrid[:-1]) / 2.
chisqgrid = np.zeros((len(masscenters), len(concencenters)))
for i in range(len(masscenters)):
for j in range(len(concencenters)):
chisqgrid[i, j] = fitmodel.ChiSqStat(
noisy_g, noisy_sigma,
fitter.model(r_mpc, masscenters[i], concencenters[j]))
probgrid = np.exp(-0.5 * (chisqgrid - np.max(chisqgrid)))
massprobgrid = np.sum(probgrid, axis=1) / np.sum(probgrid)
logprob = np.log(massprobgrid)
else:
chisqgrid = np.zeros(len(masscenters))
for i in range(len(massgrid)):
chisqgrid[i] = fitmodel.ChiSqStat(
noisy_g, noisy_sigma, fitter.model(r_mpc, masscenters[i]))
logprob = -0.5 * chisqgrid
ax = pylab.gca()
ax.plot(masscenters, np.exp(logprob - np.max(logprob)), label=noise)
print 'Max: {0}'.format(masscenters[np.argmax(logprob)])
pylab.legend()
return fig, massgrid, concengrid, chisqgrid, logprob
def consolidateFits(workdir, simtype, outdir):
failfile = open('{0}/fails'.format(outdir), 'w')
idpattern = idpatterns[simtype]
answers = cPickle.load(
open('{0}/{1}_answers.pkl'.format(workdir, simtype), 'rb'))
outputfiles = glob.glob('%s/*.out' % outdir)
nhalos = len(outputfiles)
ids = []
measured_m200s = np.zeros(nhalos)
measured_m200errs = np.zeros(nhalos)
measured_m500s = np.zeros(nhalos)
measured_m500errs = np.zeros(nhalos)
measured_cs = np.zeros(nhalos)
measured_rs = np.zeros(nhalos)
true_m200s = np.zeros(nhalos)
true_m500s = np.zeros(nhalos)
true_cs = np.zeros(nhalos)
redshifts = np.zeros(nhalos)
results = dict(ids=ids,
measured_m200s=measured_m200s,
measured_m200errs=measured_m200errs,
measured_m500s=measured_m500s,
measured_m500errss=measured_m500errs,
measured_cs=measured_cs,
measured_rs=measured_rs,
true_m200s=true_m200s,
true_m500s=true_m500s,
true_cs=true_cs,
redshifts=redshifts)
class WeirdException(Exception):
pass
#load up the environment for cosmology, and mc relation if used
config = nfwfit.readConfiguration('{0}/config.sh'.format(outdir))
simreader = nfwfit.buildSimReader(config)
nfwutils.global_cosmology.set_cosmology(simreader.getCosmology())
fitter = nfwfit.buildFitter(config)
configname = os.path.basename(outdir)
for i, output in enumerate(outputfiles):
filebase = os.path.basename(output)
match = idpattern.match(filebase)
try:
haloid = int(match.group(1))
except AttributeError as e:
print filebase
raise e
except ValueError:
haloid = match.group(1)
try:
truth = answers[haloid]
except KeyError:
print 'Failure at {0}'.format(output)
raise
ids.append(haloid)
true_m200s[i] = truth['m200']
true_m500s[i] = truth['m500']
true_cs[i] = truth['concen']
redshifts[i] = truth['redshift']
input = open(output)
measured = cPickle.load(input)
input.close()
if measured is None:
print 'Fail {0} {1}'.format(configname, haloid)
failfile.write('Fail {0} {1}\n'.format(configname, haloid))
continue
measured_m200s[i] = measured[0][
'm200'] * fitter.model.massScale * nfwutils.global_cosmology.h
measured_m200errs[i] = np.mean(np.abs(
measured[1]
['m200'])) * fitter.model.massScale * nfwutils.global_cosmology.h
if 'c200' in measured:
measured_cs[i] = measured['c200']
else:
## need to dig up the mc relation
measured_cs[i] = fitter.model.massconRelation(
np.abs(measured_m200s[i]), redshifts[i],
fitter.model.overdensity)
#####
#calculate m500
measured_rs[i] = nfwutils.rscaleConstM(np.abs(measured_m200s[i]),
measured_cs[i], redshifts[i],
fitter.model.overdensity)
measured_m500s[i] = nfwutils.Mdelta(measured_rs[i], measured_cs[i],
redshifts[i], 500)
if measured_m200s[i] < 0:
measured_m500s[i] = -measured_m500s[i]
if not np.isfinite(measured_m500s[i]):
print 'NOT FINITE'
print haloid
print measured
cPickle.dump(results, open('%s/consolidated.pkl' % outdir, 'w'))
failfile.close()
def compareNoiseProfiles(data=None):
''' Compare intrinsic noise levels to assumed shape noise'''
if data is None:
data = {}
if 'centers_mpc' not in data:
config = nfwfit.readConfiguration(
'/vol/euclid1/euclid1_1/dapple/mxxl_lensing/mxxlsnap54/general-c4-r10-n0_0-xrayNONE/config.sh'
)
simreader = nfwfit.buildSimReader(config)
nfwutils.global_cosmology.set_cosmology(simreader.getCosmology())
fitter = nfwfit.buildFitter(config)
intrnoise_profiles = []
for haloid in range(800, 880):
catbase = '/vol/euclid1/euclid1_raid1/dapple/mxxl_lensing/mxxlsnap54/halo_54_{}_0'.format(
haloid)
catalog = nfwfit.readSimCatalog(catbase, simreader, config)
r_mpc, ghat, sigma_ghat, beta_s, beta_s2, zlens = fitter.prepData(
catalog)
intrnoise_profiles.append(sigma_ghat)
intrnoise_profiles = np.row_stack(intrnoise_profiles)
intrnoise = np.mean(intrnoise_profiles, axis=0)
intrnoise_err = np.std(intrnoise_profiles, axis=0)
edges_mpc = np.linspace(config.profilemin, config.profilemax,
config.nbins + 1)
centers_mpc = (edges_mpc[1:] + edges_mpc[:-1]) / 2.
dL = nfwutils.global_cosmology.angulardist(zlens)
edges_arcmin = (edges_mpc / dL) * (180 / np.pi) * 60
bin_areas = np.pi * (edges_arcmin[1:]**2 - edges_arcmin[:-1]**2)
shapesigma = 0.25
galdensity = 20. # per sq arc min
shape_noise = shapesigma / np.sqrt(galdensity * bin_areas)
data['centers_mpc'] = centers_mpc
data['intrnoise'] = intrnoise
data['intrnoise_err'] = intrnoise_err
data['shape_noise'] = shape_noise
else:
centers_mpc = data['centers_mpc']
intrnoise = data['intrnoise']
intrnoise_err = data['intrnoise_err']
shape_noise = data['shape_noise']
fig = pylab.figure()
ax = pylab.gca()
ax.plot(centers_mpc,
intrnoise / intrnoise[2],
label='Intrinsic Noise [80 cluster avg]',
marker='None',
linestyle='-',
linewidth=2,
color=pp.colors[0])
ax.plot(centers_mpc,
shape_noise / shape_noise[2],
marker='None',
linestyle='-',
linewidth=3,
color=pp.colors[1],
label='Shape Noise')
ax.legend(loc='upper right')
ax.set_xlabel('Radius [Mpc]', fontsize=16)
ax.set_ylabel('Relative Shear Error [Arbit Norm]', fontsize=16)
fig.tight_layout()
fig.savefig('figures/relative_shear_noise.png')
fig.savefig('figures/relative_shear_noise.pdf')
fig.savefig('figures/relative_shear_noise.eps')
return fig, data
