def main():
"""
Demonstrate construction of multiple det data streams with a signal
injection
"""
#
# Generate waveform
#
print 'generating waveoform...'
waveform = pmns_utils.Waveform('shen_135135_lessvisc')
# Pick some extrinsic parameters
ext_params = ExtParams(distance=1,
ra=0.0,
dec=0.0,
polarization=0.0,
inclination=0.0,
phase=0.0,
geocent_peak_time=0.0 + 5.0)
# Construct the time series for these params
waveform.make_wf_timeseries(theta=ext_params.inclination,
phi=ext_params.phase)
#
# Generate IFO data
#
det1_data = DetData(waveform=waveform, ext_params=ext_params)
from scipy import signal
import pylab as pl
pl.figure()
pl.plot(det1_data.td_response.sample_times, det1_data.td_response.data)
pl.plot(det1_data.td_signal.sample_times, det1_data.td_signal.data)
pl.figure()
f, p = signal.welch(det1_data.td_response.data,
fs=1. / det1_data.delta_t,
nperseg=512)
pl.loglog(f, np.sqrt(p))
f, p = signal.welch(det1_data.td_signal.data,
fs=1. / det1_data.delta_t,
nperseg=512)
pl.loglog(f, np.sqrt(p))
pl.ylim(1e-25, 1e-21)
pl.show()
method='nelder-mead')
return result_theory
# --------------------------------------------------------------------
# Data Generation
#
# Generate Signal Data
#
# Signal
print ''
print '--- %s ---' % sys.argv[1]
waveform = pmns_utils.Waveform('%s' % sys.argv[1])
waveform.compute_characteristics()
# Extrinsic parameters
ext_params = simsig.ExtParams(20.0,
ra=0.0,
dec=0.0,
polarization=0.0,
inclination=0.0,
phase=0.0,
geocent_peak_time=0.25)
# Frequency range for SNR around f2 peak - we'll go between 1, 5 kHz for the
# actual match calculations, though
flow = waveform.fpeak - 150
fupp = waveform.fpeak + 150
######################################################################
# Data Generation
print >> sys.stdout, '''
-------------------------------------
Beginning pmns_pca_inference:
1) Generating Data
'''
#
# Generate Signal Data
#
print >> sys.stdout, "generating waveform..."
waveform = pmns_utils.Waveform('%s_lessvisc'%opts.waveform_name)
waveform.compute_characteristics()
te=time.time()
print >> sys.stdout, "...waveform construction took: %f sec"%(te-ts0)
# ------------------------------------------------------------------------------------
# --- Set up timing for injection: inject in center of segment
# Need to be careful. The latest, longest template must lie within the data
# segment. note also that the noise length needs to be a power of 2.
max_sig=100e-3
cbc_delta_t=10e-3 # std of gaussian with cbc timing error
time_prior_width=3*cbc_delta_t
max_sig_len=int(np.ceil(max_sig*16384))
# # Build Catalogues # fshift_center = 1000 print "building catalogues" (freqaxis, low_cat, high_cat, shift_cat, original_cat, fpeaks, low_sigmas, high_sigmas) = pmns_pca_utils.build_catalogues(waveform_names, fshift_center) delta_f = np.diff(freqaxis)[0] # Convert to magnitude/phase full_mag, full_phase = pmns_pca_utils.complex_to_polar(original_cat) shift_mag, shift_phase = pmns_pca_utils.complex_to_polar(shift_cat) low_mag, low_phase = pmns_pca_utils.complex_to_polar(low_cat) high_mag, high_phase = pmns_pca_utils.complex_to_polar(high_cat) waveform = pmns_utils.Waveform(waveform_names[1]) waveform.reproject_waveform() # Window peakidx = np.argmax(abs(waveform.hplus.data)) win = lal.CreateTukeyREAL8Window(len(waveform.hplus), 0.25) waveform.hplus.data *= win.data.data wavdata = np.zeros(16384) wavdata[:len(waveform.hplus.data)] = np.copy(waveform.hplus.data) waveform_TD = pycbc.types.TimeSeries(wavdata, delta_t=waveform.hplus.delta_t) waveform_FD = waveform_TD.to_frequencyseries() mag = abs(waveform_FD.data) phase = np.unwrap(np.angle(waveform_FD.data))
posterior['bandwidth'].samples)
durationPDF = bppu.PosteriorOneDPDF(name='duration',
posterior_samples=duration_samps)
posterior.append(durationPDF)
return posterior
# -----------------------------------------------------------------------------------------------------------
# **** MAIN **** #
posterior_file = sys.argv[1]
bsn_file = sys.argv[2]
snr_file = sys.argv[3]
waveform = pmns_utils.Waveform("shen_135135_lessvisc")
#waveform=pmns_utils.Waveform("apr_135135_lessvisc")
waveform.compute_characteristics()
waveform.reproject_waveform()
H = waveform.hplus.to_frequencyseries()
fw, axw = pl.subplots()
axw.plot(H.sample_frequencies, abs(H), color='r')
oneDMenu, twoDMenu, binSizes = oneD_bin_params()
# Get true values:
truevals = {}
for item in oneDMenu:
if item == 'frequency':
truevals[item] = waveform.fpeak
def main():
print "running pmns_pp_utils.py doesn't do much (yet)"
# Get post-proc dir
ppdir = sys.argv[1]
waveform_name = sys.argv[2]
waveform = pmns_utils.Waveform(waveform_name)
waveform.reproject_waveform()
waveform.compute_characteristics()
outdir = os.path.join(ppdir, "gmmfreqs")
try:
os.makedirs(outdir)
except OSError:
print 'e'
# Identify posterior files:
posfiles, Bfiles, snrsfiles, posparents = find_posterior_files(ppdir)
logBs = load_Bfiles(Bfiles)
SNRs = load_snrfiles(snrsfiles)
# Get the frequency posterior modes
freq_mode_means, freq_mode_covars = get_modes(posfiles, plot=True)
# Compute the fpeak accuracy, assuming the highest freq mode is the one
# which captures fpeak
freq_acc = abs(waveform.fpeak - freq_mode_means[:, 0])
#
# Summarise the results
#
freq_acc_summary = measurement_summary('GMMfpeakAccuracy', freq_acc)
fmax_mu_summary = measurement_summary('GMMfmax', freq_mode_means[:, 0])
fmid_mu_summary = measurement_summary('GMMfmid', freq_mode_means[:, 1])
fmin_mu_summary = measurement_summary('GMMfmin', freq_mode_means[:, 2])
dump_summary(
os.path.join(outdir, 'GMMfreqsSummary.txt'),
[freq_acc_summary, fmax_mu_summary, fmid_mu_summary, fmin_mu_summary])
# fmax_sigma_summary = measurement_summary('GMMfmax_sigma',
# np.sqrt(freq_mode_covars[:,0]))
# fmid_sigma_summary = measurement_summary('GMMfmid_sigma',
# np.sqrt(freq_mode_covars[:,1]))
# fmin_sigma_summary = measurement_summary('GMMfmin_sigma',
# np.sqrt(freq_mode_covars[:,2]))
#
#
# Generate Web Page & Plots
#
# SNR vs freq acc
fig, ax = plot_measurement_vs_statistic(x=SNRs, y=freq_acc, \
yerrs=None, truevals=None, param=None, xlabel=r'Injected SNR',
ylabel='GMM Frequency Accuracy [Hz]')
ax.set_yscale('log')
fig.savefig('{outdir}/GMMfpeakAccuracyvsinjSNR.png'.format(outdir=outdir))
pl.close(fig)
# logB vs freq acc
fig, ax = plot_measurement_vs_statistic(x=logBs, y=freq_acc, \
yerrs=None, truevals=None, param=None, xlabel=r'BSN',
ylabel='GMM Frequency Accuracy [Hz]')
ax.set_yscale('log')
fig.savefig('{outdir}/GMMfpeakAccuracyvsBSN.png'.format(outdir=outdir))
pl.close(fig)
# SNR vs maxfreq
fig, ax = plot_measurement_vs_statistic(x=SNRs, y=freq_mode_means[:,0], \
yerrs=None, truevals=None, param=None, xlabel=r'Injected SNR',
ylabel='GMM Max Frequency [Hz]')
fig.savefig('{outdir}/GMMMaxFreqvsinjSNR.png'.format(outdir=outdir))
pl.close(fig)
# logB vs maxfreq
fig, ax = plot_measurement_vs_statistic(x=logBs, y=freq_mode_means[:,0], \
yerrs=None, truevals=None, param=None, xlabel=r'BSN',
ylabel='GMM Max Frequency [Hz]')
fig.savefig('{outdir}/GMMMaxFreqvsBSN.png'.format(outdir=outdir))
pl.close(fig)
# SNR vs midfreq
fig, ax = plot_measurement_vs_statistic(x=SNRs, y=freq_mode_means[:,1], \
yerrs=None, truevals=None, param=None, xlabel=r'Injected SNR',
ylabel='GMM Mid Frequency [Hz]')
fig.savefig('{outdir}/GMMMidFreqvsinjSNR.png'.format(outdir=outdir))
pl.close(fig)
# logB vs midfreq
fig, ax = plot_measurement_vs_statistic(x=logBs, y=freq_mode_means[:,1], \
yerrs=None, truevals=None, param=None, xlabel=r'BSN',
ylabel='GMM Mid Frequency [Hz]')
fig.savefig('{outdir}/GMMMidFreqvsBSN.png'.format(outdir=outdir))
pl.close(fig)
# SNR vs minfreq
fig, ax = plot_measurement_vs_statistic(x=SNRs, y=freq_mode_means[:,0], \
yerrs=None, truevals=None, param=None, xlabel=r'Injected SNR',
ylabel='GMM Min Frequency [Hz]')
fig.savefig('{outdir}/GMMMinFreqvsinjSNR.png'.format(outdir=outdir))
pl.close(fig)
# logB vs minfreq
fig, ax = plot_measurement_vs_statistic(x=logBs, y=freq_mode_means[:,0], \
yerrs=None, truevals=None, param=None, xlabel=r'BSN',
ylabel='GMM Min Frequency[Hz]')
fig.savefig('{outdir}/GMMMinFreqvsBSN.png'.format(outdir=outdir))
pl.close(fig)
waveform_names = [
'apr_135135_lessvisc', 'shen_135135_lessvisc', 'dd2_135135_lessvisc',
'dd2_165165_lessvisc', 'nl3_135135_lessvisc', 'nl3_1919_lessvisc',
'tm1_135135_lessvisc', 'tma_135135_lessvisc', 'sfhx_135135_lessvisc',
'sfho_135135_lessvisc'
]
#
# Waveform
#
print ''
print '--- %s ---' % sys.argv[1]
#waveform = pmns_utils.Waveform('%s_lessvisc'%sys.argv[1])
matches = []
for name in waveform_names:
waveform = pmns_utils.Waveform(name)
waveform.compute_characteristics()
# Get optimal hplus
waveform.reproject_waveform()
h = apply_taper(waveform.hplus)
h_s = pycbc.filter.sigma(h,
low_frequency_cutoff=fmin,
high_frequency_cutoff=fmax)
#
# Spectrum
#
H = abs(h.to_frequencyseries())**2
freqs = h.to_frequencyseries().sample_frequencies.data
#datafiles=sys.argv[2]
# XXX: Hardcoding
#distances=np.array([5, 7.5, 10, 12.5, 15, 17.5, 20])
distances=np.array([20, 25, 30, 35, 40, 45, 50, 55, 60, 70, 80, 90, 100])
odds2pos = lambda logB: 1.0/(1.0+1.0/np.exp(logB))
#logBthresh=np.interp(0.9, odds2pos(np.arange(0, 10, 1e-4)), np.arange(0, 10,
# 1e-4))
found_criterion=sys.argv[2]
logBthresh=np.log(float(sys.argv[3]))
netSNRthresh=float(sys.argv[3])
wf = pmns_utils.Waveform(waveform_name+'_lessvisc')
wf.compute_characteristics()
logBs = []
netSNRs = []
freq_pdfs = []
maxfreqs = []
meanfreqs = []
credintvals = []
credintwidths = []
confintvals=[]
confintwidths=[]
#datafiles=glob.glob('LIB-PMNS_waveform-%s_distance-*.pickle'%waveform_name)
for c in xrange(len(upps)):
upps[c] = intervals[1][c] - centers[c]
lows[c] = centers[c] - intervals[0][c]
return lows, upps
# -------------------------------
# Load results
#datafiles = sys.argv[1]
datafile = sys.argv[1]
# XXX: Hardcoding
wf = pmns_utils.Waveform('dd2_135135_lessvisc')
wf.compute_characteristics()
# --- Data extraction
print >> sys.stdout, "loading %s..." % datafile
freq_axis, freq_pdfs_tmp, freq_estimates, all_sig_evs, all_noise_evs =\
pickle.load(open(datafile))
logBs = all_sig_evs - all_noise_evs
#logBthreshs = np.linspace(min(logBs), max(logBs), 100)
#logBthresh_range = stats.scoreatpercentile(logBs, [5,95])
#logBthreshs = np.linspace(min(logBthresh_range), max(logBthresh_range), 10)
logBthreshs = np.arange(-0.5, 0.5 + 0.05, 0.05)
# Preallocation
BSN_threshold=-np.inf
# --- end input
if not os.path.isdir(outputdirectory):
os.makedirs(outputdirectory)
else:
print >> sys.stdout, \
"warning: %s exists, results will be overwritten"%outputdirectory
currentdir=os.path.join(outputdirectory,'summaryfigs')
# --- Construct the injected waveform
if waveform_name not in ["av15", "av16"]:
waveform_name += "_lessvisc"
waveform = pmns_utils.Waveform('%s'%waveform_name)
waveform.reproject_waveform(0,0)
waveform.compute_characteristics()
# --- Identify files
ifospattern="V1H1L1"
engineglobpattern="lalinferencenest"
sampglobpattern="posterior"
BSNfiles = glob.glob('%s/posterior_samples/%s_%s*.dat_B.txt'%(
resultsdir, sampglobpattern, ifospattern ) )
snrfiles = glob.glob('%s/engine/%s*-%s-*_snr.txt'%(
resultsdir, engineglobpattern, ifospattern ) )