lc_name = glob.glob(folder + '/' + lc_name_template)[0]
# get flux
time, flux = np.loadtxt(lc_name, dtype=(np.float, np.float),
usecols=(1, 2), unpack=True)
# Scale the light curve so that the mean level is one.
flux /= np.median(flux)
#pl.subplot(211)
#pl.title(lc_name_template)
#pl.step(time, flux, where='mid')
# estimate event properties
eep = EEP(time, flux)
eep.estimate()
results = eep.get_params()
# Cut flux with three times of the event width.
# At least 1 seconds.
width = max(results[0] * 3., 1.)
# 3 times of the width. The data point at zero time
# must be centered.
half_width = int((width / 2.) / 0.05)
center_index = np.searchsorted(time, results[3])
#print half_width, center_index
# This make always the number of data points "odd".
flux_cut = flux[max(center_index - half_width, 0):
min(center_index + half_width + 1, len(flux))]
#pl.subplot(212)
#pl.title(results)
# Select one random event (three light curves).
results, c_id = select_random(conn)
true_id.append(c_id)
# Add noise to the selected event.
#Also, make the light curves having 200 data points.
# 20 Hz time sampling. Centered at zero.
#print len(results['flux_a']), len(results['flux_b']), len(results['flux_c'])
results['flux_a'] = add_noise(results['flux_a'], SNRs[m])
results['flux_b'] = add_noise(results['flux_b'], SNRs[m])
results['flux_c'] = add_noise(results['flux_c'], SNRs[m])
# Estimate observables for each light curve.
eep = EEP(results['time_a'], results['flux_a'])
eep.estimate()
a_eep = eep.get_params()
width = max(a_eep[0] * 3., 1.)
half_width = int((width / 2.) / 0.05)
center_index = np.searchsorted(results['time_a'], a_eep[3])
start_index = center_index - half_width
end_index = center_index + half_width + 1
#start_index = max(center_index - half_width, 0)
#end_index = min(center_index + half_width + 1, len(results['flux_a']))
results['flux_a'] = results['flux_a'][start_index:end_index]
results['time_a'] = results['time_a'][start_index:end_index]
eep = EEP(results['time_b'], results['flux_b'])
eep.estimate()
b_eep = eep.get_params()
width = max(b_eep[0] * 3., 1.)
half_width = int((width / 2.) / 0.05)
for tel in tels:
# Binning, with offset 0.
results['flux_%d' % tel] = \
binning(results['flux_%d' % tel], 0, b_bin_steps)
results['time_%d' % tel] = \
binning(results['time_%d' % tel], 0, b_bin_steps)
# Add noise.
results['flux_%d' % tel] = \
add_noise(results['flux_%d' % tel], SNRs[m])
# Estimate observables for each light curve.
for tel in tels:
eep = EEP(results['time_%d' % tel], results['flux_%d' % tel])
eep.estimate()
results['eep_%d' % tel] = eep.get_params()
width = max(results['eep_%d' % tel][0] * 3., 0.4)
half_width = int((width / 2.) * base_sampling)
center_index = np.searchsorted(results['time_%d' % tel],
results['eep_%d' % tel][3])
results['flux_%d' % tel] = \
results['flux_%d' % tel] \
[max(center_index - half_width, 0):
min(center_index + half_width + 1,
len(results['flux_%d' % tel]))]
results['time_%d' % tel] = \
results['time_%d' % tel] \
[max(center_index - half_width, 0):
min(center_index + half_width + 1,
len(results['time_%d' % tel]))]
flux /= np.median(flux)
flux = add_noise(flux, np.inf)
# flux = add_noise(flux, 3)
# Time offset from 0 to 5.
colors = "rgbmck"
base_sampling = 120
sampling = 40
b_bin_steps = base_sampling / sampling
for i in range(b_bin_steps):
down_time = binning(time, i, b_bin_steps)
down_flux = binning(flux, i, b_bin_steps)
eep = EEP(down_time, down_flux)
eep.estimate()
print "binned N: ", len(eep.flux), eep.get_params()
results = eep.get_params()
width = max(results[0] * 3.0, 0.4)
# 3 times of the width. The data point at zero time
# must be centered.
half_width = int((width / 2.0) * base_sampling)
center_index = np.searchsorted(time, results[3])
flux_cut = flux[max(center_index - half_width, 0) : min(center_index + half_width + 1, len(flux))]
pl.subplot(122)
pl.step(eep.time, eep.flux, color=colors[i % len(colors)], marker="x", where="mid")
v = pl.axis()
x = np.linspace(v[0], v[1], 1000)
y = eep.gaussian_step(eep.p1, x)
time, flux = np.loadtxt(lc_name, dtype=(np.float, np.float),
usecols=(1, 2), unpack=True)
print len(time)
# Scale the light curve so that the mean level is one.
flux /= np.median(flux)
flux = add_noise(flux, np.inf)
width = 6
down_time = time[:(time.size // width) * width]. \
reshape(-1, width).mean(axis=1)
down_flux = flux[:(flux.size // width) * width]. \
reshape(-1, width).mean(axis=1)
eep = EEP(down_time, down_flux)
eep.estimate()
print eep.p1, len(eep.flux), eep.get_params()
results = eep.get_params()
width = max(results[0] * 3., 1.)
# 3 times of the width. The data point at zero time
# must be centered.
half_width = int((width / 2.) / 0.05)
center_index = np.searchsorted(time, results[3])
#print half_width, center_index
flux_cut = flux[max(center_index - half_width, 0):
min(center_index + half_width + 1, len(flux))]
#flux_cut = flux[center_index - half_width:
# center_index + half_width + 1]
#print flux_cut, len(flux_cut)
pl.subplot(122)
