def super_worker(self, *args):
# generate parameter dict for transit function input
pars = self.init.copy()
for k in range(len(self.keys)):
pars[self.keys[k]] = args[k]
# compute white noise
noise = pars['rp']**2 / pars['sig_tol']
ndata = np.random.normal(1, noise, len(self.t))
# generate transit + variability
t = self.t - np.min(self.t)
A = pars['A'] + pars['A'] * np.sin(2 * np.pi * t / pars['PA'])
w = pars['w'] + pars['w'] * np.sin(2 * np.pi * t / pars['Pw'])
data = transit(time=self.t, values=pars) * ndata * (
1 + A * np.sin(2 * np.pi * t / w + pars['phi']))
#2018 Apr 24 - ndata is the noise signal but without any transit in it
# So results has one column that is all the transit and noise parameters,
# the next column contains the transit + noise signals,
# and the final column contains just the noise signals without any transits
# add systematic into noise "same distribution different points"
ndata = np.random.normal(1, noise, len(
self.t)) * (1 + A * np.sin(2 * np.pi * t / w + pars['phi']))
return args, data, ndata
def super_worker(self, *args):
# generate parameter dict for transit function input
pars = self.init.copy()
for k in range(len(self.keys)):
pars[self.keys[k]] = args[k]
# compute white noise
noise = pars['rp']**2 / pars['sig_tol']
ndata = np.random.normal(1, noise, len(self.t))
# generate transit + variability
t = self.t - np.min(self.t)
A = pars['A'] + pars['A'] * np.sin(2 * np.pi * t / pars['PA'])
w = pars['w'] + pars['w'] * np.sin(2 * np.pi * t / pars['Pw'])
kk = np.random.randint(2, 5)
w = w / kk
tt = transit(time=self.t, values=pars)
tt = 1 + (tt - 1) * np.random.normal(1, 0.15)
data = tt * ndata * (1 + A * np.sin(2 * np.pi * t / w + pars['phi']))
# add systematic into noise "same distribution different points"
ndata = np.random.normal(1, noise, len(
self.t)) * (1 + A * np.sin(2 * np.pi * t / w + pars['phi']))
return args, data, ndata
def generate_positives(residuals, pars, N=10000, minsnr=1.5, snrscale=2):
positive = []
snrs = []
# generate some positive examples
for i in range(10000):
# choose random sample of time, ensure we have enough data points ``
length = 0
while (length != 180):
di = np.random.choice(np.arange(len(times)))
# make sure we get enough data for NN
ti = np.random.choice(np.arange(len(times[di])))
start_t = times[di][ti]
end_t = times[di][ti] + 2 * 180 / (24 * 60)
tmask = (times[di] > start_t) & (times[di] < end_t)
length = tmask.sum()
tlength = 0
# make sure the transit length is sufficient
while (tlength < 15):
std = np.std(residuals[di][tmask])
snr = minsnr + np.random.uniform() * snrscale
snrs.append(snr)
# create parameters for injected data set
pars['ar'] = np.random.uniform() * 7 + 7
pars['per'] = np.random.uniform() * 4 + 2
pars['tm'] = np.median(
times[di][tmask]) + np.random.random() / 24 / 60
pars['rp'] = (np.abs(snr) * std)**0.5
tmodel = transit(time=times[di][tmask], values=pars)
if (snr < 0):
tmodel *= -1
tmodel += 2
tlength = (tmodel != 1).sum()
# create data and assess transit probability
data = residuals[di][tmask] * tmodel
pdata = preprocessing.scale(data)
positive.append(pdata)
return snrs, positive
# plt.show()
NPTS = 1000
# simulate transit data
from ELCA import lc_fitter, transit
from flux_decorrelation import gaussian_weights
t = np.linspace(0.85,1.05,NPTS)
init = { 'rp':0.1, 'ar':14.07, # Rp/Rs, a/Rs
'per':3.336817, 'inc':88.75, # Period (days), Inclination
'u1': 0.3, 'u2': 0, # limb darkening (linear, quadratic)
'ecc':0, 'ome':0, # Eccentricity, Arg of periastron
'a0':1, 'a1':0, # Airmass extinction terms
'a2':0, 'tm':0.95 } # tm = Mid Transit time (Days)
data = transit(time=t, values=init) #+ np.random.normal(0, 2e-4, len(t))
# simulate PSFs on detector
xcent = [15.13]
ycent = [14.96]
sigmax = [0.55]
sigmay = [0.55]
for i in range(1,NPTS):
xcent.append( xcent[i-1] + np.random.normal(0,0.005) )
ycent.append( ycent[i-1] + np.random.normal(0,0.005) )
sigmax.append( sigmax[i-1] + np.random.normal(0,0.0005) )
sigmay.append( sigmay[i-1] + np.random.normal(0,0.0005) )
# simulate images on detector
X = np.zeros((NPTS,7))
# GENERATE DATA
NDAYS = 30
CADENCE = 29.4 # minutes
NPTS = int(NDAYS * 24 * 60 / CADENCE)
# transit data
t = np.linspace(0.5, 0.5 + NDAYS, NPTS)
init = {
'rp': 0.1,
'ar': 12.0, # Rp/Rs, a/Rs
'per': 8,
'inc': 90, # Period (days), Inclination
'u1': 0.7,
'u2': 0, # limb darkening (linear, quadratic)
'ecc': 0,
'ome': 0, # Eccentricity, Arg of periastron
'a0': 1,
'a1': 0, # Airmass extinction terms
'a2': 0,
'tm': 1
} # tm = Mid Transit time (Days)
model = transit(time=t, values=init)
data = model * np.random.normal(1, 1000e-6, NPTS) + 500e-6 * np.sin(
2 * np.pi * np.arange(NPTS) / (0.5 * NPTS))
alphas, rgba_colors = time_series_eval(data, mlp, spacing=5, ws=180)
plot_time_series(t, data, alphas, rgba_colors)
# TODO implement BLS
'rp': 0.1,
'ar': 12, # Rp/Rs, a/Rs
'per': 2,
'inc': 90, # Period (days), Inclination
'u1': 0.5,
'u2': 0, # limb darkening (linear, quadratic)
'ecc': 0,
'ome': 0, # Eccentricity, Arg of periastron
'a0': 1,
'a1': 0, # Airmass extinction terms
'a2': 0,
'tm': 1
} # tm = Mid Transit time (Days)
for v in changes[key]:
init[key] = v
ax[i, j].plot(t,
transit(time=t, values=init),
label="{} = {}".format(label[key], v))
ax[i, j].legend(loc='best')
ax[i, j].set_xlabel('Time (Days)')
ax[i, j].set_ylim([0.986, 1.0005])
if j == 1:
ax[i, j].get_yaxis().set_ticks([])
else:
ax[i, j].set_ylabel('Relative Flux')
plt.savefig('transit_shape_analysis.png')