def test_calibrate_lrt_works_as_expected(self):
m = 1
df = 0.01
freq = np.arange(df, 5 + df, df)
nfreq = freq.size
rng = np.random.RandomState(100)
noise = rng.exponential(size=nfreq)
model = models.Const1D()
model.amplitude = 2.0
p = model(freq)
power = noise * p
ps = Powerspectrum()
ps.freq = freq
ps.power = power
ps.m = m
ps.df = df
ps.norm = "leahy"
loglike = PSDLogLikelihood(ps.freq, ps.power, model, m=1)
s_all = np.atleast_2d(np.ones(10) * 2.0).T
model2 = models.PowerLaw1D() + models.Const1D()
model2.x_0_0.fixed = True
loglike2 = PSDLogLikelihood(ps.freq, ps.power, model2, m=1)
pe = PSDParEst(ps)
pval = pe.calibrate_lrt(loglike, [2.0], loglike2,
[2.0, 1.0, 2.0], sample=s_all,
max_post=False, nsim=5,
seed=100)
assert pval > 0.001
def fit_gaussian(data_1d, x_loc=0.0):
nx = data_1d.shape[0]
if x_loc < 0.01:
x_loc = nx / 2.0
x_1d = np.arange(nx)
# subtract the mean or median, although it may be better to have
# the model fit a continuum with Const1D ?
datamed = np.median(data_1d)
# data_sub = data_1d - np.median(data_1d)
gauss_init = models.Gaussian1D(
amplitude=1.0, mean=x_loc,
stddev=1.5) + models.Const1D(amplitude=datamed)
fit_g = fitting.LevMarLSQFitter()
gauss = fit_g(gauss_init, x_1d, data_1d)
return gauss
def astropy_clipping_fit(x, row, variance):
gauss_model = models.Gaussian1D() + models.Const1D()
fit = fitting.LevMarLSQFitter()
clipping_fit = fitting.FittingWithOutlierRemoval(fit,
sigma_clip,
niter=30,
sigma=15.0)
# initialize fitters
fit = fitting.LevMarLSQFitter()
or_fit = fitting.FittingWithOutlierRemoval(fit,
sigma_clip,
niter=3,
sigma=6.0)
g_init = models.Gaussian1D(
amplitude=np.max(row), mean=len(x) / 2,
stddev=1.0) + models.Const1D(amplitude=np.min(row))
filtered_data, or_fitted_model = or_fit(g_init,
x,
row,
weights=1.0 / variance)
return or_fitted_model.mean_0.value
def fit_gaussians_to_spectrum(obs_flx, obs_wvl, idx_peaks):
print ' Fitting multiple ({:.0f}) gaussinas to the extracted arc spectrum'.format(len(idx_peaks))
#
median_val = np.median(obs_flx)
mean_wvl = 0 # np.mean(obs_wvl)
fit_model = models.Const1D(amplitude=np.median(obs_flx)) + polynomial.Polynomial1D(4)
for i_p, idx_p in enumerate(idx_peaks):
peak_val = obs_flx[idx_p] - median_val
# dela boljs brez nastavljenih boundov
fit_model += models.Gaussian1D(amplitude=peak_val, mean=obs_wvl[idx_p]-mean_wvl, stddev=0.1)#,
# bounds={'mean': (obs_wvl[idx_p]-0.5, obs_wvl[idx_p]+0.5),
# 'amplitude': (peak_val*0.8, peak_val*1.2)})
fit_t = fitting.LevMarLSQFitter()
fitted_model = fit_t(fit_model, obs_wvl-mean_wvl, obs_flx)
return fitted_model
def setup_class(cls):
nx = 1000000
cls.x = np.arange(nx)
cls.countrate = 10.0
cls.cerr = 2.0
cls.y = np.random.normal(cls.countrate, cls.cerr, size=cls.x.shape[0])
cls.yerr = np.ones_like(cls.y) * cls.cerr
cls.model = models.Const1D()
p_amplitude = lambda amplitude: \
scipy.stats.norm(loc=cls.countrate, scale=cls.cerr).pdf(amplitude)
cls.priors = {"amplitude": p_amplitude}
def setup_class(cls):
np.random.seed(1000)
m = 1
nfreq = 100
freq = np.arange(nfreq)
noise = np.random.exponential(size=nfreq)
power = noise * 2.0
ps = Powerspectrum()
ps.freq = freq
ps.power = power
ps.m = m
ps.n = freq.shape[0]
ps.df = freq[1] - freq[0]
ps.norm = "leahy"
cls.ps = ps
cls.a_mean, cls.a_var = 2.0, 1.0
cls.model = models.Const1D()
p_amplitude = lambda amplitude: \
scipy.stats.norm(loc=cls.a_mean, scale=cls.a_var).pdf(amplitude)
cls.priors = {"amplitude": p_amplitude}
cls.lpost = PSDPosterior(cls.ps.freq,
cls.ps.power,
cls.model,
m=cls.ps.m)
cls.lpost.logprior = set_logprior(cls.lpost, cls.priors)
cls.fitmethod = "powell"
cls.max_post = True
cls.t0 = np.array([2.0])
cls.neg = True
cls.opt = scipy.optimize.minimize(cls.lpost,
cls.t0,
method=cls.fitmethod,
args=cls.neg,
tol=1.e-10)
cls.opt.x = np.atleast_1d(cls.opt.x)
cls.optres = OptimizationResultsSubclassDummy(cls.lpost,
cls.opt,
neg=True)
def estimate_peak_width(data, min=2, max=8):
"""
Estimates the FWHM of the spectral features (arc lines) by fitting
Gaussians to the brightest peaks.
Parameters
----------
data: ndarray
1D data array (will be modified)
min: int
minimum plausible peak width
max: int
maximum plausible peak width (not inclusive)
Returns
-------
float: estimate of FWHM of features
"""
all_widths = []
for fwidth in range(min, max + 1): # plausible range of widths
data_copy = data.copy() # We'll be editing the data
widths = []
for i in range(15): # 15 brightest peaks, should ensure we get real ones
index = 2 * fwidth + np.argmax(data_copy[2 * fwidth:-2 * fwidth - 1])
data_to_fit = data_copy[index - 2 * fwidth:index + 2 * fwidth + 1]
m_init = models.Gaussian1D(stddev=0.42466 * fwidth) + models.Const1D(np.min(data_to_fit))
m_init.mean_0.bounds = [-1, 1]
m_init.amplitude_1.fixed = True
fit_it = fitting.FittingWithOutlierRemoval(fitting.LevMarLSQFitter(),
sigma_clip, sigma=3)
with warnings.catch_warnings():
# Ignore model linearity warning from the fitter
warnings.simplefilter('ignore')
m_final, _ = fit_it(m_init, np.arange(-2 * fwidth, 2 * fwidth + 1),
data_to_fit)
# Quick'n'dirty logic to remove "peaks" at edges of CCDs
if m_final.amplitude_1 != 0 and m_final.stddev_0 < fwidth:
widths.append(m_final.stddev_0 / 0.42466)
# Set data to zero so no peak is found here
data_copy[index - 2 * fwidth:index + 2 * fwidth + 1] = 0.
all_widths.append(sigma_clip(widths).mean())
return sigma_clip(all_widths).mean()
def test_transform_magnitudes_identical_input(order):
# Analogous to the test case for calculate_transform_coefficients
# above where the input magnitudes are identical, except the input
# objects have coordinates.
n_stars = 100
zero = models.Const1D(0.0)
instrumental, catalog_table = generate_tables(n_stars, zero)
calib_mags, stars_with_match, transform = \
transform_magnitudes(instrumental, catalog_table, catalog_table,
order=order)
print(calib_mags)
assert all(calib_mags == catalog_table['r_mag'])
assert all(stars_with_match)
assert len(transform.parameters) == order + 1
assert all(transform.parameters == 0)
def fitB4(coord_iso):
u = coord_iso[0]
v = coord_iso[1]
r = np.sqrt(u**2 + v**2)
E=np.zeros(len(u))
for i in range(0,len(u)):
if np.sign(u[i])>0 and np.sign(v[i])>0 : E[i]=np.arctan(v[i]/u[i])
elif np.sign(u[i])>0 and np.sign(v[i])<0 : E[i]=2*np.pi+np.arctan(v[i]/u[i])
else : E[i]=np.pi+np.arctan(v[i]/u[i])
Es = np.linspace(0,2*np.pi,num=100)
dic = {'frequency': True, 'phase': True}
g_init = (models.Sine1D(amplitude=0.1, frequency=1.5/np.pi, fixed=dic)
+models.Sine1D(amplitude=0.1, frequency=2/np.pi, fixed=dic)
+models.Sine1D(amplitude=0.1, frequency=1.5/np.pi,
phase=0.25, fixed=dic)
+models.Sine1D(amplitude=0.1, frequency=2/np.pi,
phase=0.25, fixed=dic)
+models.Const1D(amplitude=1.0,fixed={'amplitude':True}))
fit = fitting.LevMarLSQFitter()
or_fit = fitting.FittingWithOutlierRemoval(fit, sigma_clip, niter=3, sigma=3.0)
filtered_data, or_fitted_model = or_fit(g_init, E, r)
fitted_model = fit(g_init, E, r)
plt.figure(figsize=(8,5))
plt.plot(E, r, 'gx', label="original data")
plt.plot(E, filtered_data, 'r+', label="filtered data")
plt.plot(Es, fitted_model(Es), 'g-',
label="model fitted w/ original data")
plt.plot(Es, or_fitted_model(Es), 'r--',
label="model fitted w/ filtered data")
plt.legend(loc=2, numpoints=1)
return or_fitted_model[3].amplitude.value
def test_calibrate_highest_outlier_works_with_sampling(self):
m = 1
nfreq = 100
seed = 100
freq = np.linspace(1, 10, nfreq)
rng = np.random.RandomState(seed)
noise = rng.exponential(size=nfreq)
model = models.Const1D()
model.amplitude = 2.0
p = model(freq)
power = noise * p
ps = Powerspectrum()
ps.freq = freq
ps.power = power
ps.m = m
ps.df = freq[1] - freq[0]
ps.norm = "leahy"
nsim = 5
lpost = PSDPosterior(ps.freq, ps.power, model, m=1)
p_amplitude = lambda amplitude: \
scipy.stats.norm(loc=1.0, scale=1.0).pdf(
amplitude)
priors = {"amplitude": p_amplitude}
lpost.logprior = set_logprior(lpost, priors)
pe = PSDParEst(ps)
with catch_warnings(RuntimeWarning):
pval = pe.calibrate_highest_outlier(lpost, [2.0],
sample=None,
max_post=True,
seed=seed,
nsim=nsim,
niter=10,
nwalkers=20,
burnin=10)
assert pval > 0.001
def fit_gaussian_peak(x, y, guess=None):
# fit a Gaussian to a peak
#
# Return model fitted object
# peak_model_fit.parameters = [offset, amplitude, mean, standard deviation]
if guess is None:
# help make some guesses
offset = np.min(y)
guess = [offset, np.max(y) - offset, x[np.argmax(y)], 1.0]
g1 = models.Gaussian1D(guess[1], guess[2], guess[3])
amp = models.Const1D(guess[0])
peak_model = amp + g1
fitter = fitting.LevMarLSQFitter()
peak_model_fit = fitter(peak_model, x, y)
return peak_model_fit
def plot_lum_vs_lag():
rmid_list = get_total_rmid_list()
zfinal = pickle.load(open(Location.project_loca +
"/info_database/zfinal.pkl"))
lag_list = list()
lag_err_list = list()
lum_list = list()
lum_err_list = list()
fig = plt.figure()
for each in rmid_list:
try:
lag, lag_err = get_lag(each, zfinal)
lum, lum_err = get_lum(each, zfinal)
if lag < 3.0 * lag_err or lum < 3.0 * lum_err:
raise Exception
lag_list.append(lag)
lag_err_list.append(lag_err)
lum_list.append(lum)
lum_err_list.append(lum_err)
except Exception:
continue
# plt.errorbar([lum], [lag])
# fig.text(np.log10(lum), np.log10(lag), str(each))
lag = np.log10(np.array(lag_list))
lum = np.log10(np.array(lum_list))
lag_err = np.array(lag_err_list) / np.array(lag_list)
lum_err = np.array(lum_err_list) / np.array(lum_list)
theory = models.Linear1D(0.5, -10.0, fixed={"slope": True})
obs = models.Const1D(0.0)
fitter = fitting.LinearLSQFitter()
theory_fit = fitter(theory, lum, lag, weights = lag_err ** 2.0)
print(theory_fit.parameters)
obs_fit = fitter(obs, lum, lag, weights = lag_err ** 2.0)
print(obs_fit.parameters)
rcs = np.sum((obs_fit(lum) - lag) ** 2.0) ** 0.5 / (len(rmid_list) - 1.0)
print(rcs)
print(len(lum))
plt.errorbar(lum, lag, xerr=lum_err, yerr=lag_err, fmt='o')
# plt.plot(lum, theory_fit(lum))
plt.plot(lum, obs_fit(lum))
plt.show()
def fit_gaussians_to_spectrum_linebyline(obs_flx, obs_wvl, idx_peaks, d_wvl=2):
print ' Fitting {:.0f} individual gaussinas to the extracted arc spectrum'.format(len(idx_peaks))
fitted_vals = [] # [mean, std] combinations
for i_p, idx_p in enumerate(idx_peaks):
# data subset
idx_data_use = np.logical_and(obs_wvl > obs_wvl[idx_p]-d_wvl, obs_wvl < obs_wvl[idx_p]+d_wvl)
use_flx = obs_flx[idx_data_use]
use_wvl = obs_wvl[idx_data_use]
#
median_val = np.median(use_flx)
fit_model = models.Const1D(amplitude=np.median(use_flx))
peak_val = obs_flx[idx_p] - median_val
# dela boljs brez nastavljenih boundov
fit_model += models.Gaussian1D(amplitude=peak_val, mean=obs_wvl[idx_p], stddev=0.1)#,
# bounds={'mean': (obs_wvl[idx_p]-0.5, obs_wvl[idx_p]+0.5),
# 'amplitude': (peak_val*0.8, peak_val*1.2)})
fit_t = fitting.LevMarLSQFitter()
fitted_model = fit_t(fit_model, use_wvl, use_flx)
# print fitted_model.param_names
fitted_vals.append(fitted_model.parameters[2:])
return np.array(fitted_vals)
def refine_velocity_guess(spectrum, axis, v_guess, detected_line, return_fit = False):
"""
Refines a velocity guess with better accuracy, avoiding to have discrete guess corresponding to pixel numbers.
A gaussian function is fitted to estimate the line position.
Parameters
----------
spectrum : 1D :class:`~numpy:numpy.ndarray`
The spectrum containing emission lines
axis : 1D :class:`~numpy:numpy.ndarray`
The corresponding axis, in wavenumber
v_guess : float
A first guess on the velocity, typically obtained from :func:`guess_source_velocity`
detected_line : str
Names of the main line present in the spectrum (as defined `here <http://celeste.phy.ulaval.ca/orcs-doc/introduction.html#list-of-available-lines>`_)
return_fit : bool, Default = False
(Optional) If True, returns the fit parameters, for further investigation
Returns
-------
v : float
The updated velocity guess
"""
from orb.utils.spectrum import line_shift, compute_radial_velocity
from orb.core import Lines
from astropy.modeling import models, fitting
line_rest = Lines().get_line_cm1(detected_line)
mu = line_rest + line_shift(v_guess, line_rest, wavenumber=True)
G0 = models.Gaussian1D(amplitude=np.nanmax(spectrum), mean=mu, stddev=11)
G0.mean.max = line_rest + line_shift(v_guess-25, line_rest, wavenumber=True)
G0.mean.min = line_rest + line_shift(v_guess+25, line_rest, wavenumber=True)
C = models.Const1D(amplitude = np.nanmedian(spectrum))
model = C+G0
fitter = fitting.LevMarLSQFitter()
fit = fitter(model, axis, spectrum)
if return_fit:
return compute_radial_velocity(fit.mean_1,line_rest, wavenumber=True), fit
else:
return compute_radial_velocity(fit.mean_1,line_rest, wavenumber=True)
def __init__(self, x, y, dy, initial_models=None):
QObject.__init__(self)
self.x = x
self.y = y
self.dy = dy
if initial_models is None:
initial_models = [models.Const1D(0.0)]
self.models = initial_models
self.ui = ModelBrowserUI(self, self.models, self.x, self.y)
self.plot, self.resid = _build_axes(self.ui.canvas.fig)
self._draw(preserve_limits=False)
self.mouse_handler = ModelEventHandler(self.plot, self.active_model)
add_callback(self.mouse_handler, 'model', self.set_model)
self.ui.fit.pressed.connect(nonpartial(self.fit))
self._sync_model_list()
def setup_class(cls):
np.random.seed(150)
cls.nlor = 3
cls.x_0_0 = 0.5
cls.x_0_1 = 2.0
cls.x_0_2 = 7.5
cls.amplitude_0 = 200.0
cls.amplitude_1 = 100.0
cls.amplitude_2 = 50.0
cls.fwhm_0 = 0.1
cls.fwhm_1 = 1.0
cls.fwhm_2 = 0.5
cls.whitenoise = 2.0
cls.model = models.Lorentz1D(cls.amplitude_0, cls.x_0_0, cls.fwhm_0) + \
models.Lorentz1D(cls.amplitude_1, cls.x_0_1, cls.fwhm_1) + \
models.Lorentz1D(cls.amplitude_2, cls.x_0_2, cls.fwhm_2) + \
models.Const1D(cls.whitenoise)
freq = np.linspace(0.01, 10.0, 10.0 / 0.01)
p = cls.model(freq)
noise = np.random.exponential(size=len(freq))
power = p * noise
cls.ps = Powerspectrum()
cls.ps.freq = freq
cls.ps.power = power
cls.ps.df = cls.ps.freq[1] - cls.ps.freq[0]
cls.ps.m = 1
cls.t0 = np.asarray(
[200.0, 0.5, 0.1, 100.0, 2.0, 1.0, 50.0, 7.5, 0.5, 2.0])
cls.parest, cls.res = fit_lorentzians(cls.ps, cls.nlor, cls.t0)
def abtov2v3(channel, **kwargs):
# Construct the reference data model in general JWST imager type
input_model = datamodels.ImageModel()
# Convert input of type '1A' into the band and channel that pipeline needs
theband, thechan = bandchan(channel)
# Set the filter in the data model meta header
input_model.meta.instrument.band = theband
input_model.meta.instrument.channel = thechan
# If passed input refs keyword, unpack and use it
if ('refs' in kwargs):
therefs = kwargs['refs']
# Otherwise use default reference files
else:
therefs = setreffiles_cdp6(channel)
# The pipeline transform actually uses the triple
# (alpha,beta,lambda) -> (v2,v3,lambda)
basedistortion = miri.abl_to_v2v3l(input_model, therefs)
# At present, the pipeline uses v2,v3 in degrees so convert to arcsec
distortion = basedistortion | models.Scale(3600.) & models.Scale(
3600.) & models.Identity(1)
# Therefore we need to hack a reasonable wavelength onto our input, run transform,
# then hack it back off again
thewave = midwave(channel)
# Duplicate the beta value at first, then replace with wavelength value
map = models.Mapping(
(0, 1, 1
)) | models.Identity(1) & models.Identity(1) & models.Const1D(thewave)
map.inverse = models.Mapping((0, 1), n_inputs=3)
allmap = map | distortion | map.inverse
allmap.inverse = map | distortion.inverse | map.inverse
# Return the distortion object that can then be queried
return allmap
def setup_class(cls):
m = 1
nfreq = 100000
freq = np.arange(nfreq)
noise = np.random.exponential(size=nfreq)
power = noise * 2.0
ps = Powerspectrum()
ps.freq = freq
ps.power = power
ps.m = m
ps.df = freq[1] - freq[0]
ps.norm = "leahy"
cls.ps = ps
cls.a_mean, cls.a_var = 2.0, 1.0
cls.model = models.Const1D()
p_amplitude = lambda amplitude: \
scipy.stats.norm(loc=cls.a_mean, scale=cls.a_var).pdf(amplitude)
cls.priors = {"amplitude": p_amplitude}
cls.lpost = PSDPosterior(cls.ps.freq,
cls.ps.power,
cls.model,
m=cls.ps.m)
cls.lpost.logprior = set_logprior(cls.lpost, cls.priors)
cls.fitmethod = "BFGS"
cls.max_post = True
cls.t0 = [2.0]
cls.neg = True
cls.opt = scipy.optimize.minimize(cls.lpost,
cls.t0,
method=cls.fitmethod,
args=cls.neg,
tol=1.e-10)
def setup_class(cls):
m = 1
nfreq = 1000000
freq = np.arange(nfreq)
noise = np.random.exponential(size=nfreq)
power = noise * 2.0
ps = Powerspectrum()
ps.freq = freq
ps.power = power
ps.m = m
ps.df = freq[1] - freq[0]
ps.norm = "leahy"
cls.ps = ps
cls.a_mean, cls.a_var = 2.0, 1.0
cls.model = models.Const1D()
p_amplitude = lambda amplitude: \
scipy.stats.norm(loc=cls.a_mean, scale=cls.a_var).pdf(amplitude)
cls.priors = {"amplitude": p_amplitude}
def test_fitting_backend():
np.random.seed(42)
x, y = build_spectrum()
spectrum = Spectrum1D(flux=y * u.Jy, spectral_axis=x * u.um)
g1f = models.Gaussian1D(0.7 * u.Jy, 4.65 * u.um, 0.3 * u.um, name='g1')
g2f = models.Gaussian1D(2.0 * u.Jy, 5.55 * u.um, 0.3 * u.um, name='g2')
g3f = models.Gaussian1D(-2. * u.Jy, 8.15 * u.um, 0.2 * u.um, name='g3')
zero_level = models.Const1D(1. * u.Jy, name='const1d')
model_list = [g1f, g2f, g3f, zero_level]
expression = "g1 + g2 + g3 + const1d"
# Returns the initial model
fm, fitted_spectrum = fb.fit_model_to_spectrum(spectrum,
model_list,
expression,
run_fitter=False)
parameters_expected = np.array(
[0.7, 4.65, 0.3, 2., 5.55, 0.3, -2., 8.15, 0.2, 1.])
assert np.allclose(fm.parameters, parameters_expected, atol=1e-5)
# Returns the fitted model
fm, fitted_spectrum = fb.fit_model_to_spectrum(spectrum,
model_list,
expression,
run_fitter=True)
parameters_expected = np.array([
1.0104705, 4.58956282, 0.19590464, 2.39892026, 5.49867754, 0.10834472,
-1.66902953, 8.19714439, 0.09535613, 3.99125545
])
assert np.allclose(fm.parameters, parameters_expected, atol=1e-5)
def setup_class(cls):
cls.m = 10
nfreq = 1000000
freq = np.arange(nfreq)
noise = scipy.stats.chi2(2. * cls.m).rvs(size=nfreq) / np.float(cls.m)
power = noise
ps = Powerspectrum()
ps.freq = freq
ps.power = power
ps.m = cls.m
ps.df = freq[1] - freq[0]
ps.norm = "leahy"
cls.ps = ps
cls.a_mean, cls.a_var = 2.0, 1.0
cls.model = models.Const1D()
p_amplitude = lambda amplitude: \
scipy.stats.norm(loc=cls.a_mean, scale=cls.a_var).pdf(amplitude)
cls.priors = {"amplitude": p_amplitude}
def test_simulate_highest_outlier_works(self):
m = 1
nfreq = 100000
seed = 100
freq = np.linspace(1, 10, nfreq)
rng = np.random.RandomState(seed)
noise = rng.exponential(size=nfreq)
model = models.Const1D()
model.amplitude = 2.0
p = model(freq)
power = noise * p
ps = Powerspectrum()
ps.freq = freq
ps.power = power
ps.m = m
ps.df = freq[1] - freq[0]
ps.norm = "leahy"
nsim = 10
loglike = PSDLogLikelihood(ps.freq, ps.power, model, m=1)
s_all = np.atleast_2d(np.ones(nsim) * 2.0).T
pe = PSDParEst(ps)
res = pe.fit(loglike, [2.0], neg=True)
maxpow_sim = pe.simulate_highest_outlier(s_all,
loglike, [2.0],
max_post=False,
seed=seed)
assert maxpow_sim.shape[0] == nsim
assert np.all(maxpow_sim > 20.00) and np.all(maxpow_sim < 31.0)
def create_channel_selector(alpha, lam, channel, beta):
if channel == 1:
nslice = 21
elif channel == 2:
nslice = 17
elif channel == 3:
nslice = 16
elif channel == 4:
nslice = 12
else:
raise ValueError("Incorrect channel #")
ind = []
for i in range(5):
for j in range(5):
ind.append((i, j))
selector = {}
# In the paper the formula is (x-xs)^j*y^i, so the 'x' corresponds
# to y in modeling. - swapped in Mapping
axs = alpha.field('x_s')
lxs = lam.field('x_s')
for i in range(nslice):
ashift = models.Shift(axs[i])
lshift = models.Shift(lxs[i])
palpha = models.Polynomial2D(8)
plam = models.Polynomial2D(8)
for index, coeff in zip(ind, alpha[i][1:]):
setattr(palpha, 'c{0}_{1}'.format(index[0], index[1]), coeff)
for index, coeff in zip(ind, lam[i][1:]):
setattr(plam, 'c{0}_{1}'.format(index[0], index[1]), coeff)
alpha_model = ashift & models.Identity(1) | palpha
lam_model = lshift & models.Identity(1) | plam
beta_model = models.Const1D(beta[0] + (i - 1) * beta[1])
selector[i] = models.Mapping((1, 0, 1, 0, 0)) | alpha_model & lam_model & beta_model
return selector
# We will search for pulsations over a range of frequencies around the known pulsation period.
df = (period_ranges[1] - period_ranges[0]) / period_bins
frequencies = 1 / np.arange(period_ranges[0], period_ranges[1], df)
freq, efstat = epoch_folding_search(times, frequencies, nbin=nbin)
pulse_frequency = freq[np.where(efstat == max(efstat))[0][0]]
print('pulse frequency', pulse_frequency)
_ = write_files('pulse_frequency_NuSTAR', pulse_frequency)
#fitting epoch folding distribution with Lorentzian curve
g_init = models.Lorentz1D(
amplitude=max(efstat) - min(efstat),
x_0=pulse_frequency,
fwhm=pulse_frequency / 500) + models.Const1D(amplitude=min(efstat))
fit_g = fitting.LevMarLSQFitter()
bin_max = [np.where(efstat == max(efstat))[0][0]][0]
bin_left_min = 0
bin_right_min = len(efstat)
for j in range(min(bin_max, len(efstat) - bin_max) - 2):
d_efstat = efstat[bin_max - j - 1] - efstat[bin_max - j]
if d_efstat > 0 and max(efstat) - efstat[bin_max - j] > (max(efstat) -
min(efstat)) / 2:
bin_left_min = bin_max - j
break
for j in range(min(bin_max, len(efstat) - bin_max) - 2):
d_efstat = efstat[bin_max + j + 1] - efstat[bin_max + j]
if d_efstat > 0 and max(efstat) - efstat[bin_max + j] > (max(efstat) -
min(efstat)) / 2:
def test_LabelMapper():
transform = models.Const1D(12.3)
lm = selector.LabelMapper(inputs=('x', 'y'), mapper=transform, inputs_mapping=(1,))
x = np.linspace(3, 11, 20)
assert_allclose(lm(x, x), transform(x))
def test_cube_fitting_backend():
np.random.seed(42)
SIGMA = 0.1 # noise in data
TOL = 0.4 # test tolerance
# Flux cube oriented as in JWST data. To build a Spectrum1D
# instance with this, one need to transpose it so the spectral
# axis direction corresponds to the last index.
flux_cube = np.zeros((SPECTRUM_SIZE, IMAGE_SIZE, IMAGE_SIZE))
# Generate list of all spaxels to be fitted
_spx = [[(x, y) for x in range(IMAGE_SIZE)] for y in range(IMAGE_SIZE)]
spaxels = [item for sublist in _spx for item in sublist]
# Fill cube spaxels with spectra that differ from
# each other only by their noise component.
x, _ = build_spectrum()
for spx in spaxels:
flux_cube[:, spx[0], spx[1]] = build_spectrum(sigma=SIGMA)[1]
# Transpose so it can be packed in a Spectrum1D instance.
flux_cube = flux_cube.transpose(1, 2, 0)
spectrum = Spectrum1D(flux=flux_cube*u.Jy, spectral_axis=x*u.um)
# Initial model for fit.
g1f = models.Gaussian1D(0.7*u.Jy, 4.65*u.um, 0.3*u.um, name='g1')
g2f = models.Gaussian1D(2.0*u.Jy, 5.55*u.um, 0.3*u.um, name='g2')
g3f = models.Gaussian1D(-2.*u.Jy, 8.15*u.um, 0.2*u.um, name='g3')
zero_level = models.Const1D(1.*u.Jy, name='const1d')
model_list = [g1f, g2f, g3f, zero_level]
expression = "g1 + g2 + g3 + const1d"
# Fit to all spaxels.
fitted_parameters, fitted_spectrum = fb.fit_model_to_spectrum(
spectrum, model_list, expression)
# Check that parameter results are formatted as expected.
assert type(fitted_parameters) == list
assert len(fitted_parameters) == 225
for m in fitted_parameters:
if m['x'] == 3 and m['y'] == 2:
fitted_model = m['model']
assert type(fitted_model[0].amplitude.value) == np.float64
assert fitted_model[0].amplitude.unit == u.Jy
assert type(fitted_model[0] == params.Parameter)
assert type(fitted_model[0].mean.value) == np.float64
assert fitted_model[0].mean.unit == u.um
# Check that spectrum result is formatted as expected.
assert type(fitted_spectrum) == Spectrum1D
assert len(fitted_spectrum.shape) == 3
assert fitted_spectrum.shape == (IMAGE_SIZE, IMAGE_SIZE, SPECTRUM_SIZE)
assert fitted_spectrum.flux.unit == u.Jy
# The important point here isn't to check the accuracy of the
# fit, which was already tested elsewhere. We are mostly
# interested here in checking the correctness of the data
# packaging into the output products.
assert np.allclose(fitted_model[0].amplitude.value, 1.09, atol=TOL)
assert np.allclose(fitted_model[1].amplitude.value, 2.4, atol=TOL)
assert np.allclose(fitted_model[2].amplitude.value, -1.7, atol=TOL)
assert np.allclose(fitted_model[0].mean.value, 4.6, atol=TOL)
assert np.allclose(fitted_model[1].mean.value, 5.5, atol=TOL)
assert np.allclose(fitted_model[2].mean.value, 8.2, atol=TOL)
assert np.allclose(fitted_model[0].stddev.value, 0.2, atol=TOL)
assert np.allclose(fitted_model[1].stddev.value, 0.1, atol=TOL)
assert np.allclose(fitted_model[2].stddev.value, 0.1, atol=TOL)
assert np.allclose(fitted_model[3].amplitude.value, 4.0, atol=TOL)
def test_const1d(tmpdir, standard_version):
helpers.assert_roundtrip_tree({"model": astmodels.Const1D(amplitude=5.)},
tmpdir,
init_options={"version": standard_version})
astmodels.RotateNative2Celestial(5.63, -72.5, 180),
astmodels.Multiply(3),
astmodels.Multiply(10 * u.m),
astmodels.RotateCelestial2Native(5.63, -72.5, 180),
astmodels.EulerAngleRotation(23, 14, 2.3, axes_order='xzx'),
astmodels.Mapping((0, 1), n_inputs=3),
astmodels.Shift(2. * u.deg),
astmodels.Scale(3.4 * u.deg),
astmodels.RotateNative2Celestial(5.63 * u.deg, -72.5 * u.deg, 180 * u.deg),
astmodels.RotateCelestial2Native(5.63 * u.deg, -72.5 * u.deg, 180 * u.deg),
astmodels.RotationSequence3D([1.2, 2.3, 3.4, .3], 'xyzx'),
astmodels.SphericalRotationSequence([1.2, 2.3, 3.4, .3], 'xyzy'),
astmodels.AiryDisk2D(amplitude=10., x_0=0.5, y_0=1.5),
astmodels.Box1D(amplitude=10., x_0=0.5, width=5.),
astmodels.Box2D(amplitude=10., x_0=0.5, x_width=5., y_0=1.5, y_width=7.),
astmodels.Const1D(amplitude=5.),
astmodels.Const2D(amplitude=5.),
astmodels.Disk2D(amplitude=10., x_0=0.5, y_0=1.5, R_0=5.),
astmodels.Ellipse2D(amplitude=10., x_0=0.5, y_0=1.5, a=2., b=4.,
theta=0.1),
astmodels.Exponential1D(amplitude=10., tau=3.5),
astmodels.Gaussian1D(amplitude=10., mean=5., stddev=3.),
astmodels.Gaussian2D(amplitude=10.,
x_mean=5.,
y_mean=5.,
x_stddev=3.,
y_stddev=3.),
astmodels.KingProjectedAnalytic1D(amplitude=10., r_core=5., r_tide=2.),
astmodels.Logarithmic1D(amplitude=10., tau=3.5),
astmodels.Lorentz1D(amplitude=10., x_0=0.5, fwhm=2.5),
astmodels.Moffat1D(amplitude=10., x_0=0.5, gamma=1.2, alpha=2.5),
'Trapezoid1D':
models.Trapezoid1D(1.0, 1.0, 1.0, 1.0),
'Moffat1D':
models.Moffat1D(1.0, 1.0, 1.0, 1.0),
'ExponentialCutoffPowerLaw1D':
models.ExponentialCutoffPowerLaw1D(1.0, 1.0, 1.0, 1.0),
'BrokenPowerLaw1D':
models.BrokenPowerLaw1D(1.0, 1.0, 1.0, 1.0),
'LogParabola1D':
models.LogParabola1D(1.0, 1.0, 1.0, 1.0),
'PowerLaw1D':
models.PowerLaw1D(1.0, 1.0, 1.0),
'Linear1D':
models.Linear1D(1.0, 0.0),
'Const1D':
models.Const1D(0.0),
'Redshift':
models.Redshift(0.0),
'Scale':
models.Scale(1.0),
'Shift':
models.Shift(0.0),
'Sine1D':
models.Sine1D(1.0, 1.0),
'Chebyshev1D':
models.Chebyshev1D(1),
'Legendre1D':
models.Legendre1D(1),
'Polynomial1D':
models.Polynomial1D(1),
}
def fit_lorentzians(ps,
nlor,
starting_pars,
fit_whitenoise=True,
max_post=False,
priors=None,
fitmethod="L-BFGS-B"):
"""
Fit a number of Lorentzians to a power spectrum, possibly including white
noise. Each Lorentzian has three parameters (amplitude, centroid position,
full-width at half maximum), plus one extra parameter if the white noise
level should be fit as well. Priors for each parameter can be included in
case `max_post = True`, in which case the function will attempt a
Maximum-A-Posteriori fit. Priors must be specified as a dictionary with one
entry for each parameter.
The parameter names are `(amplitude_i, x_0_i, fwhm_i)` for each `i` out of
a total of `N` Lorentzians. The white noise level has a parameter
`amplitude_(N+1)`. For example, a model with two Lorentzians and a
white noise level would have parameters:
[amplitude_0, x_0_0, fwhm_0, amplitude_1, x_0_1, fwhm_1, amplitude_2].
Parameters
----------
ps : Powerspectrum
A Powerspectrum object with the data to be fit
nlor : int
The number of Lorentzians to fit
starting_pars : iterable
The list of starting guesses for the optimizer. See explanation above
for ordering of parameters in this list.
fit_whitenoise : bool, optional, default True
If True, the code will attempt to fit a white noise level along with
the Lorentzians. Be sure to include a starting parameter for the
optimizer in `starting_pars`!
max_post : bool, optional, default False
If True, perform a Maximum-A-Posteriori fit of the data rather than a
Maximum Likelihood fit. Note that this requires priors to be specified,
otherwise this will cause an exception!
priors : {dict | None}, optional, default None
Dictionary with priors for the MAP fit. This should be of the form
{"parameter name": probability distribution, ...}
fitmethod : string, optional, default "L-BFGS-B"
Specifies an optimization algorithm to use. Supply any valid option for
`scipy.optimize.minimize`.
Returns
-------
parest : PSDParEst object
A PSDParEst object for further analysis
res : OptimizationResults object
The OptimizationResults object storing useful results and quantities
relating to the fit
Example
-------
We start by making an example power spectrum with three Lorentzians
>>> np.random.seed(400)
>>> nlor = 3
>>> x_0_0 = 0.5
>>> x_0_1 = 2.0
>>> x_0_2 = 7.5
>>> amplitude_0 = 150.0
>>> amplitude_1 = 50.0
>>> amplitude_2 = 15.0
>>> fwhm_0 = 0.1
>>> fwhm_1 = 1.0
>>> fwhm_2 = 0.5
We will also include a white noise level:
>>> whitenoise = 2.0
>>> model = models.Lorentz1D(amplitude_0, x_0_0, fwhm_0) + \\
... models.Lorentz1D(amplitude_1, x_0_1, fwhm_1) + \\
... models.Lorentz1D(amplitude_2, x_0_2, fwhm_2) + \\
... models.Const1D(whitenoise)
>>> freq = np.linspace(0.01, 10.0, 10.0/0.01)
>>> p = model(freq)
>>> noise = np.random.exponential(size=len(freq))
>>> power = p*noise
>>> ps = Powerspectrum()
>>> ps.freq = freq
>>> ps.power = power
>>> ps.df = ps.freq[1] - ps.freq[0]
>>> ps.m = 1
Now we have to guess starting parameters. For each Lorentzian, we have
amplitude, centroid position and fwhm, and this pattern repeats for each
Lorentzian in the fit. The white noise level is the last parameter.
>>> t0 = [150, 0.4, 0.2, 50, 2.3, 0.6, 20, 8.0, 0.4, 2.1]
We're ready for doing the fit:
>>> parest, res = fit_lorentzians(ps, nlor, t0)
`res` contains a whole array of useful information about the fit, for
example the parameters at the optimum:
>>> p_opt = res.p_opt
"""
model = models.Lorentz1D()
if nlor > 1:
for i in range(nlor - 1):
model += models.Lorentz1D()
if fit_whitenoise:
model += models.Const1D()
return fit_powerspectrum(ps,
model,
starting_pars,
max_post=max_post,
priors=priors,
fitmethod=fitmethod)
