def test_model_image(imsize, nproc, count, chunk=8):
psf = None
model_orig = create_model()
imfit = Imfit(model_orig,
psf=psf,
quiet=True,
nproc=nproc,
chunk_size=chunk)
shape = (imsize, imsize)
imfit.getModelImage(shape)
t1 = time.time()
imfit._modelObject._testCreateModelImage(count)
return time.time() - t1
def fit_spectra(full_spectra, full_noise, flux_unit, init_model):
PSF = moffat_psf(fwhm=2.4, size=9)
fitter = Imfit(init_model, PSF, quiet=True)
models = []
for i in xrange(0, full_spectra.shape[-1]):
# print '#'*50, i
f__l = full_spectra[..., i] / flux_unit
err__l = full_noise[..., i] / flux_unit
fitter.fit(f__l, err__l)
if fitter.nValidPixels < 1000:
print '(%d) Too many masked pixels.' % i
if fitter.nPegged < 0:
print '(%d) Pegged parameters.' % i
m = fitter.getModelDescription()
# print m.bulge.I_e.value, m.bulge.r_e.value, m.disk.I_0.value, m.disk.h.value
models.append(m)
return models
def fit_image(flux, noise, guess_model, PSF=None,
mode='LM', insist=False, quiet=True, nproc=None, use_cash_statistics=False):
'''
Doc me!
'''
imfit = Imfit(guess_model, PSF, quiet=quiet, nproc=nproc)
imfit.fit(flux, noise,
mode=mode,
error_type='sigma',
use_cash_statistics=(mode != 'LM' and use_cash_statistics))
logger.debug('Valid pix: %d | Iter.: %d | pegged: %d | fit stat.: %f | BIC: %f' % \
(imfit.nValidPixels, imfit.nIter, imfit.nPegged, imfit.reducedFitStatistic, imfit.BIC))
fitted_model = imfit.getModelDescription()
logger.debug(' Fitted model:\n%s\n\n' % str(fitted_model))
if not imfit.fitConverged or imfit.nPegged > 0:
logger.warn('Bad fit: did not converge or pegged parameter.')
if mode == 'LM' and insist:
logger.warn(' Retrying using N-M simplex.')
imfit.fit(flux, noise,
mode='NM',
error_type='sigma',
use_cash_statistics=use_cash_statistics)
fitted_model = imfit.getModelDescription()
if imfit.fitConverged:
logger.debug(' N-M simplex fit statistic: %f | BIC: %f' % (imfit.reducedFitStatistic, imfit.BIC))
else:
logger.warn(' Bad fit: N-M simplex did not converge.')
logger.debug(' Fitted model:\n%s\n\n' % str(fitted_model))
return fitted_model, imfit.fitConverged, imfit.reducedFitStatistic
def test_fitting():
psf = gaussian_psf(2.5, size=9)
model_orig = create_model()
imfit = Imfit(model_orig, psf=psf, quiet=False)
noise_level = 0.1
shape = (100, 100)
image = imfit.getModelImage(shape)
noise = image * noise_level
image += (np.random.random(shape) * noise)
mask = np.zeros_like(image, dtype='bool')
imfit.fit(image, noise, mask)
model_fitted = imfit.getModelDescription()
orig_params = get_model_param_array(model_orig)
fitted_params = get_model_param_array(model_fitted)
assert_allclose(orig_params, fitted_params, rtol=noise_level)
def get_images(db):
K = fitsQ3DataCube(db, smooth=False)
mask = ~K.qMask
qSignal = np.ma.array(K.qSignal, mask=mask)
qNoise = np.ma.array(K.qNoise, mask=mask)
# PA, ba = K.getEllipseParams()
model = galaxy_model(K.x0, K.y0, bulge=True, disk=True,
I_e=1, r_e=12, n=1.5, PA_b=90.0, ell_b=0.5,
I_0=1, h=12, PA_d=90, ell_d=0.5)
PSF = moffat_psf(fwhm=2.4, size=9)
fitter = Imfit(model, PSF, quiet=True)
fitter.fit(qSignal, qNoise)
model_image = fitter.getModelImage()
model = fitter.getModelDescription()
bulge_image, disk_image = create_model_images(qSignal.shape, PSF, model)
bulge_image = np.ma.array(bulge_image, mask=mask)
disk_image = np.ma.array(disk_image, mask=mask)
bulge_frac = bulge_image / model_image
disk_frac = disk_image / model_image
bulge_noise = np.sqrt(bulge_frac) * qNoise
disk_noise = np.sqrt(disk_frac) * qNoise
return model, qSignal, qNoise, model_image, bulge_image, bulge_noise, disk_image, disk_noise
def test_fitting():
psf = gaussian_psf(2.5, size=9)
model_orig = create_model()
imfit = Imfit(model_orig, psf=psf, quiet=False)
noise_level = 0.1
shape = (100, 100)
image = imfit.getModelImage(shape)
noise = image * noise_level
image += (np.random.random(shape) * noise)
mask = np.zeros_like(image, dtype='bool')
imfit.fit(image, noise, mask)
model_fitted = imfit.getModelDescription()
orig_params = get_model_param_array(model_orig)
fitted_params = get_model_param_array(model_fitted)
assert_allclose(orig_params, fitted_params, rtol=noise_level)
model.x0.setLimitsRel(5, 5)
model.y0.setValue(flux[0].shape[0] / 2)
model.y0.setLimitsRel(5, 5)
model.addFunction(psf_func)
print 'Initial PSF model:'
print model
good = []
chi2 = []
flag = []
fitmodel = []
modelflux = []
for i, f in enumerate(flux):
star = path.basename(cubes[i])
print 'Fitting PSF %d of %d (%s)' % (i, len(flux), star)
imfit = Imfit(model, quiet=True)
imfit.fit(f, mode='NM', use_model_for_errors=False, use_cash_statistics=False)
_goodfraction = float(imfit.nValidPixels) / float(f.size)
print ' Good pixels: %.1f %%' % (_goodfraction * 100)
good.append(_goodfraction)
print ' Fit converged? %s' % imfit.fitConverged
flag.append(not imfit.fitConverged)
print ' Fit statistic: %f' % imfit.fitStatistic
chi2.append(imfit.fitStatistic)
_fitmodel = imfit.getModelDescription()
fitmodel.append(_fitmodel)
model_f = imfit.getModelImage()
modelflux.append(model_f)
imradius = 10
def rad_prof(model, bins):
imfit = Imfit(model, quiet=False)
image = imfit.getModelImage((model.y0.value * 2, model.x0.value * 2))
return radialProfile(image, bins, x0, y0, pa=0.0, ba=1.0)
def create_image(shape, PSF, model):
imfit = Imfit(model, PSF, quiet=True)
image = imfit.getModelImage(shape)
return image
def model_image(model, shape, PSF=None, nproc=None):
imfit = Imfit(model, PSF, quiet=True, nproc=nproc)
return imfit.getModelImage(shape)
model.y0.setLimitsRel(3, 3)
model.addFunction(psf_func)
print 'Initial PSF model:'
print model
psfmodels = []
psfflags = []
goodfraction = []
chi2 = []
for i in xrange(psfflux.shape[0]):
wl = psf_wl[i]
pf = psfflux[i] - psfbg[i]
pf[pf <= 0] = np.ma.masked
print 'Fitting PSF %d of %d (%s \AA)' % (i, fill_flux.shape[0], wl)
imfit = Imfit(model, quiet=True)
imfit.fit(pf, mode='NM', use_model_for_errors=False)
_goodfraction = float(imfit.nValidPixels) / float(psfflux[i].size)
goodfraction.append(_goodfraction)
flagged = not imfit.fitConverged or _goodfraction < 0.5
psfflags.append(flagged)
print ' Fit statistic: %f' % imfit.fitStatistic
print ' Reduced fit statistic: %f' % imfit.reducedFitStatistic
print ' Valid pix: %.1f %%' % (_goodfraction * 100.0)
print ' Flagged? %s' % flagged
chi2.append(imfit.fitStatistic)
fitmodel = imfit.getModelDescription()
modelimage = imfit.getModelImage()
psfmodels.append(fitmodel)
print fitmodel