from sklearn.manifold import LocallyLinearEmbedding
from astroML.datasets import fetch_sdss_specgals
from astroML.datasets import fetch_sdss_spectrum
data = fetch_sdss_specgals()
print data.dtype.names
ngals = 326
nwavel = 3855
plates = data['plate'][:ngals]
mjds = data['mjd'][:ngals]
fiberIDs = data['fiberID'][:ngals]
h_alpha = data['h_alpha_flux'][:ngals]
bptclass = data['bptclass'][:ngals]
specdata = np.zeros((ngals, nwavel))
i = 0
for plate, mjd, fiberID in zip(plates, mjds, fiberIDs):
tempdata = fetch_sdss_spectrum(plate, mjd, fiberID)
specdata[i, :] = tempdata.spectrum/tempdata.spectrum.mean()
i += 1
# Apply LLE
k = 7
for fignum, n in enumerate([2, 3]):
lle = LocallyLinearEmbedding(k, n)
lle.fit(specdata)
proj = lle.transform(specdata)
pl.subplot(2, 1, fignum+1)
pl.scatter(proj[:,0], proj[:,1], c=bptclass, s=50)
pl.colorbar()
pl.show()
the SDSS server for the given plate, fiber, and mjd, downloads the spectrum,
and plots the result.
"""
# Author: Jake VanderPlas <*****@*****.**>
# License: BSD
# The figure is an example from astroML: see http://astroML.github.com
from matplotlib import pyplot as plt
from astroML.datasets import fetch_sdss_spectrum
#------------------------------------------------------------
# Fetch single spectrum
plate = 1615
mjd = 53166
fiber = 513
spec = fetch_sdss_spectrum(plate, mjd, fiber)
#------------------------------------------------------------
# Plot the resulting spectrum
ax = plt.axes()
ax.plot(spec.wavelength(), spec.spectrum, '-k', label='spectrum')
ax.plot(spec.wavelength(), spec.error, '-', color='gray', label='error')
ax.legend(loc=4)
ax.set_title('Plate = %(plate)i, MJD = %(mjd)i, Fiber = %(fiber)i' % locals())
ax.text(0.05, 0.95, 'z = %.2f' % spec.z, size=16,
ha='left', va='top', transform=ax.transAxes)
ax.set_xlabel(r'$\lambda (\AA)$')
def fetch_and_shift_spectra(n_spectra,
outfile,
primtarget=TARGET_GALAXY,
zlim=(0, 0.7),
loglam_start=3.5,
loglam_end=3.9,
Nlam=1000):
"""
This function queries CAS for matching spectra, and then downloads
them and shifts them to a common redshift binning
"""
# First query for the list of spectra to download
plate, mjd, fiber = query_plate_mjd_fiber(n_spectra, primtarget,
zlim[0], zlim[1])
# Set up arrays to hold information gathered from the spectra
spec_cln = np.zeros(n_spectra, dtype=np.int32)
lineindex_cln = np.zeros(n_spectra, dtype=np.int32)
log_NII_Ha = np.zeros(n_spectra, dtype=np.float32)
log_OIII_Hb = np.zeros(n_spectra, dtype=np.float32)
z = np.zeros(n_spectra, dtype=np.float32)
zerr = np.zeros(n_spectra, dtype=np.float32)
spectra = np.zeros((n_spectra, Nlam), dtype=np.float32)
mask = np.zeros((n_spectra, Nlam), dtype=np.bool)
# Calculate new wavelength coefficients
new_coeff0 = loglam_start
new_coeff1 = (loglam_end - loglam_start) / Nlam
# Now download all the needed spectra, and resample to a common
# wavelength bin.
n_spectra = len(plate)
num_skipped = 0
i = 0
while i < n_spectra:
sys.stdout.write(' %i / %i spectra\r' % (i + 1, n_spectra))
sys.stdout.flush()
try:
spec = fetch_sdss_spectrum(plate[i], mjd[i], fiber[i])
except HTTPError:
num_skipped += 1
print("%i, %i, %i not found" % (plate[i], mjd[i], fiber[i]))
i += 1
continue
spec_rebin = spec.restframe().rebin(new_coeff0, new_coeff1, Nlam)
if np.all(spec_rebin.spectrum == 0):
num_skipped += 1
print("%i, %i, %i is all zero" % (plate[i], mjd[i], fiber[i]))
continue
spec_cln[i] = spec.spec_cln
lineindex_cln[i], (log_NII_Ha[i], log_OIII_Hb[i])\
= spec.lineratio_index()
z[i] = spec.z
zerr[i] = spec.zerr
spectra[i] = spec_rebin.spectrum
mask[i] = spec_rebin.compute_mask(0.5, 5)
i += 1
sys.stdout.write('\n')
N = i
print(" %i spectra skipped" % num_skipped)
print(" %i spectra processed" % N)
print("saving to %s" % outfile)
np.savez(outfile,
spectra=spectra[:N],
mask=mask[:N],
coeff0=new_coeff0,
coeff1=new_coeff1,
spec_cln=spec_cln[:N],
lineindex_cln=lineindex_cln[:N],
log_NII_Ha=log_NII_Ha[:N],
log_OIII_Hb=log_OIII_Hb[:N],
z=z[:N],
zerr=zerr[:N])
def plot_image_spec_sdss_galaxy(sdss_obj, save_figure=False):
#plot image and spectrum of a specified SDSS galaxy
# input sdss_obj = individual SDSS main galaxy data base entry
# save_figure specifies whether PDF of the figure will be saved in fig/ subdirectory for the record
#
# define plot with 2 horizonthal panels of appropriate size
fig, (ax0, ax1) = plt.subplots(1, 2, figsize=(6, 2.))
# set an appropriate font size for labels
plt.rc('font',size=8)
#get RA and DEC of the galaxy and form the filename to save SDSS image
RA = sdss_obj['ra']; DEC = sdss_obj['dec']; scale=0.25
outfile = image_home_dir()+str(sdss_obj['objID'])+'.jpg'
fetch_sdss_image(outfile, RA, DEC, scale)
img = Image.open(outfile)
# do not plot pixel axis labels
ax0.axis('off')
ax0.imshow(img)
# fetch SDSS spectrum using plate number, epoch, and fiber ID
plate = sdss_obj['plate']; mjd = sdss_obj['mjd']; fiber = sdss_obj['fiberID']
spec = fetch_sdss_spectrum(plate, mjd, fiber)
# normalize spectrum for plotting
spectrum = 0.5 * spec.spectrum / spec.spectrum.max()
lam = spec.wavelength()
text_kwargs = dict(ha='center', va='center', alpha=0.5, fontsize=10)
# set axis limits for spectrum plot
ax1.set_xlim(3000, 10000)
ax1.set_ylim(0, 0.6)
color = np.zeros(5)
for i, f, c, loc in zip([0,1,2,3,4],'ugriz', 'bgrmk', [3500, 4600, 6100, 7500, 8800]):
data_home = setup.sdss_filter_dir()
archive_file = os.path.join(data_home, '%s.dat' % f)
if not os.path.exists(archive_file):
raise ValueError("Error in plot_img_spec_sdss_galaxy: filter file '%s' does not exist!" % archive_file )
F = open(archive_file)
filt = np.loadtxt(F, unpack=True)
ax1.fill(filt[0], filt[2], ec=c, fc=c, alpha=0.4)
fsp = interp1d(filt[0],filt[2], bounds_error=False, fill_value=0.0)
ax1.text(loc, 0.03, f, color=c, **text_kwargs)
# compute magnitude in each band using simple Simpson integration of spectrum and filter using eq 1.2 in the notes
fspn = lam*fsp(lam)/simps(fsp(lam)/lam,lam)
lamf = fspn; #lamf = lamf.clip(0.0)
specf = spec.spectrum * lamf
color[i] = -2.5 * np.log10(simps(specf,lam))
# print estimated g-r color for the object
grcatcol = sdss_obj['modelMag_g']-sdss_obj['modelMag_r']
print "computed g-r = ", color[1]-color[2]
print "catalog g-r=", grcatcol
ax1.plot(lam, spectrum, '-k', lw=0.5, label=r'$(g-r)=%.2f$'%grcatcol )
ax1.set_xlabel(r'$\lambda {(\rm \AA)}$')
ax1.set_ylabel(r'$\mathrm{norm.\ specific\ flux}\ \ \ f_{\lambda}$')
#ax.set_title('Plate = %(plate)i, MJD = %(mjd)i, Fiber = %(fiber)i' % locals())
#ax.set_title('%s' % objects[obj_]['name'])
ax1.legend(frameon=False)
if save_figure:
plt.savefig('fig/gal_img_spec'+'_'+str(sdss_obj['objID'])+'.pdf',bbox_inches='tight')
plt.subplots_adjust(wspace = 0.2)
plt.show()
return
primtarget=TARGET_GALAXY_RED
zlim=(0.2, 0.6)
plate, mjd, fiber = query_plate_mjd_fiber(100, primtarget, zlim[0], zlim[1])
agg = SpecMeanAggregator()
lam = agg.lam
zdist = []
for plate_n, mjd_n, fiber_n in zip(plate, mjd, fiber):
sys.stdout.write("{0}.{1}.{2} \r".format(plate_n, mjd_n, fiber_n))
sys.stdout.flush()
spec = fetch_sdss_spectrum(plate_n, mjd_n, fiber_n)
zdist.append(spec.z)
agg.add_spec(spec)
sys.stdout.write('\n')
spec, dspec = agg.reduce()
fig, ax = plt.subplots(2, 1, figsize=(8, 8))
ax[0].plot(lam, spec, '-k')
#ax[0].fill_between(lam, spec - dspec, spec + dspec,
# color='#AAAAAA', alpha=0.3)
ax[0].set_xlim(2500, 7500)
ax[0].set_ylim(0, 0.7)
ax[0].set_xlabel(r'$\lambda')
#---------------------------------------------------------------------- # This function adjusts matplotlib settings for a uniform feel in the textbook. # Note that with usetex=True, fonts are rendered with LaTeX. This may # result in an error if LaTeX is not installed on your system. In that case, # you can set usetex to False. from astroML.plotting import setup_text_plots setup_text_plots(fontsize=8, usetex=True) #------------------------------------------------------------ # Fetch the spectrum from SDSS database & pre-process plate = 659 mjd = 52199 fiber = 381 data = fetch_sdss_spectrum(plate, mjd, fiber) lam = data.wavelength() spec = data.spectrum # wavelengths are logorithmically spaced: we'll work in log(lam) loglam = np.log10(lam) flag = (lam > 4000) & (lam < 5000) lam = lam[flag] loglam = loglam[flag] spec = spec[flag] lam = lam[:-1] loglam = loglam[:-1] spec = spec[:-1]
z_galaxies = d_galaxies['z']
modelMag_u_galaxies = d_galaxies['modelMag_u']
modelMag_i_galaxies = d_galaxies['modelMag_i']
modelMag_r_galaxies = d_galaxies['modelMag_r']
modelMag_g_galaxies = d_galaxies['modelMag_g']
properties = [h_alpha_flux_galaxies,extinction_r_galaxies,velDisp_galaxies,z_galaxies,\
modelMag_u_galaxies-modelMag_g_galaxies,modelMag_u_galaxies-modelMag_i_galaxies]
nproperites = ["H_alpha","extinction_r","velocity Dispersion","redshift","u-g","u-i"]
#print mjd_galaxies
spectra = []
for mjd,plate,fiberID in zip(mjd_galaxies,plate_galaxies,fiberID_galaxies):
spectra.append(fetch_sdss_spectrum(plate,mjd,fiberID))
#print spectra[0].wavelength()
#print spectra[1].wavelength()
#nwav = len(spectra[0].wavelength())
wavs0 = spectra[0].wavelength()
wavmin = wavs0.min()
wavmax = wavs0.max()
nwav = 5000
wavs = np.linspace(wavmin,wavmax,nwav)
data = np.zeros((Ngals,nwav))
def fetch_and_shift_spectra(n_spectra,
outfile,
primtarget=TARGET_GALAXY,
zlim=(0, 0.7),
loglam_start=3.5,
loglam_end=3.9,
Nlam=1000):
"""
This function queries CAS for matching spectra, and then downloads
them and shifts them to a common redshift binning
"""
# First query for the list of spectra to download
plate, mjd, fiber = query_plate_mjd_fiber(n_spectra, primtarget,
zlim[0], zlim[1])
# Set up arrays to hold information gathered from the spectra
spec_cln = np.zeros(n_spectra, dtype=np.int32)
lineindex_cln = np.zeros(n_spectra, dtype=np.int32)
log_NII_Ha = np.zeros(n_spectra, dtype=np.float32)
log_OIII_Hb = np.zeros(n_spectra, dtype=np.float32)
z = np.zeros(n_spectra, dtype=np.float32)
zerr = np.zeros(n_spectra, dtype=np.float32)
spectra = np.zeros((n_spectra, Nlam), dtype=np.float32)
mask = np.zeros((n_spectra, Nlam), dtype=np.bool)
specerr = np.zeros((n_spectra, Nlam), dtype=np.float32)
# also save plate, mjd, fiber to allow reference to SDSS data
plates = np.zeros(n_spectra, dtype=np.int32)
mjds = np.zeros(n_spectra, dtype=np.int32)
fibers = np.zeros(n_spectra, dtype=np.int32)
# Calculate new wavelength coefficients
new_coeff0 = loglam_start
new_coeff1 = (loglam_end - loglam_start) / Nlam
# Now download all the needed spectra, and resample to a common
# wavelength bin.
n_spectra = len(plate)
num_skipped = 0
# changed counter and loop so that skipped spectra do not create gaps in arrays
j = 0
for i in range(n_spectra):
sys.stdout.write(' %i / %i spectra\r' % (i + 1, n_spectra))
sys.stdout.flush()
try:
spec = fetch_sdss_spectrum(plate[i], mjd[i], fiber[i], data_home='/epyc/users/sportill/specAE/cache')
except HTTPError:
num_skipped += 1
print("%i, %i, %i not found" % (plate[i], mjd[i], fiber[i]))
continue
spec_rebin = spec.restframe().rebin(new_coeff0, new_coeff1, Nlam)
if spec.z < zlim[0] or spec.z > zlim[1]:
num_skipped += 1
print("%i, %i, %i outside redshift range" % (plate[i], mjd[i], fiber[i]))
continue
if np.all(spec_rebin.spectrum == 0):
num_skipped += 1
print("%i, %i, %i is all zero" % (plate[i], mjd[i], fiber[i]))
continue
#if spec.spec_cln < 2 or spec.spec_cln > 3:
# num_skipped += 1
# print("%i, %i, %i is not a galaxy spectrum" % (plate[i], mjd[i], fiber[i]))
# continue
spec_cln[j] = spec.spec_cln
lineindex_cln[j], (log_NII_Ha[j], log_OIII_Hb[j])\
= spec.lineratio_index()
z[j] = spec.z
zerr[j] = spec.zerr
spectra[j] = spec_rebin.spectrum
mask[j] = spec_rebin.compute_mask(0.5, 5)
assert((mask[j] == 0).any())
specerr[j] = spec_rebin.error
plates[j] = plate[i]
mjds[j] = mjd[i]
fibers[j] = fiber[i]
j += 1
sys.stdout.write('\n')
N = j
print(" %i spectra skipped" % num_skipped)
print(" %i spectra processed" % N)
print("saving to %s" % outfile)
np.savez(outfile,
spectra=spectra[:N],
mask=mask[:N],
spec_err=specerr[:N],
coeff0=new_coeff0,
coeff1=new_coeff1,
spec_cln=spec_cln[:N],
lineindex_cln=lineindex_cln[:N],
log_NII_Ha=log_NII_Ha[:N],
log_OIII_Hb=log_OIII_Hb[:N],
z=z[:N],
zerr=zerr[:N],
plate=plates[:N],
mjd=mjds[:N],
fiber=fibers[:N])
def fetch_and_shift_spectra(n_spectra,
outfile,
primtarget=TARGET_GALAXY,
zlim=(0, 0.7),
loglam_start=3.5,
loglam_end=3.9,
Nlam=1000):
"""
This function queries CAS for matching spectra, and then downloads
them and shifts them to a common redshift binning
"""
# First query for the list of spectra to download
plate, mjd, fiber = query_plate_mjd_fiber(n_spectra, primtarget,
zlim[0], zlim[1])
print('plate:', plate,'mjd:', mjd, 'fiber:', fiber) # Print identifiers to check against other script
# Set up arrays to hold information gathered from the spectra
spec_cln = np.zeros(n_spectra, dtype=np.int32)
lineindex_cln = np.zeros(n_spectra, dtype=np.int32)
log_NII_Ha = np.zeros(n_spectra, dtype=np.float32)
log_OIII_Hb = np.zeros(n_spectra, dtype=np.float32)
z = np.zeros(n_spectra, dtype=np.float32)
zerr = np.zeros(n_spectra, dtype=np.float32)
spectra = np.zeros((n_spectra, Nlam), dtype=np.float32)
spec_err = np.zeros((n_spectra, Nlam), dtype=np.float32)
mask = np.zeros((n_spectra, Nlam), dtype=np.bool)
# Calculate new wavelength coefficients
new_coeff0 = loglam_start
new_coeff1 = (loglam_end - loglam_start) / Nlam
# Now download all the needed spectra, and resample to a common
# wavelength bin.
n_spectra = len(plate)
num_skipped = 0
i = 0
while i < n_spectra:
sys.stdout.write(' %i / %i spectra\r' % (i + 1, n_spectra))
sys.stdout.flush()
try:
spec = fetch_sdss_spectrum(plate[i], mjd[i], fiber[i])
except HTTPError:
num_skipped += 1
print("%i, %i, %i not found" % (plate[i], mjd[i], fiber[i]))
i += 1
continue
spec_rebin = spec.restframe().rebin(new_coeff0, new_coeff1, Nlam) # Redshift corrected
# spec_rebin = spec.rebin(new_coeff0, new_coeff1, Nlam) # Not redshift corrected
if np.all(spec_rebin.spectrum == 0):
num_skipped += 1
print("%i, %i, %i is all zero" % (plate[i], mjd[i], fiber[i]))
i += 1
continue
spec_cln[i] = spec.spec_cln
lineindex_cln[i], (log_NII_Ha[i], log_OIII_Hb[i])\
= spec.lineratio_index()
z[i] = spec.z
zerr[i] = spec.zerr
spectra[i] = spec_rebin.spectrum
spec_err[i] = spec_rebin.error
mask[i] = spec_rebin.compute_mask(0.5, 5)
i += 1
sys.stdout.write('\n')
N = i
print(" %i spectra skipped" % num_skipped)
print(" %i spectra processed" % N)
print("saving to %s" % outfile)
np.savez(outfile,
spectra=spectra[:N],
spec_err=spec_err[:N],
mask=mask[:N],
coeff0=new_coeff0,
coeff1=new_coeff1,
spec_cln=spec_cln[:N],
lineindex_cln=lineindex_cln[:N],
log_NII_Ha=log_NII_Ha[:N],
log_OIII_Hb=log_OIII_Hb[:N],
z=z[:N],
zerr=zerr[:N])
def plot_image_spec_sdss_galaxy(sdss_obj, save_figure=False):
#plot image and spectrum of a specified SDSS galaxy
# input sdss_obj = individual SDSS main galaxy data base entry
# save_figure specifies whether PDF of the figure will be saved in fig/ subdirectory for the record
#
# define plot with 2 horizonthal panels of appropriate size
fig, (ax0, ax1) = plt.subplots(1, 2, figsize=(6, 2.))
# set an appropriate font size for labels
plt.rc('font', size=8)
#get RA and DEC of the galaxy and form the filename to save SDSS image
RA = sdss_obj['ra']
DEC = sdss_obj['dec']
scale = 0.25
outfile = image_home_dir() + str(sdss_obj['objID']) + '.jpg'
fetch_sdss_image(outfile, RA, DEC, scale)
img = Image.open(outfile)
# do not plot pixel axis labels
ax0.axis('off')
ax0.imshow(img)
# fetch SDSS spectrum using plate number, epoch, and fiber ID
plate = sdss_obj['plate']
mjd = sdss_obj['mjd']
fiber = sdss_obj['fiberID']
spec = fetch_sdss_spectrum(plate, mjd, fiber)
# normalize spectrum for plotting
spectrum = 0.5 * spec.spectrum / spec.spectrum.max()
lam = spec.wavelength()
text_kwargs = dict(ha='center', va='center', alpha=0.5, fontsize=10)
# set axis limits for spectrum plot
ax1.set_xlim(3000, 10000)
ax1.set_ylim(0, 0.6)
color = np.zeros(5)
for i, f, c, loc in zip([0, 1, 2, 3, 4], 'ugriz', 'bgrmk',
[3500, 4600, 6100, 7500, 8800]):
data_home = setup.sdss_filter_dir()
archive_file = os.path.join(data_home, '%s.dat' % f)
if not os.path.exists(archive_file):
raise ValueError(
"Error in plot_img_spec_sdss_galaxy: filter file '%s' does not exist!"
% archive_file)
F = open(archive_file)
filt = np.loadtxt(F, unpack=True)
ax1.fill(filt[0], filt[2], ec=c, fc=c, alpha=0.4)
fsp = interp1d(filt[0], filt[2], bounds_error=False, fill_value=0.0)
ax1.text(loc, 0.03, f, color=c, **text_kwargs)
# compute magnitude in each band using simple Simpson integration of spectrum and filter using eq 1.2 in the notes
fspn = lam * fsp(lam) / simps(fsp(lam) / lam, lam)
lamf = fspn
#lamf = lamf.clip(0.0)
specf = spec.spectrum * lamf
color[i] = -2.5 * np.log10(simps(specf, lam))
# print estimated g-r color for the object
grcatcol = sdss_obj['modelMag_g'] - sdss_obj['modelMag_r']
print("computed g-r = ", color[1] - color[2])
print("catalog g-r=", grcatcol)
ax1.plot(lam, spectrum, '-k', lw=0.5, label=r'$(g-r)=%.2f$' % grcatcol)
ax1.set_xlabel(r'$\lambda {(\rm \AA)}$')
ax1.set_ylabel(r'$\mathrm{norm.\ specific\ flux}\ \ \ f_{\lambda}$')
#ax.set_title('Plate = %(plate)i, MJD = %(mjd)i, Fiber = %(fiber)i' % locals())
#ax.set_title('%s' % objects[obj_]['name'])
ax1.legend(frameon=False)
if save_figure:
plt.savefig('fig/gal_img_spec' + '_' + str(sdss_obj['objID']) + '.pdf',
bbox_inches='tight')
plt.subplots_adjust(wspace=0.2)
plt.show()
return
def fitted_parameters(self, burnin=None, thin=1):
if burnin is None:
burnin = int(self.sampler.chain.shape[1] * 0.5)
warn("no burnin given, using burnin of 50%")
fitted = np.percentile(self.sampler.chain[:, burnin::thin].reshape(
-1, self.sampler.chain.shape[-1]), (16, 50, 84),
axis=0)
return pd.DataFrame(fitted.T,
columns=['lower', 'median', 'upper'],
index=self.parameter_names)
if __name__ == '__main__':
from astroML.datasets import fetch_sdss_spectrum
spec = fetch_sdss_spectrum(1975, 53734, 1).restframe()
x, y, e = spec.wavelength(), spec.spectrum, spec.error
norm_const = np.median(spec.spectrum[(x >= 5500) & (x <= 5600)])
y /= norm_const
e /= norm_const
model = SpectrumModel()
ha_group = model.add_region(6450, 6775, 'halpha')
hb_group = model.add_region(4800, 5050, 'hbeta')
NIIa = ha_group.add_emission_line('NIIa', 6549.86)
NIIb = ha_group.add_emission_line('NIIb',
6585.27,
amplitude=NIIa.amplitude / 0.34,
width=NIIa.width)
