def make_set(ntrain,
slines,
outroot=None,
tol=1 * u.arcsec,
igmsp_survey='SDSS_DR7',
frac_without=0.,
seed=1234,
zmin=None,
zmax=4.5,
high=False,
slls=False,
mix=False,
low_s2n=False):
""" Generate a training set
Parameters
----------
ntrain : int
Number of training sightlines to generate
slines : Table
Table of sightlines without DLAs (usually from SDSS or BOSS)
igmsp_survey : str, optional
Dataset name for spectra
frac_without : float, optional
Fraction of sightlines (on average) without a DLA
seed : int, optional
outroot : str, optional
Root for output filenames
root+'.fits' for spectra
root+'.json' for DLA info
zmin : float, optional
Minimum redshift for training; defaults to min(slines['ZEM'])
zmax : float, optional
Maximum redshift to train on
mix : bool, optional
Mix of SLLS and DLAs
low_s2n : bool, optional
Reduce the S/N artificially, i.e. add noise
Returns
-------
"""
from linetools.spectra.utils import collate
# Init and checks
igmsp = IgmSpec()
assert igmsp_survey in igmsp.groups
rstate = np.random.RandomState(seed)
rfrac = rstate.random_sample(ntrain)
if zmin is None:
zmin = np.min(slines['ZEM'])
rzem = zmin + rstate.random_sample(ntrain) * (zmax - zmin)
fNHI = init_fNHI(slls=slls, mix=mix, high=high)
all_spec = []
full_dict = {}
# Begin looping
for qq in range(ntrain):
print("qq = {:d}".format(qq))
full_dict[qq] = {}
# Grab sightline
isl = np.argmin(np.abs(slines['ZEM'] - rzem[qq]))
full_dict[qq]['sl'] = isl # sightline
specl, meta = igmsp.spectra_from_coord(
(slines['RA'][isl], slines['DEC'][isl]),
groups=['SDSS_DR7'],
tol=tol,
verbose=False)
assert len(meta) == 1
spec = specl
# Meta data for header
mdict = {}
for key in meta.keys():
mdict[key] = meta[key][0]
mhead = Header(mdict)
# Clear?
if rfrac[qq] < frac_without:
spec.meta['headers'][0] = mdict.copy() #mhead
all_spec.append(spec)
full_dict[qq]['nDLA'] = 0
continue
# Insert at least one DLA
spec, dlas = insert_dlas(spec,
mhead['zem_GROUP'],
rstate=rstate,
fNHI=fNHI,
slls=slls,
mix=mix,
high=high,
low_s2n=low_s2n)
spec.meta['headers'][0] = mdict.copy() #mhead
all_spec.append(spec)
full_dict[qq]['nDLA'] = len(dlas)
for kk, dla in enumerate(dlas):
full_dict[qq][kk] = {}
full_dict[qq][kk]['NHI'] = dla.NHI
full_dict[qq][kk]['zabs'] = dla.zabs
# Generate one object
final_spec = collate(all_spec)
# Write?
if outroot is not None:
# Spectra
final_spec.write_to_hdf5(outroot + '.hdf5')
# Dict -> JSON
gdict = ltu.jsonify(full_dict)
ltu.savejson(outroot + '.json', gdict,
overwrite=True) #, easy_to_read=True)
# Return
return final_spec, full_dict
def fig_lowz_hiz(outfil='fig_loz_hiz.png'):
""" Show varying IGM transmission
"""
#hdlls_path = '/u/xavier/paper/LLS/Optical/Data/DR1/Spectra/'
esi_path = '/u/xavier/Keck/ESI/RedData/'
hst_path = '/u/xavier/HST/Cycle23/z1IGM/Archive/PG1206+459/'
#
igmsp = IgmSpec()
idicts = [
dict(filename='Data/3C273_STIS_E140M_F.fits'),
dict(filename=hst_path + 'PG1206+459_E230M_f.fits'),
dict(coord='J212329.50-005052.9', group=['HD-LLS_DR1']),
dict(coord='J113621.00+005021.0', group=['HD-LLS_DR1']),
dict(coord='J113246.5+120901.6', group=['ESI_DLA']),
dict(filename=esi_path +
'J1148+5251/SDSSJ1148+5251_stack.fits'), # z=6
]
lbls = [
'HST/STIS: 3C273',
'HST/STIS: PG1206+459',
'Keck/HIRES: J2123-0050', # 2.26
'Magellan/MIKE: J1136+0050', # 3.43
'Keck/ESI: J1132+1209', # 5.17
'Keck/ESI: J1148+5251', # 6.4
]
zems = [0.17, 1.16, 2.26, 3.43, 5.17, 6.4]
xrest = np.array([1080, 1200.])
ymnx = [-0.1, 1.1]
# Cut down
idicts = [idicts[0], idicts[3]]
lbls = [lbls[0], lbls[3]]
zems = [zems[0], zems[3]]
lw = 1.
csz = 19.
# Start the plot
#fig = plt.figure(figsize=(5.0, 8.0))
fig = plt.figure(figsize=(8.0, 5.0))
plt.clf()
gs = gridspec.GridSpec(len(lbls), 1)
# Loop
for qq, lbl in enumerate(lbls):
# Grab data
idict = idicts[qq]
if 'coord' in idict.keys():
qdict = {}
for key in idict.keys():
if key not in ['coord', 'group']:
qdict[key] = idict[key]
spec, meta = igmsp.spectra_from_coord(
idict['coord'], tol=5. * u.arcsec,
groups=idict['group']) #, query_dict=qdict)
if meta is None:
print("Bad coord?")
pdb.set_trace()
elif len(meta) > 1:
pdb.set_trace()
else:
spec = lsio.readspec(idict['filename'])
if lbl == 'HST/STIS: 3C273':
hdu = fits.open('Data/3C273_STIS_E140M_c.fits')
conti_3c273 = hdu[0].data
spec.co = conti_3c273
spec.normed = True
# Spectrum
ax = plt.subplot(gs[qq])
ax.set_xlim(xrest * (1 + zems[qq]) / 1215.67 - 1)
ax.set_ylim(ymnx)
if qq == 3:
ax.set_ylabel('Normalized Flux')
if qq == len(lbls) - 1:
ax.set_xlabel(r'Redshift of Ly$\alpha$')
ax.plot(spec.wavelength.value / 1215.6701 - 1,
spec.flux,
'k',
linewidth=lw)
# Label
#ax.text(0.05, 0.95, lbl+' zem={:0.1f}'.format(zems[qq]), color='blue',
# transform=ax.transAxes, size=csz, ha='left', bbox={'facecolor':'white'})
#
set_fontsize(ax, 12.)
# Layout and save
#plt.subplots_adjust(hspace=0)
plt.tight_layout(pad=0.2, h_pad=0.0, w_pad=0.4)
plt.savefig(outfil, dpi=600)
plt.close()
# Finish
print('Writing {:s}'.format(outfil))