def __init__(self,
f_mod_interp,
bandCoefAmplitudes, bandCoefPositions, bandCoefWidths,
lines_pos, lines_width,
var_C, var_L, alpha_C, alpha_L,
redshiftGridGP,
use_interpolators=True,
lambdaRef=4.5e3,
g_AB=1.0):
DL = approx_DL()
self.bands = np.arange(bandCoefAmplitudes.shape[0])
if isinstance(f_mod_interp, int):
self.mean_fct = None
self.nt = f_mod_interp
else:
self.mean_fct = Photoz_linear_sed_basis(f_mod_interp)
self.nt = f_mod_interp.shape[0]
# self.mean_fct = Photoz_mean_function(
# alpha, bandCoefAmplitudes, bandCoefPositions, bandCoefWidths,
# g_AB=g_AB, lambdaRef=lambdaRef, DL_z=DL)
self.kernel = Photoz_kernel(
bandCoefAmplitudes, bandCoefPositions, bandCoefWidths,
lines_pos, lines_width, var_C, var_L, alpha_C, alpha_L,
g_AB=g_AB, DL_z=DL, redshiftGrid=redshiftGridGP,
use_interpolators=use_interpolators)
self.redshiftGridGP = redshiftGridGP
def __init__(self,
bandCoefAmplitudes,
bandCoefPositions,
bandCoefWidths,
lines_pos,
lines_width,
var_C,
var_L,
alpha_T,
alpha_C,
alpha_L,
redshiftGridGP,
use_interpolators=True,
lambdaRef=4.5e3,
g_AB=1.0):
DL = approx_DL()
self.bands = np.arange(bandCoefAmplitudes.shape[0])
self.kernel = Photoz_SN_kernel(bandCoefAmplitudes,
bandCoefPositions,
bandCoefWidths,
lines_pos,
lines_width,
var_C,
var_L,
alpha_T,
alpha_C,
alpha_L,
g_AB=g_AB,
DL_z=DL,
redshiftGrid=redshiftGridGP,
use_interpolators=use_interpolators)
self.redshiftGridGP = redshiftGridGP
def __init__(self,
f_mod_interp,
bandCoefAmplitudes, bandCoefPositions, bandCoefWidths,
lines_pos, lines_width,
var_C, var_L, alpha_C, alpha_L,
redshiftGridGP,
use_interpolators=True,
lambdaRef=4.5e3,
g_AB=1.0):
DL = approx_DL()
self.bands = np.arange(bandCoefAmplitudes.shape[0])
if isinstance(f_mod_interp, int):
self.mean_fct = None
self.nt = f_mod_interp
else:
self.mean_fct = Photoz_linear_sed_basis(f_mod_interp)
self.nt = f_mod_interp.shape[0]
# self.mean_fct = Photoz_mean_function(
# alpha, bandCoefAmplitudes, bandCoefPositions, bandCoefWidths,
# g_AB=g_AB, lambdaRef=lambdaRef, DL_z=DL)
self.kernel = Photoz_kernel(
bandCoefAmplitudes, bandCoefPositions, bandCoefWidths,
lines_pos, lines_width, var_C, var_L, alpha_C, alpha_L,
g_AB=g_AB, DL_z=DL, redshiftGrid=redshiftGridGP,
use_interpolators=use_interpolators)
self.redshiftGridGP = redshiftGridGP
def __init__(self,
alpha,
fcoefs_amp,
fcoefs_mu,
fcoefs_sig,
g_AB=1.0,
lambdaRef=4.5e3,
DL_z=None,
name='photoz_mf'):
""" Constructor."""
# If luminosity_distance function not provided, use approximation
if DL_z is None:
self.DL_z = approx_DL()
else:
self.DL_z = DL_z
self.g_AB = g_AB
assert lambdaRef > 1e2 and lambdaRef < 1e5
self.lambdaRef = lambdaRef
self.fourpi = 4 * np.pi
self.sqrthalfpi = np.sqrt(np.pi / 2)
self.alpha = alpha
assert fcoefs_amp.shape[0] == fcoefs_mu.shape[0] and\
fcoefs_amp.shape[0] == fcoefs_sig.shape[0]
self.fcoefs_amp = np.array(fcoefs_amp)
self.fcoefs_mu = np.array(fcoefs_mu)
self.fcoefs_sig = np.array(fcoefs_sig)
self.numCoefs = fcoefs_amp.shape[1]
self.norms = np.sqrt(2*np.pi)\
* np.sum(self.fcoefs_amp * self.fcoefs_sig / self.fcoefs_mu, axis=1)
self.fcoefs_amp *= self.fcoefs_mu
def __init__(self, numTypes, maglim=None):
self.numTypes = numTypes
self.params = OrderedDict({})
self.paramranges = OrderedDict({})
for i in range(numTypes - 1):
self.params['pt' + str(i + 1)] = 0.5
self.paramranges['pt' + str(i + 1)] = [0.2, 0.9]
if maglim is not None:
self.maglim = maglim
self.DL = approx_DL()
else:
self.maglim = None
self.lumFct = powerLawLuminosityFct() # p(lum | z)
# p(z, t)
self.nofz = ComovingVolumeEfect()
self.children = [self.lumFct] + [self.nofz]
def __init__(self,
bandCoefAmplitudes, bandCoefPositions, bandCoefWidths,
lines_pos, lines_width,
var_C, var_L, alpha_T, alpha_C, alpha_L,
redshiftGridGP,
use_interpolators=True,
lambdaRef=4.5e3,
g_AB=1.0):
DL = approx_DL()
self.bands = np.arange(bandCoefAmplitudes.shape[0])
self.kernel = Photoz_SN_kernel(
bandCoefAmplitudes, bandCoefPositions, bandCoefWidths,
lines_pos, lines_width, var_C, var_L, alpha_T, alpha_C, alpha_L,
g_AB=g_AB, DL_z=DL, redshiftGrid=redshiftGridGP,
use_interpolators=use_interpolators)
self.redshiftGridGP = redshiftGridGP
def __init__(self,
tabulatedWavelength,
tabulatedSpectrum,
photometricBands,
redshiftGrid=None,
order=15):
self.DL = approx_DL()
self.photometricBands = photometricBands
self.numBands = len(photometricBands)
self.sed_interp = interp1d(tabulatedWavelength,
tabulatedSpectrum,
bounds_error=False,
fill_value="extrapolate")
if redshiftGrid is None:
self.redshiftGrid = np.logspace(np.log10(1e-2), np.log10(2.0), 350)
else:
self.redshiftGrid = redshiftGrid
self.fbcoefs = {}
self.fbinterps = {}
self.logfbinterps = {}
self.order = order
self.fmodgrid = np.zeros(
(self.redshiftGrid.size, len(photometricBands)))
self.bandNames = []
for ib, filt in enumerate(photometricBands):
self.bandNames.append(filt.bandName)
for iz in range(self.redshiftGrid.size):
opz = (self.redshiftGrid[iz] + 1)
xf_z = filt.wavelengthGrid / opz
yf_z = filt.tabulatedResponse
ysed = self.sed_interp(xf_z)
facz = opz**2. / (4 * np.pi *
self.DL(self.redshiftGrid[iz])**2.)
ysedext = facz * ysed
self.fmodgrid[iz, ib] =\
np.trapz(ysedext * yf_z, x=xf_z) / filt.norm
self.fbinterps[filt.bandName] = UnivariateSpline(self.redshiftGrid,
self.fmodgrid[:,
ib],
s=0)
self.fbcoefs[filt.bandName] = np.polyfit(
self.redshiftGrid, np.log(self.fmodgrid[:, ib]),
self.order - 1)
self.logfbinterps[filt.bandName] =\
np.poly1d(self.fbcoefs[filt.bandName])
def __init__(self,
fcoefs_amp,
fcoefs_mu,
fcoefs_sig,
lines_mu,
lines_sig,
var_C,
var_L,
alpha_C,
alpha_L,
g_AB=1.0,
DL_z=None,
redshiftGrid=None,
use_interpolators=True):
""" Constructor."""
self.use_interpolators = use_interpolators
if DL_z is None:
self.DL_z = approx_DL()
else:
self.DL_z = DL_z
self.g_AB = g_AB
self.fourpi = 4 * np.pi
self.lines_mu = np.array(lines_mu)
self.lines_sig = np.array(lines_sig)
self.numLines = self.lines_mu.size
assert fcoefs_amp.shape[0] == fcoefs_mu.shape[0] and\
fcoefs_amp.shape[0] == fcoefs_sig.shape[0]
self.fcoefs_amp = fcoefs_amp
self.fcoefs_mu = fcoefs_mu
self.fcoefs_sig = fcoefs_sig
self.numCoefs = fcoefs_amp.shape[1]
self.numBands = fcoefs_amp.shape[0]
self.norms = np.sqrt(2*np.pi)\
* np.sum(self.fcoefs_amp * self.fcoefs_sig, axis=1)
# Initialize parameters and link them.
self.var_C = var_C
self.var_L = var_L
self.alpha_C = alpha_C
self.alpha_L = alpha_L
if redshiftGrid is None:
self.redshiftGrid = np.linspace(0, 4, num=160)
else:
self.redshiftGrid = copy(redshiftGrid)
self.nz = self.redshiftGrid.size
self.construct_interpolators()
def __init__(self,
tabulatedWavelength,
tabulatedSpectrum,
photometricBands,
redshiftGrid=None,
order=15):
self.DL = approx_DL()
self.photometricBands = photometricBands
self.numBands = len(photometricBands)
self.sed_interp = interp1d(tabulatedWavelength, tabulatedSpectrum)
if redshiftGrid is None:
self.redshiftGrid = np.logspace(np.log10(1e-2), np.log10(2.0), 350)
else:
self.redshiftGrid = redshiftGrid
self.fbcoefs = {}
self.fbinterps = {}
self.order = order
for filt in photometricBands:
fmodgrid = np.zeros((self.redshiftGrid.size, ))
for iz in range(self.redshiftGrid.size):
opz = (self.redshiftGrid[iz] + 1)
xf_z = filt.wavelengthGrid / opz
yf_z = filt.tabulatedResponse
ysed = self.sed_interp(xf_z)
facz = opz**2. / (4 * np.pi *
self.DL(self.redshiftGrid[iz])**2.)
ysedext = facz * ysed
fmodgrid[iz] =\
np.trapz(ysedext * yf_z, x=xf_z) / filt.norm
# self.fbinterps[filt.bandName] = UnivariateSpline(
# self.redshiftGrid, fmodgrid, s=0)
self.fbcoefs[filt.bandName] = np.polyfit(self.redshiftGrid,
np.log(fmodgrid),
self.order - 1)
self.fbinterps[filt.bandName] =\
np.poly1d(self.fbcoefs[filt.bandName])
def getDataFromFile(params,
firstLine,
lastLine,
prefix="",
ftype="catalog",
getXY=True,
CV=False):
"""
Returns an iterator to parse an input catalog file.
Returns the fluxes, redshifts, etc, and also GP inputs if getXY=True.
"""
if ftype == "gpparams":
with open(params[prefix + 'paramFile']) as f:
for line in itertools.islice(f, firstLine, lastLine):
data = np.fromstring(line, dtype=float, sep=' ')
B = int(data[0])
z = data[1]
ell = data[2]
bands = data[3:3 + B]
flatarray = data[3 + B:]
X = np.zeros((B, 3))
for off, iband in enumerate(bands):
X[off, 0] = iband
X[off, 1] = z
X[off, 2] = ell
yield z, ell, bands, X, B, flatarray
if ftype == "catalog":
DL = approx_DL()
bandIndices, bandNames, bandColumns, bandVarColumns, redshiftColumn,\
refBandColumn = readColumnPositions(params, prefix=prefix)
bandCoefAmplitudes, bandCoefPositions, bandCoefWidths, norms\
= readBandCoefficients(params)
refBandNorm = norms[params['bandNames'].index(params[prefix +
'referenceBand'])]
if CV:
bandIndicesCV, bandNamesCV, bandColumnsCV,\
bandVarColumnsCV, redshiftColumnCV =\
readColumnPositions(params, prefix=prefix+'CV_', refFlux=False)
with open(params[prefix + 'catFile']) as f:
for line in itertools.islice(f, firstLine, lastLine):
data = np.array(line.split(' '), dtype=float)
refFlux = data[refBandColumn]
normedRefFlux = refFlux * refBandNorm
if redshiftColumn >= 0:
z = data[redshiftColumn]
else:
z = -1
# drop bad values and find how many bands are valid
mask = np.isfinite(data[bandColumns])
mask &= np.isfinite(data[bandVarColumns])
mask &= data[bandColumns] > 0.0
mask &= data[bandVarColumns] > 0.0
bandsUsed = np.where(mask)[0]
numBandsUsed = mask.sum()
if z > -1:
ell = normedRefFlux * 4 * np.pi \
* params['fluxLuminosityNorm'] * DL(z)**2 * (1+z)
if (refFlux <= 0) or (not np.isfinite(refFlux))\
or (z < 0) or (numBandsUsed <= 1):
print("Skipping galaxy: refflux=", refFlux, "z=", z,
"numBandsUsed=", numBandsUsed)
continue # not valid data - skip to next valid object
fluxes = data[bandColumns[mask]]
fluxesVar = data[bandVarColumns[mask]] +\
(params['training_extraFracFluxError'] * fluxes)**2
if CV:
maskCV = np.isfinite(data[bandColumnsCV])
maskCV &= np.isfinite(data[bandVarColumnsCV])
maskCV &= data[bandColumnsCV] > 0.0
maskCV &= data[bandVarColumnsCV] > 0.0
bandsUsedCV = np.where(maskCV)[0]
numBandsUsedCV = maskCV.sum()
fluxesCV = data[bandColumnsCV[maskCV]]
fluxesCVVar = data[bandVarColumnsCV[maskCV]] +\
(params['training_extraFracFluxError'] * fluxesCV)**2
if not getXY:
if CV:
yield z, normedRefFlux,\
bandIndices[mask], fluxes, fluxesVar,\
bandIndicesCV[maskCV], fluxesCV, fluxesCVVar
else:
yield z, normedRefFlux,\
bandIndices[mask], fluxes, fluxesVar,\
None, None, None
if getXY:
Y = np.zeros((numBandsUsed, 1))
Yvar = np.zeros((numBandsUsed, 1))
X = np.ones((numBandsUsed, 3))
for off, iband in enumerate(bandIndices[mask]):
X[off, 0] = iband
X[off, 1] = z
X[off, 2] = ell
Y[off, 0] = fluxes[off]
Yvar[off, 0] = fluxesVar[off]
if CV:
yield z, normedRefFlux,\
bandIndices[mask], fluxes, fluxesVar,\
bandIndicesCV[maskCV], fluxesCV, fluxesCVVar,\
X, Y, Yvar
else:
yield z, normedRefFlux,\
bandIndices[mask], fluxes, fluxesVar,\
None, None, None,\
X, Y, Yvar
def getDataFromFile(params, firstLine, lastLine,
prefix="", ftype="catalog", getXY=True, CV=False):
"""
Returns an iterator to parse an input catalog file.
Returns the fluxes, redshifts, etc, and also GP inputs if getXY=True.
"""
if ftype == "gpparams":
with open(params[prefix+'paramFile']) as f:
for line in itertools.islice(f, firstLine, lastLine):
data = np.fromstring(line, dtype=float, sep=' ')
B = int(data[0])
z = data[1]
ell = data[2]
bands = data[3:3+B]
flatarray = data[3+B:]
X = np.zeros((B, 3))
for off, iband in enumerate(bands):
X[off, 0] = iband
X[off, 1] = z
X[off, 2] = ell
yield z, ell, bands, X, B, flatarray
if ftype == "catalog":
DL = approx_DL()
bandIndices, bandNames, bandColumns, bandVarColumns, redshiftColumn,\
refBandColumn = readColumnPositions(params, prefix=prefix)
bandCoefAmplitudes, bandCoefPositions, bandCoefWidths, norms\
= readBandCoefficients(params)
refBandNorm = norms[params['bandNames']
.index(params[prefix+'referenceBand'])]
if CV:
bandIndicesCV, bandNamesCV, bandColumnsCV,\
bandVarColumnsCV, redshiftColumnCV =\
readColumnPositions(params, prefix=prefix+'CV_', refFlux=False)
with open(params[prefix+'catFile']) as f:
for line in itertools.islice(f, firstLine, lastLine):
data = np.array(line.split(' '), dtype=float)
refFlux = data[refBandColumn]
normedRefFlux = refFlux * refBandNorm
if redshiftColumn >= 0:
z = data[redshiftColumn]
else:
z = -1
# drop bad values and find how many bands are valid
mask = np.isfinite(data[bandColumns])
mask &= np.isfinite(data[bandVarColumns])
mask &= data[bandColumns] > 0.0
mask &= data[bandVarColumns] > 0.0
bandsUsed = np.where(mask)[0]
numBandsUsed = mask.sum()
if z > -1:
ell = normedRefFlux * 4 * np.pi \
* params['fluxLuminosityNorm'] * DL(z)**2 * (1+z)
if (refFlux <= 0) or (not np.isfinite(refFlux))\
or (z < 0) or (numBandsUsed <= 1):
print("Skipping galaxy: refflux=", refFlux,
"z=", z, "numBandsUsed=", numBandsUsed)
continue # not valid data - skip to next valid object
fluxes = data[bandColumns[mask]]
fluxesVar = data[bandVarColumns[mask]] +\
(params['training_extraFracFluxError'] * fluxes)**2
if CV:
maskCV = np.isfinite(data[bandColumnsCV])
maskCV &= np.isfinite(data[bandVarColumnsCV])
maskCV &= data[bandColumnsCV] > 0.0
maskCV &= data[bandVarColumnsCV] > 0.0
bandsUsedCV = np.where(maskCV)[0]
numBandsUsedCV = maskCV.sum()
fluxesCV = data[bandColumnsCV[maskCV]]
fluxesCVVar = data[bandVarColumnsCV[maskCV]] +\
(params['training_extraFracFluxError'] * fluxesCV)**2
if not getXY:
if CV:
yield z, normedRefFlux,\
bandIndices[mask], fluxes, fluxesVar,\
bandIndicesCV[maskCV], fluxesCV, fluxesCVVar
else:
yield z, normedRefFlux,\
bandIndices[mask], fluxes, fluxesVar,\
None, None, None
if getXY:
Y = np.zeros((numBandsUsed, 1))
Yvar = np.zeros((numBandsUsed, 1))
X = np.ones((numBandsUsed, 3))
for off, iband in enumerate(bandIndices[mask]):
X[off, 0] = iband
X[off, 1] = z
X[off, 2] = ell
Y[off, 0] = fluxes[off]
Yvar[off, 0] = fluxesVar[off]
if CV:
yield z, normedRefFlux,\
bandIndices[mask], fluxes, fluxesVar,\
bandIndicesCV[maskCV], fluxesCV, fluxesCVVar,\
X, Y, Yvar
else:
yield z, normedRefFlux,\
bandIndices[mask], fluxes, fluxesVar,\
None, None, None,\
X, Y, Yvar