def plec():
index, a, b = ce.PLEC_Index, ce.PLEC_Expfactor, ce.PLEC_Exp_Index
cutoff = (1 / a)**(1 / b)
prefactor = flux * np.exp(a * pivot**b)
return Models.PLSuperExpCutoff(p=[prefactor, index, cutoff, b],
e0=pivot,
free=free)
def lookup(self, name):
""" return an ExtendedSource object by name, None if not found """
aname = self.alias.get(name,name) #alias will be the new name
try:
i = list(self.names).index(name)
except ValueError:
return None
source = self.sources[i]
# make a new object copied from original
if source.model.name=='BrokenPowerLaw': #convert this
model = Models.LogParabola()
else: model = source.model
if model.name=='LogParabola':
model.free[-1]=False # E_break ne free
### seems to be necessary for some models created from
if model.mappers[0].__class__.__name__== 'LimitMapper':
print 'wrong mappers: converting model for source %s, model %s' % (name, model.name)
model = eval('Models.%s(p=%s)' % (model.name, list(model.get_all_parameters())))
extsource = sources.ExtendedSource(name=self.realname(aname),
skydir=source.skydir,
model = model,
dmodel= source.spatial_model
)
if extsource.model.name=='LogParabola':
assert sum( extsource.model.free)<4, 'E_break is free?'
return extsource
def fit_isotropic(roi, nbands=8, folder='isotropic_fit'):
""" fit only the front and back"""
from uw.like import Models
iso_source = roi.get_source('isotrop')
old_model=iso_source.model.copy()
roi.sources.set_model(Models.FrontBackConstant(), 'isotrop')
iso_model=iso_source.model
roi.reinitialize()
cx = []
for eband in range(nbands):
print '*** Energy Band {}: iso counts {}'.format( eband,
[t[1].counts.round() for t in roi[2*eband:2*eband+2]])
roi.select(eband)
iso_model[0]=1; iso_model[1]=1
roi.fit([0,1]);
u = iso_model.get_all_parameters();
cx.append(u)
roi.select()
roi.sources.set_model(old_model)
if folder is not None:
if not os.path.exists(folder): os.mkdir(folder)
filename= '{}/{}.pickle'.format(folder, roi.name)
pickle.dump(np.array(cx), open(filename, 'w'))
print 'wrote file {}'.format(filename)
return np.array(cx)
def __init__(self, cat_entry):
"""return a Pandas.DataFrame for the band flux information
cat_entry: pandas.Series object
Adds columns with conversion to differential energy flux, in eV units
"""
cols = 'Flux Unc_Flux Index Npred Sqrt_TS'.split()
t = [cat_entry[cname + '_Band'] for cname in cols]
bf = pd.DataFrame(t, index=cols).T
# add columns with energy flux
# evaluate energy flux = e**2 * df/de at geometric mean, assuming powerlaw
# convert to units eV/(cm**2 * s)
eb = energy_bounds
r = BandFlux.conversion_factors(eb.LowerEnergy, eb.UpperEnergy,
bf.Index)
bf['eflux'] = bf.Flux * r
bf['eflux_unc'] = (bf.Unc_Flux) * r
self.bf = bf
# construct the spectral model object if info is there
# TODO: deal with exponential
s = cat_entry
#self.name = cat_entry['Source_Name']
self.model = None
self.nickname = cat_entry['NickName']
if 'Flux_Density' in cat_entry:
if s['SpectrumType'] == 'LogParabola':
self.model = Models.LogParabola(
[
s['Flux_Density'], s['Spectral_Index'], s['beta'],
s['Pivot_Energy']
],
free=[True, True, False, False])
def __init__(self, roi, nbands, folder):
""" fit the front and back"""
from uw.like import Models
iso_source = roi.get_source('isotrop')
old_model = iso_source.model.copy()
roi.sources.set_model(Models.FrontBackConstant(), 'isotrop')
iso_model = iso_source.model
roi.reinitialize()
cx = []
dx = []
for eband in range(nbands):
roi.select(eband)
print '*** Energy Band {}: iso counts {}'.format(
eband, [t[1].counts.round() for t in roi.selected])
iso_model[0] = 1
iso_model[1] = 1
n = len(roi.selected)
roi.fit(range(n))
u = iso_model.get_all_parameters()
du = np.array([iso_model.error(i) for i in range(2)])
cx.append(u)
dx.append(du)
roi.select()
roi.sources.set_model(old_model)
if folder is not None:
if not os.path.exists(folder): os.mkdir(folder)
filename = '{}/{}.pickle'.format(folder, roi.name)
pickle.dump(dict(val=cx, err=dx), open(filename, 'w'))
print 'wrote file {}'.format(filename)
self.val = cx
self.err = dx
def __call__(self, s):
sdir = s.skydir
if hasattr(s, 'spatial_model') and s.spatial_model is not None:
return True
for i, t in enumerate(self.sdir):
dist = np.degrees(t.difference(sdir))
if dist < self.tol and s.ts > self.ts_min:
self.tags[i] = True
self.assoc[i] = (s.name, dist, s.ts)
if self.rename and s.name != self.psr_names[i]:
print 'Skymodel: renaming %s(%d) to %s' % (
s.name, s.index, self.psr_names[i])
s.name = self.psr_names[i]
if s.model.name == 'ExpCutoff': return True
flux = s.model[0]
if flux > 1e-18:
print 'Skymodel: replacing model for: %s(%d): pulsar name: %s' % (
s.name, s.index, self.psr_names[i])
s.model = Models.ExpCutoff()
s.free = s.model.free.copy()
else:
print 'Apparent pulsar %s(%d), %s, is very weak, flux=%.2e <1e-13: leave as powerlaw' % (
s.name, s.index, self.psr_names[i], flux)
return True
if s.model.name == 'ExpCutoff':
print 'Skymodel setup warning: %s (%d) not in LAT pulsar list, should not be expcutoff' % (
s.name, s.index)
return True
def model(st, flux, index, beta, pivot, cutoff, b):
if st == 'PowerLaw':
return Models.PowerLaw(p=[flux, index], e0=pivot)
elif st == 'PLSuperExpCutoff':
prefactor = flux * np.exp((pivot / cutoff)**b)
return Models.PLSuperExpCutoff(p=[prefactor, index, cutoff, b],
e0=pivot)
elif st == 'PLExpCutoff':
prefactor = flux * np.exp((pivot / cutoff))
return Models.PLSuperExpCutoff(
p=[prefactor, index, cutoff, 1.0], e0=pivot)
elif st == 'LogParabola':
return Models.LogParabola(p=[flux, index, beta, pivot])
elif st == 'PowerLaw2': ### same as PowerLaw in table
return Models.PowerLaw(p=[flux, index], e0=pivot)
else:
raise Exception(
'unexpected spectrum type {}, source name {}'.format(
st, name))
def pmodel(source):
""" create a pointlike model from a Series object from DataFrame row
"""
expandbits = lambda x: [(x & 2**i) != 0 for i in range(4)]
modelname, freebits, e0, norm, norm_unc, pindex, pindex_unc, index2,index2_unc, cutoff, cutoff_unc=\
[source[x] for x in 'modelname freebits e0 flux flux_unc pindex pindex_unc index2 index2_unc cutoff cutoff_unc'.split()]
free = expandbits(freebits)
errors = [norm_unc, pindex_unc]
if modelname == 'LogParabola':
if np.abs(index2) < 2e-3:
modelname = 'PowerLaw'
model = Models.PowerLaw(p=[norm, pindex], e0=e0)
else:
# if index2<0:
# print 'Source %s has beta (%.2f) <0: setting to 0.' % ( source.name, index2, )
# index2=0
model = Models.LogParabola(p=[norm, pindex, index2, e0])
model.free[-1] = False
errors.append(index2_unc)
elif modelname == 'PLSuperExpCutoff':
prefactor = np.exp(-(e0 / cutoff)**index2)
model = Models.PLSuperExpCutoff(
p=[norm / prefactor, pindex, cutoff, index2], e0=e0)
errors[0] /= prefactor
errors += [cutoff_unc, index2_unc]
elif modelname == 'ExpCutoff':
prefactor = np.exp(-(e0 / cutoff))
model = Models.PLSuperExpCutoff(
p=[norm / prefactor, pindex, cutoff, 0.], e0=e0)
model.free[3] == False
errors[0] /= prefactor
errors += [cutoff_unc]
else:
raise Exception('model "%s" not recognized' % modelname)
map(model.set_error, range(len(errors)), errors)
model.free[:] = free[:model.len()]
return model
def fit_diffuse(roi, nbands=8, select=[0,1,2], restore=False, folder='diffuse_fit'):
"""
Perform indpendent fits to the gal, iso_front, and iso_back for each of the first nbands bands
select: None or list of variables
"""
from uw.like import Models
# freeze all free sources, thaw gal and iso
roi.thaw('Norm', 'ring')
iso_model =roi.sources.find_source('isotrop').model.copy()
roi.sources.set_model(Models.FrontBackConstant(), 'isotrop')
roi.reinitialize()
# do the fitting
dpars=[]
energies = []
for ie in range(nbands):
roi.select(ie);
roi.fit(select, ignore_exception=True)
energies.append(int(roi.energies[0]))
dpars.append( roi.sources.parameters.get_parameters()[:3])
t =np.power(10, dpars)
df = pd.DataFrame(t, columns=['gal iso_front iso_back'.split()])
df.index=energies
if restore:
# does not seem to work, comment this out for now
# restore sources
roi.sources.diffuse_normalization *= df
for s,f in zip(free_sources, saved_free):
s.model.free=f
roi.sources.find_source('ring').model.free[0]=False
roi.sources.set_model(iso_model, 'isotrop')
roi.reinitialize()
roi.select()
roi.fit() # needed to restore gradient, at least.
# update the pickle file
write_pickle(roi)
elif folder is not None:
# simply save results
if not os.path.exists(folder):
os.mkdir(folder)
filename= '{}/{}.pickle'.format(folder, roi.name)
pickle.dump(df, open(filename, 'w'))
print 'wrote file {}'.format(filename)
return df
def lp():
index, beta = ce.LP_Index, ce.LP_beta
return Models.LogParabola(p=[flux, index, beta, pivot],
free=free)
def ExpCutoff(*pars):
model = Models.ExpCutoff(p=pars, free=[True, True, False])
return model
def PLSuperExpCutoff(*pars):
model = Models.PLSuperExpCutoff(p=pars, free=[True, True, False, False])
return model
def ExpCutoff(*pars):
model = Models.ExpCutoff(p=pars, free=[True, True, False])
sources.set_default_bounds(model)
return model
def PLSuperExpCutoff(*pars):
model = Models.PLSuperExpCutoff(p=pars, free={True, True, False, False})
sources.set_default_bounds(model)
return model
def LogParabola(*pars):
model = Models.LogParabola(p=pars, free=[True, True, False, False])
sources.set_default_bounds(model)
return model
def PowerLaw(*pars):
model = Models.PowerLaw(p=pars)
sources.set_default_bounds(model)
return model
def LogParabola(*pars, **kw):
m = Models.LogParabola(p=pars, **kw)
m.free[3] = False
return m
def PowerLaw(*pars):
model = Models.PowerLaw(p=pars)
return model
def FBconstant(f, b, **kw):
return Models.FrontBackConstant(f, b, **kw)
def Constant(*pars, **kw):
return Models.Constant(p=pars, **kw)
def PLSuperExpCutoff(*pars, **kw):
return Models.PLSuperExpCutoff(p=pars, **kw)
def ExpCutoff(*pars, **kw):
return Models.ExpCutoff(p=pars, **kw)
def PowerLaw(*pars, **kw):
return Models.PowerLaw(p=pars, **kw)
def PSR_default():
return Models.PLSuperExpCutoff(p=(1e-13, 1.25, 1500, 0.67),
free=[True, True, True, False])
def LogParabola(*pars):
model = Models.LogParabola(p=pars, free=[True, True, False, False])
return model
def pl():
index, beta = ce.PL_Index, 0
return Models.LogParabola(p=[flux, index, beta, pivot],
free=free)
def load_auxcat(self, tol=0.2):
""" modify the list of pointsources from entries in the auxcat: for now:
* add it not there
* move there, at new ra,dec
* remove if ra<0
"""
if self.auxcat is None or self.auxcat == '':
return
cat = self.auxcat
if not os.path.exists(cat):
cat = os.path.expandvars(os.path.join('$FERMI', 'catalog', cat))
if not os.path.exists(cat):
cat = os.path.join(self.folder, self.auxcat)
if not os.path.exists(cat):
raise Exception(
'auxilliary source catalog "%s" not found locally (%s) or in $FERMI/catalog'
% (self.auxcat, self.folder))
ext = os.path.splitext(cat)[-1]
if ext == '.pickle':
ss = pd.load(cat).itertuples()
dataframe = True
print 'loading auxcat from DataFrame'
elif ext == '.csv':
cc = pd.read_csv(cat)
cols = list(cc.columns)
ss = cc.itertuples()
i_eflux = cols.index('eflux') + 1
i_pindex = cols.index('pindex') + 1
i_e0 = cols.index('e0') + 1
dataframe = True
print 'loading auxcat from csv'
else:
ss = makerec.load(cat)
dataframe = False
names = [s.name for s in self.point_sources]
sdirs = [s.skydir for s in self.point_sources]
def check_near(sdir):
closest = np.degrees(np.min(np.array(map(sdir.difference, sdirs))))
return closest
toremove = []
print 'process auxcat %s' % cat
for s in ss:
if dataframe:
# from csv: construct full model from parameters
sname, sra, sdec = s[1:4]
n0 = s[i_eflux] / (s[i_e0]**2 * 1e6)
model = Models.LogParabola(p=[n0, s[i_pindex], 0., s[i_e0]])
model.free[2:] = False
else: # for comatibility: default from pater
sname, sra, sdec = s.name, s.ra, s.dec
model = self.newmodel
if type(self.newmodel) == types.StringType:
model = eval(self.newmodel)
elif model is None:
pass
else:
model = self.newmodel.copy(
) # make sure to get a new object
if not sname.startswith('SEED'): # allow underscores
sname = sname.replace('_', ' ')
if sname not in names:
skydir = SkyDir(float(sra), float(sdec))
close = check_near(skydir)
if close < tol:
print '\tsource %s close to another source, reject' % sname
continue
index = self.hpindex(skydir)
if model is not None:
model.free[
0] = True # must have at least one free parameter to be set up properly in an ROI
self.point_sources.append(
sources.PointSource(name=sname,
skydir=skydir,
index=index,
model=model))
print '\tadded new source %s at ROI %d (%.1f deg )' % (
sname, index, close)
else:
print '\t source %s is in the model:' % sname, # will remove if ra<0' % sname
ps = self.point_sources[names.index(sname)]
if float(sra) <= 0:
toremove.append(ps)
print ' removed.'
else:
newskydir = SkyDir(float(sra), float(sdec))
print 'moved from %s to %s' % (ps.skydir, newskydir)
ps.skydir = newskydir
for ps in toremove:
self.point_sources.remove(ps)
