def _weights2map_ ( weights_files ) :
from ostap.core.core import cpp, std, Ostap
MAP = std.map ( 'std::string', 'std::string' )
weights = WeightsFiles ( weights_files )
_weights = MAP()
for method , xml in weights.files.iteritems() :
_weights [ method ] = xml
assert not _weights .empty() , \
'Invalid MAP size %s for' % ( _weights.size() , weights )
assert not _weights.empty() , "Invalid weights_files: %s" % weights.files
return _weights
def combined_hdata(sample, varset, histograms, name='', title=''):
MAP = std.map('std::string', 'TH1*')
PAIR = std.pair('const std::string', 'TH1*')
mm = MAP()
for key in histograms:
mm.insert(PAIR(key, histograms[key]))
name = name if name else dsID()
title = title if title else 'Data for simultaneous fit/%s' % sample.GetName(
)
varlst = ROOT.RooArgList()
if isinstance(varset, ROOT.RooAbsReal): varlst.add(varset)
else:
for v in varset:
varlst.add(v)
return ROOT.RooDataHist(name, title, varlst, sample, mm)
def _inputs2map_ ( inputs ) :
"""Convert input structure to Ostap.TMVA.MAPS
"""
from ostap.core.core import cpp, std, Ostap
MAP = std.map ( 'std::string', 'std::string' )
_inputs = MAP()
assert isinstance ( inputs , ( dict , tuple , list ) ) , \
'Invalid type of "inputs": %s' % inputs
if isinstance ( inputs , dict ) :
for k , v in inputs.itertems() : _inputs[k] = v
elif isinstance ( inputs , ( tuple , list ) ) :
for i in inputs :
if isinstance ( i , str ) : k , v = i , i
else : k , v = i
_inputs[k] = v
##
assert not _inputs .empty() and _inputs.size() == len ( inputs ), \
'Invalid MAP size %s for %s' % ( _inputs.size() , inputs )
return _inputs
def addChoppingResponse(dataset,
chopper,
N,
inputs,
weights_files,
category_name='chopping',
prefix='tmva_',
suffix='_response',
aux=0.9):
"""
Helper function to add TMVA response into dataset
>>> tar_file = trainer.tar_file
>>> dataset = ...
>>> inputs = [ 'var1' , 'var2' , 'var2' ]
>>> dataset.addTMVAResponce ( inputs , tar_file , prefix = 'tmva_' )
"""
assert isinstance(N, int) and 1 < N < 10000, 'Invalid "N" %s' % N
if isinstance(chopper, str):
if chopper in dataset:
chopper = getattr(varset, chopper)
else:
varset = dataset.get()
varlist = ROOT.RooArgList()
for v in varset:
varlist.add(v)
chopper = ROOT.RooFormulaVar('chopping', chopper, varlist)
logger.debug('Create chopping function %s' % chopper)
assert isinstance(chopper,
ROOT.RooAbsReal), 'Invalid choper type %s' % chopper
category = ROOT.RooCategory(
category_name, 'Chopping category: (%s)%%%d' % (chopper.GetTitle(), N))
for i in range(N):
if N < 10: cn = category_name + '_%d' % i
if N < 100: cn = category_name + '_%02d' % i
elif N < 1000: cn = category_name + '_%03d' % i
elif N < 10000: cn = category_name + '_%04d' % i
else: cn = category_name + '_%d' % i
category.defineType(cn, i)
## decode inputs&weights
from ostap.tools.tmva import _inputs2map_, _weights2map_
_inputs = _inputs2map_(inputs)
files = WeightsFiles(weights_files).files
files_ = [_weights2map_(f) for f in files]
from ostap.core.core import cpp, std, Ostap
MAP = std.map('std::string', 'std::string')
MAPS = std.vector(MAP)
_maps = MAPS()
for m in files_:
_maps.push_back(m)
sc = Ostap.TMVA.addChoppingResponse(dataset, chopper, category, N, _inputs,
_maps, prefix, suffix, aux)
if sc.isFailure():
logger.error('Error from Ostap::TMVA::addChoppingResponse %s' % sc)
return sc
def combined_hdata ( sample ,
varset ,
histograms ,
name = '' ,
title = '' ) :
"""Create combined binned dataset for simultaneous fit
- combine 2D histograms:
>>> sample = ROOT.RooCategory ( 'sample' , 'fitting sample' , 'A' , 'B' )
>>> hA = ...
>>> hB = ...
>>> var = ROOT.RooRealVar ( ... )
>>> ds = combined_hdata ( sample , var , { 'A' : hA , 'B' : hB } )
- combine 2D histograms:
>>> sample = ROOT.RooCategory ( 'sample' , 'fitting sample' , 'A' , 'B' )
>>> hA = ...
>>> hB = ...
>>> xvar = ROOT.RooRealVar ( ... )
>>> yvar = ROOT.RooRealVar ( ... )
>>> ds = combined_hdata ( sample , (xvar, yvar) , { 'A' : hA , 'B' : hB } )
- combine 3D histograms:
>>> sample = ROOT.RooCategory ( 'sample' , 'fitting sample' , 'A' , 'B' )
>>> hA = ...
>>> hB = ...
>>> xvar = ROOT.RooRealVar ( ... )
>>> yvar = ROOT.RooRealVar ( ... )
>>> zvar = ROOT.RooRealVar ( ... )
>>> ds = combined_hdata ( sample , (xvar, yvar, zvar) , { 'A' : hA , 'B' : hB } )
"""
MAP = std.map ( 'std::string' , 'TH1*' )
PAIR = std.pair ( 'const std::string' , 'TH1*' )
mm = MAP()
d1 = 0
d2 = 0
d3 = 0
labels = sample.labels ()
for label in labels :
histo = histograms.pop ( label )
if isinstance ( histo , ROOT.TH3 ) : d3 += 1
elif isinstance ( histo , ROOT.TH2 ) : d2 += 1
elif isinstance ( histo , ROOT.TH1 ) : d1 += 1
mm.insert ( PAIR ( label , histo ) )
assert not historgams, 'Unknown histograms: %s' % histograms.keys()
assert ( d3 or d2 ) and ( d3 or d1 ) and ( d2 or d1 ), \
'Histograms of different dimensions cannot be combined !'
name = name if name else dsID()
title = title if title else 'Data for simultaneous fit/%s' % sample.GetName()
varlst = ROOT.RooArgList()
if isinstance ( varset , ROOT.RooAbsReal ) : varlst.add ( varset )
else :
for v in varset : varlst.add ( v )
assert ( d3 and 3 == len ( varlst ) ) or \
( d2 and 2 == len ( varlst ) ) or \
( d1 and 1 == len ( varlst ) ) , \
'Invalid dimension of dataset!'
return ROOT.RooDataHist ( name , title , varlst , sample , mm )