def MatrixInvertRoot(M, tol=1e-32):
M_root = TMatrixD(np.shape(M)[0], np.shape(M)[1])
for i in range(np.shape(M)[0]):
for j in range(np.shape(M)[1]):
M_root[i][j] = M[i,j]
M_root.SetTol(tol)
M_root.Invert()
M_inv = np.zeros_like(M)
for i in range(np.shape(M)[0]):
for j in range(np.shape(M)[1]):
M_inv[i,j] = M_root[i][j]
return M_inv
def Matrix(rows, cols, type='F'):
if type == 'F':
return TMatrixF(rows, cols)
elif type == 'D':
return TMatrixD(rows, cols)
raise TypeError("No matrix for type '%s'" % type)
def test1MatrixElementAssignment(self):
"""Matrix lookup has to be non-const to allow assigment"""
m = TMatrixD(5, 5)
self.assert_(not 'const' in type(m[0]).__name__)
# test assignment
m[1][2] = 3.
self.assertEqual(m[1][2], 3.)
m[1, 2] = 4.
self.assertEqual(m[1][2], 4.)
def test08ElementAccess(self):
"""Test access to elements in matrix and array objects."""
n = 3
v = TVectorF(n)
m = TMatrixD(n, n)
for i in range(n):
self.assertEqual(v[i], 0.0)
for j in range(n):
self.assertEqual(m[i][j], 0.0)
def get_condition_number(unfoldingObject):
probMatrix = unfoldingObject.GetProbabilityMatrix('ProbMatrix')
# probabilityMatrix = unfoldingObject.GetProbabilityMatrix( probMatrix, TUnfold.kHistMapOutputVert)
probMatrix.Print()
# probMatrix.Draw('COLZ')
raw_input()
nBins = probMatrix.GetNbinsX()
m = TMatrixD(nBins, nBins)
for xbin in range(1, probMatrix.GetNbinsX()):
for ybin in range(1, probMatrix.GetNbinsY()):
m[xbin, ybin] = probMatrix.GetBinContent(xbin, ybin)
svd = TDecompSVD(m)
svd.Decompose()
svd.Print()
sig = svd.GetSig()
sig.Print()
nSig = len(sig)
sigmaMax = sig[0]
sigmaMin = sig[nSig - 2]
condition = sigmaMax / max(0, sigmaMin)
# condition = 1
print condition
return condition
ex=np.array(len(data_y)*[0],dtype=np.float)
# ... and pass to TGraphErros object
gr=TGraphErrors(nPoints,data_x,data_y,ex,ey)
gr.SetTitle("TGraphErrors mit Fit")
gr.Draw("AP");
Pol=["pol1","pol2","pol3","pol4","pol5","pol6","pol7"]
# Polifit
for i in range(len(polynom)):
gr.Fit(polynom[i],"V")
c1.Update()
gr.Draw('AP')
fitrp = TVirtualFitter.GetFitter()
nPar = fitrp.GetNumberTotalParameters()
covmat = TMatrixD(nPar, nPar,fitrp.GetCovarianceMatrix())
print('The Covariance Matrix is: ')
covmat.Print()
cormat = TMatrixD(covmat)
for i in range(nPar):
for j in range(nPar):
cormat[i][j] = cormat[i][j] / (np.sqrt(covmat[i][i]) * np.sqrt(covmat[j][j]))
print('The Correlation Matrix is: ')
cormat.Print()
raw_input('Press <ret> to continue -> ')
#Legendre Poly
for i in range(len(legendre)):
gr.Fit(legendre[i],"V")
c1.Update()
def process(self, event):
if not eval(self.skimFunction):
return True
decayMode1 = -1
decayMode2 = -1
if self.cfg_ana.l1type == 'tau':
decayMode1 = event.leg1.decayMode()
if self.cfg_ana.l2type == 'tau':
decayMode2 = event.leg2.decayMode()
# RIC: some PF muons/electron can get the wrong mass assigned.
# Peg their masses to the PDG values
if self.cfg_ana.l1type == 'muon':
mass1 = 0.10566 # PDG mass [GeV]
elif self.cfg_ana.l1type == 'ele':
mass1 = 0.51100e-3 # PDG mass [GeV]
else:
mass1 = event.leg1.mass()
if self.cfg_ana.l2type == 'muon':
mass2 = 0.10566 # PDG mass [GeV]
elif self.cfg_ana.l2type == 'ele':
mass2 = 0.51100e-3 # PDG mass [GeV]
else:
mass2 = event.leg2.mass()
leg1 = measuredTauLepton(self.legType[self.cfg_ana.l1type],
event.leg1.pt(), event.leg1.eta(),
event.leg1.phi(), mass1, decayMode1)
leg2 = measuredTauLepton(self.legType[self.cfg_ana.l2type],
event.leg2.pt(), event.leg2.eta(),
event.leg2.phi(), mass2, decayMode2)
measuredLeptons = std.vector('svFitStandalone::MeasuredTauLepton')()
# RIC: not really needed since SVfit internally sorts the inputs
if hasattr(self.cfg_ana, 'order') and self.cfg_ana.order == '12':
measuredLeptons.push_back(leg1)
measuredLeptons.push_back(leg2)
else:
measuredLeptons.push_back(leg2)
measuredLeptons.push_back(leg1)
metcov = TMatrixD(2, 2)
a_metcov = array.array('d', [
event.diLepton.mvaMetSig(0, 0),
event.diLepton.mvaMetSig(1, 0),
event.diLepton.mvaMetSig(0, 1),
event.diLepton.mvaMetSig(1, 1)
])
metcov.SetMatrixArray(a_metcov)
mex = event.diLepton.met().px()
mey = event.diLepton.met().py()
svfit = SVfitAlgo(measuredLeptons, mex, mey, metcov,
2 * self.cfg_ana.verbose)
if hasattr(self.cfg_ana, 'integrateOverVisPtResponse') and \
self.cfg_ana.integrateOverVisPtResponse:
shift = not((self.cfg_ana.l1type == 'tau' and \
leg1.decayMode() not in [0, 1, 10]) or \
(self.cfg_ana.l2type == 'tau' and \
leg2.decayMode() not in [0, 1, 10]))
svfit.shiftVisPt(shift, self.inputFile_visPtResolution)
# add an additional logM(tau,tau) term to the nll
# to suppress tails on M(tau,tau) (default is false)
# svfit.addLogM(False)
if self.cfg_ana.integration == 'VEGAS':
svfit.integrateVEGAS()
elif self.cfg_ana.integration == 'MarkovChain':
svfit.integrateMarkovChain()
else:
print 'The integration method must be defined in the cfg as "integration".'
print 'Options: [VEGAS, MarkovChain]'
raise
# debug
if self.cfg_ana.verbose:
if abs(event.diLepton.svfitMass() - svfit.mass()) > 0.01:
print 'WARNING: run {RUN}, lumi {LUMI}, event {EVT}'.format(
RUN=str(event.run),
LUMI=str(event.lumi),
EVT=str(event.eventId))
print 'precomputed svfit mass ', event.diLepton.svfitMass()
print 'svfit mass computed here ', svfit.mass()
# method override
event.diLepton.svfitMass = svfit.mass
event.diLepton.svfitMassError = svfit.massUncert
event.diLepton.svfitTransverseMass = svfit.transverseMass
# add also the pt, eta and phi as computed by SVfit
if self.cfg_ana.integration == 'MarkovChain':
event.diLepton.svfitPt = svfit.pt
event.diLepton.svfitPtError = svfit.ptUncert
event.diLepton.svfitEta = svfit.eta
event.diLepton.svfitPhi = svfit.phi
event.diLepton.svfitMET = svfit.fittedMET
event.diLepton.svfitTaus = svfit.fittedTauLeptons