def tmva_process(info, data):
"""
Create a TMVA reader and predict data.
:param rep.estimators.tmva._AdditionalInformationPredict info: additional information
:param pandas.DataFrame data: data to predict
"""
import ROOT
reader = ROOT.TMVA.Reader()
for feature in data.columns:
reader.AddVariable(feature, array.array('f', [0.]))
model_type, sigmoid_function = info.model_type
reader.BookMVA(info.method_name, info.xml_file)
signal_efficiency = None
if model_type == 'classification' and sigmoid_function is not None and 'sig_eff' in sigmoid_function:
signal_efficiency = float(sigmoid_function.strip().split('=')[1])
assert 0.0 <= signal_efficiency <= 1., 'signal efficiency must be in [0, 1], not {}'.format(
signal_efficiency)
if signal_efficiency is not None:
predictions = evaluate_reader(reader, info.method_name, data, aux=signal_efficiency)
else:
predictions = evaluate_reader(reader, info.method_name, data)
return predictions
def tmva_process(info, data):
"""
Create a TMVA reader and predict data.
:param rep.estimators.tmva._AdditionalInformationPredict info: additional information
:param pandas.DataFrame data: data to predict
"""
import ROOT
reader = ROOT.TMVA.Reader()
for feature in data.columns:
reader.AddVariable(feature, array.array('f', [0.]))
model_type, sigmoid_function = info.model_type
reader.BookMVA(info.method_name, info.xml_file)
signal_efficiency = None
if model_type == 'classification' and sigmoid_function is not None and 'sig_eff' in sigmoid_function:
signal_efficiency = float(sigmoid_function.strip().split('=')[1])
assert 0.0 <= signal_efficiency <= 1., 'signal efficiency must be in [0, 1], not {}'.format(
signal_efficiency)
if signal_efficiency is not None:
predictions = evaluate_reader(reader,
info.method_name,
data,
aux=signal_efficiency)
else:
predictions = evaluate_reader(reader, info.method_name, data)
return predictions
def ImportXml(self,train,test,ModelName):
""" Import the given Xml model """
reader = ROOT.TMVA.Reader()
#Filling reader used for eval
for Var in self.DataSet.LVariables:
reader.AddVariable(Var,array('f', [0.]))
xml = './dataset/weights/TMVAClassification_'+ModelName+'.weights.xml'
reader.BookMVA(ModelName,xml)
return evaluate_reader(reader,ModelName,test.Events), evaluate_reader(reader,ModelName,train.Events) # retuns the Classifcation Score
def plot_clf_results_tmva(reader,
x_train,
y_train,
w_train,
x_test,
y_test,
w_test,
nbins=30,
figname="BDTOutput_tmva.png",
verbose="False"):
decisions = []
weights = []
for x, y, w in ((x_train, y_train, w_train), (x_test, y_test, w_test)):
w *= 1. / np.sum(w)
dsig = evaluate_reader(reader, "BDT", x[y > 0.5])
wsig = w[y > 0.5]
dbkg = evaluate_reader(reader, "BDT", x[y < 0.5])
wbkg = w[y < 0.5]
decisions += [dsig, dbkg]
weights += [wsig, wbkg]
plot_clf_results(tuple(decisions), tuple(weights), nbins, figname, verbose)
def get_tmva_test_results(directory, variables, name=''):
# TMVA reader
reader = r.TMVA.Reader()
for var in variables:
#vtype = 'i' if var in ['nJet','tau0_decaymode','tau1_decaymode','ntags','ntags_loose'] else 'f'
reader.AddVariable(var, array('f', [0]))
reader.BookMVA('BDT', directory + 'weights/TMVA_BDT.weights.xml')
# Get testing dataset
filename = directory + 'tmva_output.root'
x_test = rec2array(root2array(filename, 'TestTree', variables))
y_test = 1 - root2array(filename, 'TestTree', 'classID')
w_test = root2array(filename, 'TestTree', 'weight')
y_decision = evaluate_reader(reader, "BDT", x_test)
return (y_test, y_decision, w_test, name)
def evaluate(config, tree, names, transform=None):
output = []
dtype = []
for name in names:
setup = load(config, name.split("_")[1])
data = rec2array(tree2array(tree.raw(), list(transform(setup["variables"])) if transform else setup["variables"]))
if name.startswith("sklearn"):
fn = os.path.join(config["mvadir"], name + ".pkl")
with open(fn, 'rb') as fd:
bdt, label = pickle.load(fd)
scores = []
if len(data) > 0:
scores = bdt.predict_proba(data)[:, 1]
output += [scores]
dtype += [(name, 'float64')]
fn = os.path.join(config["mvadir"], name + ".xml")
reader = r.TMVA.Reader("Silent")
for var in setup['variables']:
reader.AddVariable(var, array('f', [0.]))
reader.BookMVA("BDT", fn)
scores = evaluate_reader(reader, "BDT", data)
output += [scores]
dtype += [(name.replace("sklearn", "tmvalike"), 'float64')]
f = r.TFile(os.path.join(config.get("mvadir", config.get("indir", config["outdir"])), "mapping.root"), "READ")
if f.IsOpen():
likelihood = f.Get("hTargetBinning")
def lh(values):
return likelihood.GetBinContent(likelihood.FindBin(*values))
indices = dict((v, n) for n, (v, _) in enumerate(dtype))
tt = output[indices['tmvalike_tt']]
ttZ = output[indices['tmvalike_ttZ']]
if len(tt) == 0:
output += [[]]
else:
output += [np.apply_along_axis(lh, 1, np.array([tt, ttZ]).T)]
dtype += [('tmvalike_likelihood', 'float64')]
f.Close()
data = np.array(zip(*output), dtype)
tree.mva(array2tree(data))
def test(self, test_data, classification_variables, training_sample):
"""
Definition:
-----------
Testing method for RootTMVA; it loads the latest model from the "weights" sub-folder
Args:
-----
data = dictionary, containing "X", "y", "w" for the set to evaluate performance on, where:
X = ndarray of dim (# examples, # features)
y = array of dim (# examples) with target values
w = array of dim (# examples) with event weights
classification_variables = list of names of variables used for classification
training_sample = string that specifies the file name of the sample to use as a training (e.g. "SM_merged" or "X350_hh")
Returns:
--------
yhat = the array of BDT outputs, of dimensions (n_events)
"""
logging.getLogger("root_tmva").info("Evaluating performance...")
# -- Construct reader and add variables to it:
logging.getLogger("root_tmva").info("Construct TMVA reader and add variables to it")
reader = TMVA.Reader()
for v_name in classification_variables:
reader.AddVariable(v_name, array.array("f", [0]))
# -- Load TMVA results
reader.BookMVA("BDT", os.path.join(self.output_directory, training_sample, self.name, "weights", "TMVAClassification_BDT.weights.xml"))
# reader.BookMVA("BDT", os.path.join(self.output_directory, training_sample, self.name, "weights", "TMVAClassification_Fisher.weights.xml"))
# -- Load skl_BDT results (used for testing only)
# reader.BookMVA("BDT", os.path.join(self.output_directory, training_sample, "skl_BDT", "classifier", "skl_BDT_TMVA.weights.xml"))
yhat = evaluate_reader(reader, "BDT", test_data["X"])
# -- add binary classification labels
yhat_class = np.zeros(len(yhat))
yhat_class[yhat >= 0] = 1
return yhat, yhat_class
# Train an MLP
if ROOT_VERSION >= '6.07/04':
BookMethod = factory.BookMethod
else:
BookMethod = TMVA.Factory.BookMethod
BookMethod(data, 'MLP', 'MLP',
'NeuronType=tanh:NCycles=200:HiddenLayers=N+2,2:'
'TestRate=5:EstimatorType=MSE')
factory.TrainAllMethods()
# Classify the test dataset with the BDT
reader = TMVA.Reader()
for n in range(2):
reader.AddVariable('f{0}'.format(n), array('f', [0.]))
reader.BookMVA('MLP', 'weights/classifier_MLP.weights.xml')
class_proba = evaluate_reader(reader, 'MLP', X_test)
# Plot the decision boundaries
plot_colors = "rgb"
plot_step = 0.02
class_names = "ABC"
cmap = plt.get_cmap('Paired')
fig = plt.figure(figsize=(5, 5))
fig.patch.set_alpha(0)
x_min, x_max = X[:, 0].min() - 1, X[:, 0].max() + 1
y_min, y_max = X[:, 1].min() - 1, X[:, 1].max() + 1
xx, yy = np.meshgrid(np.arange(x_min, x_max, plot_step),
np.arange(y_min, y_max, plot_step))
Z = evaluate_reader(reader, 'MLP', np.c_[xx.ravel(), yy.ravel()])
if ROOT_VERSION >= '6.07/04':
BookMethod = factory.BookMethod
else:
BookMethod = TMVA.Factory.BookMethod
BookMethod(data, 'BDT', 'BDT1',
'nCuts=20:NTrees=1:MaxDepth=4:BoostType=AdaBoostR2:'
'SeparationType=RegressionVariance')
BookMethod(data, 'BDT', 'BDT2',
'nCuts=20:NTrees=300:MaxDepth=4:BoostType=AdaBoostR2:'
'SeparationType=RegressionVariance')
factory.TrainAllMethods()
# Predict the regression target
reader = TMVA.Reader()
reader.AddVariable('x', array('f', [0.]))
reader.BookMVA('BDT1', 'weights/regressor_BDT1.weights.xml')
reader.BookMVA('BDT2', 'weights/regressor_BDT2.weights.xml')
y_1 = evaluate_reader(reader, 'BDT1', X)
y_2 = evaluate_reader(reader, 'BDT2', X)
# Plot the results
plt.figure()
plt.scatter(X, y, c="k", label="training samples")
plt.plot(X, y_1, c="g", label="1 tree", linewidth=2)
plt.plot(X, y_2, c="r", label="300 trees", linewidth=2)
plt.xlabel("data")
plt.ylabel("target")
plt.title("Boosted Decision Tree Regression")
plt.legend()
plt.savefig('RootNumpy')
factory.BookMethod('BDT', 'BDT',
'NTrees=100:nEventsMin=30:MaxDepth=3:BoostType=AdaBoost:AdaBoostBeta=0.2:SeparationType=GiniIndex:nCuts=10:PruneStrength=2:PruneMethod=ExpectedError')
#factory.BookMethod('kBDT', 'BDT',
# 'Fisher:VarTransform=None:CreateMVAPdfs:'
# 'PDFInterpolMVAPdf=Spline2:NbinsMVAPdf=50:'
# 'NsmoothMVAPdf=10')
print 'added all the events and booked the method'
factory.TrainAllMethods()
print 'trained'
# Classify the test dataset with the classifier
reader = TMVA.Reader()
for n in range(n_vars):
reader.AddVariable('f{0}'.format(n), array('f', [0.]))
reader.BookMVA('BDT', 'weights/classifier_BDT.weights.xml')
twoclass_output = evaluate_reader(reader, 'BDT', X_test)
plot_colors = "br"
plot_step = 0.02
class_names = "AB"
cmap = plt.get_cmap('bwr')
plt.figure(figsize=(10, 5))
# Plot the decision boundaries
plt.subplot(121)
x_min, x_max = X[:, 0].min() - 1, X[:, 0].max() + 1
y_min, y_max = X[:, 1].min() - 1, X[:, 1].max() + 1
'''
xx, yy = np.meshgrid(np.arange(x_min, x_max, plot_step),
np.arange(y_min, y_max, plot_step))
# The following line is necessary if events have been added individually:
factory.PrepareTrainingAndTestTree(TCut('1'), 'NormMode=EqualNumEvents')
# Train a classifier
factory.BookMethod('Fisher', 'Fisher',
'Fisher:VarTransform=None:CreateMVAPdfs:'
'PDFInterpolMVAPdf=Spline2:NbinsMVAPdf=50:'
'NsmoothMVAPdf=10')
factory.TrainAllMethods()
# Classify the test dataset with the classifier
reader = TMVA.Reader()
for n in range(n_vars):
reader.AddVariable('f{0}'.format(n), array('f', [0.]))
reader.BookMVA('Fisher', 'weights/classifier_Fisher.weights.xml')
twoclass_output = evaluate_reader(reader, 'Fisher', X_test)
plot_colors = "br"
plot_step = 0.02
class_names = "AB"
cmap = plt.get_cmap('bwr')
plt.figure(figsize=(10, 5))
# Plot the decision boundaries
plt.subplot(121)
x_min, x_max = X[:, 0].min() - 1, X[:, 0].max() + 1
y_min, y_max = X[:, 1].min() - 1, X[:, 1].max() + 1
xx, yy = np.meshgrid(np.arange(x_min, x_max, plot_step),
np.arange(y_min, y_max, plot_step))
def main(weights, picklename, filename, treename='bTag_AntiKt2PV0TrackJets'):
'''
evaluate the tmva method after transforming input data into right format
Args:
-----
weights: .xml file out of mv2 training containing bdt parameters
picklename: name of the output pickle to store new mv2 values
filename: .root file with ntuples used to evaluate the tmva method
treename: (optional) name of the TTree to consider
Returns:
--------
status
Raises:
-------
nothing yet, but to be improved
'''
print 'Parsing XML file...'
# -- Load XML file
tree = ET.parse(weights)
root = tree.getroot()
# -- Get list of variable names from XML file
var_list = [
var.attrib['Label']
for var in root.findall('Variables')[0].findall('Variable')
]
# -- Count the input variables that go into MV2:
n_vars = len(var_list)
print 'Loading .root file for evaluation...'
# -- Get ntuples:
df = pup.root2panda(
filename,
treename,
branches=[
'jet_pt', 'jet_eta', 'jet_phi', 'jet_m', 'jet_ip2d_pu',
'jet_ip2d_pc', 'jet_ip2d_pb', 'jet_ip3d_pu', 'jet_ip3d_pc',
'jet_ip3d_pb', 'jet_sv1_vtx_x', 'jet_sv1_vtx_y', 'jet_sv1_vtx_z',
'jet_sv1_ntrkv', 'jet_sv1_m', 'jet_sv1_efc', 'jet_sv1_n2t',
'jet_sv1_sig3d', 'jet_jf_n2t', 'jet_jf_ntrkAtVx', 'jet_jf_nvtx',
'jet_jf_nvtx1t', 'jet_jf_m', 'jet_jf_efc', 'jet_jf_sig3d',
'jet_jf_deta', 'jet_jf_dphi', 'PVx', 'PVy', 'PVz'
])
# -- Insert default values, calculate MV2 variables from the branches in df
df = transformVars(df)
# -- Map ntuple names to var_list
names_mapping = {
'pt': 'jet_pt',
'abs(eta)': 'abs(jet_eta)',
'ip2': 'jet_ip2',
'ip2_c': 'jet_ip2_c',
'ip2_cu': 'jet_ip2_cu',
'ip3': 'jet_ip3',
'ip3_c': 'jet_ip3_c',
'ip3_cu': 'jet_ip3_cu',
'sv1_ntkv': 'jet_sv1_ntrkv',
'sv1_mass': 'jet_sv1_m',
'sv1_efrc': 'jet_sv1_efc',
'sv1_n2t': 'jet_sv1_n2t',
'sv1_Lxy': 'jet_sv1_Lxy',
'sv1_L3d': 'jet_sv1_L3d',
'sv1_sig3': 'jet_sv1_sig3d',
'sv1_dR': 'jet_sv1_dR',
'jf_n2tv': 'jet_jf_n2t',
'jf_ntrkv': 'jet_jf_ntrkAtVx',
'jf_nvtx': 'jet_jf_nvtx',
'jf_nvtx1t': 'jet_jf_nvtx1t',
'jf_mass': 'jet_jf_m',
'jf_efrc': 'jet_jf_efc',
'jf_dR': 'jet_jf_dR',
'jf_sig3': 'jet_jf_sig3d'
}
print 'Initializing TMVA...'
# -- TMVA: Initialize reader, add empty variables and weights from training
reader = TMVA.Reader()
for n in range(n_vars):
reader.AddVariable(var_list[n], array('f', [0]))
reader.BookMVA('BDTG akt2', weights)
print 'Creating feature matrix...'
# -- Get features for each event and store them in X_test
X_buf = []
for event in df[[names_mapping[var] for var in var_list]].values:
X_buf.extend(
np.array([normalize_type(jet) for jet in event]).T.tolist())
X_test = np.array(X_buf)
print 'Evaluating!'
# -- TMVA: Evaluate!
twoclass_output = evaluate_reader(reader, 'BDTG akt2', X_test)
# -- Reshape the MV2 output into event-jet format
reorganized = match_shape(twoclass_output, df['jet_pt'])
import cPickle
print 'Saving new MV2 weights in {}'.format(picklename)
cPickle.dump(reorganized, open(picklename, 'wb'))
# -- Write the new branch to the tree (currently de-activated)
#add_branch(reorganized, filename, treename, 'jet_mv2c20_new')
print 'Done. Success!'
return 0
def cl3d_fixtures(clusters, tcs):
# print clusters.columns
# for backward compatibility
if clusters.empty:
return clusters
clusters.rename(columns={
'clusters_id': 'clusters',
'clusters_n': 'nclu'
},
inplace=True)
# clusters['hwQual'] = clusters['quality']
do_compute_hoe = False
do_compute_layer_energy = False
if 'hoe' not in clusters.columns:
do_compute_hoe = True
if 'layer_energy' not in clusters.columns:
do_compute_layer_energy = True
def compute_layer_energy3(cluster, do_layer_energy=True, do_hoe=False):
components = tcs[tcs.id.isin(cluster.clusters)]
hist, bins = np.histogram(components.layer.values,
bins=range(0, 29, 2),
weights=components.energy.values)
results = []
if do_layer_energy:
results.append(hist)
if do_hoe:
em_energy = np.sum(hist)
hoe = -1
if em_energy != 0:
hoe = max(0, cluster.energy - em_energy) / em_energy
results.append(hoe)
return results
def compute_layer_energy2(cluster, do_layer_energy=True, do_hoe=False):
components = tcs[tcs.id.isin(cluster.clusters)]
hist, bins = np.histogram(components.layer.values,
bins=range(0, 29, 2),
weights=components.energy.values)
if do_layer_energy:
cluster['layer_energy'] = hist
if do_hoe:
em_energy = np.sum(hist)
hoe = -1
if em_energy != 0:
hoe = max(0, cluster.energy - em_energy) / em_energy
cluster['hoe'] = hoe
return cluster
if do_compute_hoe or do_compute_layer_energy:
# clusters = clusters.apply(lambda cl: compute_layer_energy2(cl,
# do_compute_layer_energy,
# do_compute_hoe), axis=1)
new_columns = []
if do_compute_layer_energy:
new_columns.append('layer_energy')
if do_compute_hoe:
new_columns.append('hoe')
clusters[new_columns] = clusters.apply(
lambda cl: compute_layer_energy3(cl, do_compute_layer_energy,
do_compute_hoe),
result_type='expand',
axis=1)
clusters['ptem'] = clusters.pt / (1 + clusters.hoe)
clusters['eem'] = clusters.energy / (1 + clusters.hoe)
if False:
clusters['bdt_pu'] = rnptmva.evaluate_reader(
classifiers.mva_pu_classifier_builder(), 'BDT',
clusters[['pt', 'eta', 'maxlayer', 'hoe', 'emaxe', 'szz']])
clusters['bdt_pi'] = rnptmva.evaluate_reader(
classifiers.mva_pi_classifier_builder(), 'BDT',
clusters[['pt', 'eta', 'maxlayer', 'hoe', 'emaxe', 'szz']])
return clusters
print 'Training takes ', stop - start, 's'
if not args.quiet:
print 'Training done'
print 'Outputs in directory : ', args.outdir
# evaluate training results
if args.evaluate:
factory.TestAllMethods()
factory.EvaluateAllMethods()
reader = TMVA.Reader()
for v in var:
#vtype = 'i' if v in ['nJet','tau0_decaymode','tau1_decaymode','ntags','ntags_loose'] else 'f'
reader.AddVariable(v, array('f', [0]))
reader.BookMVA('BDT', args.outdir + 'weights/TMVA_BDT.weights.xml')
y_decision = evaluate_reader(reader, "BDT", x_test)
util.plot_clf_results_tmva(reader,
x_train,
y_train,
w_train,
x_test,
y_test,
w_test,
figname=args.outdir + "bdtoutput.png",
verbose=(not args.quiet))
util.plot_roc((y_test, y_decision, w_test),
figname=args.outdir + 'roc.png')
if ROOT_VERSION >= '6.07/04':
BookMethod = factory.BookMethod
else:
BookMethod = TMVA.Factory.BookMethod
BookMethod(
data, 'BDT', 'BDT1', 'nCuts=20:NTrees=1:MaxDepth=4:BoostType=AdaBoostR2:'
'SeparationType=RegressionVariance')
BookMethod(
data, 'BDT', 'BDT2', 'nCuts=20:NTrees=300:MaxDepth=4:BoostType=AdaBoostR2:'
'SeparationType=RegressionVariance')
factory.TrainAllMethods()
# Predict the regression target
reader = TMVA.Reader()
reader.AddVariable('x', array('f', [0.]))
reader.BookMVA('BDT1', 'weights/regressor_BDT1.weights.xml')
reader.BookMVA('BDT2', 'weights/regressor_BDT2.weights.xml')
y_1 = evaluate_reader(reader, 'BDT1', X)
y_2 = evaluate_reader(reader, 'BDT2', X)
# Plot the results
plt.figure()
plt.scatter(X, y, c="k", label="training samples")
plt.plot(X, y_1, c="g", label="1 tree", linewidth=2)
plt.plot(X, y_2, c="r", label="300 trees", linewidth=2)
plt.xlabel("data")
plt.ylabel("target")
plt.title("Boosted Decision Tree Regression")
plt.legend()
plt.savefig('RootNumpy')
reader2j.AddVariable("hmmphics", arr())
reader2j.AddVariable("j1pt", arr())
reader2j.AddVariable("j1eta", arr())
reader2j.AddVariable("j2pt", arr())
reader2j.AddVariable("detajj", arr())
reader2j.AddVariable("dphijj", arr())
reader2j.AddVariable("mjj", arr())
reader2j.AddVariable("met", arr())
reader2j.AddVariable("zepen", arr())
reader2j.AddVariable("njets", arr())
reader2j.AddVariable("drmj", arr())
reader2j.AddVariable("m1ptOverMass", arr())
reader2j.AddVariable("m2ptOverMass", arr())
reader2j.AddVariable("m1eta", arr())
reader2j.AddVariable("m2eta", arr())
reader2j.AddSpectator("hmerr", arr())
reader2j.AddSpectator("weight", arr())
reader2j.AddSpectator("hmass", arr())
reader2j.AddSpectator("nbjets", arr())
reader2j.AddSpectator("bdtucsd_inclusive", arr())
reader2j.AddSpectator("bdtucsd_01jet", arr())
reader2j.AddSpectator("bdtucsd_2jet", arr())
for n in range(n_vars):
reader.AddVariable('f{0}'.format(n), )
reader.BookMVA('TMVAClassification_BDTG.weights.2jet_bveto',
'TMVAClassification_BDTG.weights.2jet_bveto.xml')
X = np.zeros((1, 24), dtype='f')
Z = evaluate_reader(reader, 'TMVAClassification_BDTG.weights.2jet_bveto', X)
def main(weights, picklename, filename, treename = 'bTag_AntiKt2PV0TrackJets'):
'''
evaluate the tmva method after transforming input data into right format
Args:
-----
weights: .xml file out of mv2 training containing bdt parameters
picklename: name of the output pickle to store new mv2 values
filename: .root file with ntuples used to evaluate the tmva method
treename: (optional) name of the TTree to consider
Returns:
--------
status
Raises:
-------
nothing yet, but to be improved
'''
print 'Parsing XML file...'
# -- Load XML file
tree = ET.parse(weights)
root = tree.getroot()
# -- Get list of variable names from XML file
var_list = [var.attrib['Label'] for var in root.findall('Variables')[0].findall('Variable')]
# -- Count the input variables that go into MV2:
n_vars = len(var_list)
print 'Loading .root file for evaluation...'
# -- Get ntuples:
df = pup.root2panda(filename, treename, branches = ['jet_pt', 'jet_eta','jet_phi', 'jet_m', 'jet_ip2d_pu',
'jet_ip2d_pc', 'jet_ip2d_pb', 'jet_ip3d_pu', 'jet_ip3d_pc','jet_ip3d_pb',
'jet_sv1_vtx_x', 'jet_sv1_vtx_y', 'jet_sv1_vtx_z', 'jet_sv1_ntrkv',
'jet_sv1_m','jet_sv1_efc','jet_sv1_n2t','jet_sv1_sig3d',
'jet_jf_n2t','jet_jf_ntrkAtVx','jet_jf_nvtx','jet_jf_nvtx1t','jet_jf_m',
'jet_jf_efc','jet_jf_sig3d', 'jet_jf_deta', 'jet_jf_dphi', 'PVx', 'PVy', 'PVz' ])
# -- Insert default values, calculate MV2 variables from the branches in df
df = transformVars(df)
# -- Map ntuple names to var_list
names_mapping = {
'pt':'jet_pt',
'abs(eta)':'abs(jet_eta)',
'ip2':'jet_ip2',
'ip2_c':'jet_ip2_c',
'ip2_cu':'jet_ip2_cu',
'ip3':'jet_ip3',
'ip3_c':'jet_ip3_c',
'ip3_cu':'jet_ip3_cu',
'sv1_ntkv':'jet_sv1_ntrkv',
'sv1_mass':'jet_sv1_m',
'sv1_efrc':'jet_sv1_efc',
'sv1_n2t':'jet_sv1_n2t',
'sv1_Lxy':'jet_sv1_Lxy',
'sv1_L3d':'jet_sv1_L3d',
'sv1_sig3':'jet_sv1_sig3d',
'sv1_dR': 'jet_sv1_dR',
'jf_n2tv':'jet_jf_n2t',
'jf_ntrkv':'jet_jf_ntrkAtVx',
'jf_nvtx':'jet_jf_nvtx',
'jf_nvtx1t':'jet_jf_nvtx1t',
'jf_mass':'jet_jf_m',
'jf_efrc':'jet_jf_efc',
'jf_dR':'jet_jf_dR',
'jf_sig3':'jet_jf_sig3d'
}
print 'Initializing TMVA...'
# -- TMVA: Initialize reader, add empty variables and weights from training
reader = TMVA.Reader()
for n in range(n_vars):
reader.AddVariable(var_list[n], array('f', [0] ) )
reader.BookMVA('BDTG akt2', weights)
print 'Creating feature matrix...'
# -- Get features for each event and store them in X_test
X_buf = []
for event in df[[names_mapping[var] for var in var_list]].values:
X_buf.extend(np.array([normalize_type(jet) for jet in event]).T.tolist())
X_test = np.array(X_buf)
print 'Evaluating!'
# -- TMVA: Evaluate!
twoclass_output = evaluate_reader(reader, 'BDTG akt2', X_test)
# -- Reshape the MV2 output into event-jet format
reorganized = match_shape(twoclass_output, df['jet_pt'])
import cPickle
print 'Saving new MV2 weights in {}'.format(picklename)
cPickle.dump(reorganized, open(picklename, 'wb'))
# -- Write the new branch to the tree (currently de-activated)
#add_branch(reorganized, filename, treename, 'jet_mv2c20_new')
print 'Done. Success!'
return 0