def convert_bdt__AdaBoost(sklearn_bdt_clf, input_var_list, tmva_outfile_xml):
# classificator
clf = sklearn_bdt_clf
if clf.n_classes_ != 2:
sys.exit("Error: Number of classes in sklearn classifier is not equal 2.")
# Order of variables must be _exactly_ as in the training numpy array
# E.g.
# var_list = [
# ('m_el_pt', 'F'),
# ('m_el_eta', 'F'),
# ('m_el_sigd0PV', 'F'),
# ('m_el_z0SinTheta', 'F'),
# ('m_el_etcone20Dpt', 'F'),
# ('m_el_ptcone20Dpt', 'F')
# ]
var_list = input_var_list
# Run-time parameters
NTrees = clf.n_estimators
#<MethodSetup Method="BDT::BDT">
# <GeneralInfo>
# <Info name="Creator" value="musthero"/>
# <Info name="AnalysisType" value="Classification"/>
# <Options>
# <Option name="NodePurityLimit" modified="No">5.000000e-01</Option>
#<Weights NTrees="2" AnalysisType="0">
#-------------
# <MethodSetup>
MethodSetup = ET.Element("MethodSetup", Method="BDT::BDT")
#<Variables>
Variables = ET.SubElement(MethodSetup, "Variables", NVar=str(len(var_list)))
for ind, val in enumerate(var_list):
name = val[0]
var_type = val[1]
Variable = ET.SubElement(Variables, "Variable", VarIndex=str(ind), Type=val[1],
Expression=name, Label=name, Title=name, Unit="", Internal=name,
Min="0.0e+00", Max="0.0e+00")
# <GeneralInfo>
GeneralInfo = ET.SubElement(MethodSetup, "GeneralInfo")
Info_Creator = ET.SubElement(GeneralInfo, "Info", name="Creator", value="Koza4ok (skTMVA)")
Info_AnalysisType = ET.SubElement(GeneralInfo, "Info", name="AnalysisType", value="Classification")
# <Options>
Options = ET.SubElement(MethodSetup, "Options")
Option_NodePurityLimit = ET.SubElement(Options, "Option", name="NodePurityLimit", modified="No").text = str(NodePurityLimit)
Option_BoostType = ET.SubElement(Options, "Option", name="BoostType", modified="Yes").text = "AdaBoost"
# <Weights>
Weights = ET.SubElement(MethodSetup, "Weights", NTrees=str(NTrees), AnalysisType="0")
for idx, dt in enumerate(clf.estimators_):
tree_weight = clf.estimator_weights_[idx]
# <BinaryTree type="DecisionTree" boostWeight="9.2106320437773737e-01" itree="0">
BinaryTree = ET.SubElement(Weights, "BinaryTree", type="DecisionTree", boostWeight=str(tree_weight), itree=str(idx))
build_xml_tree__AdaBoost(dt, 0, "s", -1, BinaryTree)
# Create XML-tree structure and save it to file
tree = ET.ElementTree(MethodSetup)
tree.write(tmva_outfile_xml)
def convert_bdt__Grad(sklearn_bdt_clf, input_var_list, tmva_outfile_xml):
# classificator
clf = sklearn_bdt_clf
if clf.loss_.K != 1:
sys.exit("Error: Only binary classification is supported for regression trees.")
if clf.n_classes_ != 2:
sys.exit("Error: Number of classes in sklearn classifier is not equal 2.")
# Order of variables must be _exactly_ as in the training numpy array
# E.g.
# var_list = [
# ('m_el_pt', 'F'),
# ('m_el_eta', 'F'),
# ('m_el_sigd0PV', 'F'),
# ('m_el_z0SinTheta', 'F'),
# ('m_el_etcone20Dpt', 'F'),
# ('m_el_ptcone20Dpt', 'F')
# ]
var_list = input_var_list
# Run-time parameters
NTrees = clf.n_estimators
#<MethodSetup Method="BDT::BDT">
# <GeneralInfo>
# <Info name="Creator" value="musthero"/>
# <Info name="AnalysisType" value="Classification"/>
# <Options>
# <Option name="NodePurityLimit" modified="No">5.000000e-01</Option>
#<Weights NTrees="2" AnalysisType="0">
#-------------
# <MethodSetup>
MethodSetup = ET.Element("MethodSetup", Method="BDT::BDT")
#<Variables>
Variables = ET.SubElement(MethodSetup, "Variables", NVar=str(len(var_list)))
for ind, val in enumerate(var_list):
name = val[0]
var_type = val[1]
Variable = ET.SubElement(Variables, "Variable", VarIndex=str(ind), Type=val[1],
Expression=name, Label=name, Title=name, Unit="", Internal=name,
Min="0.0e+00", Max="0.0e+00")
# <GeneralInfo>
GeneralInfo = ET.SubElement(MethodSetup, "GeneralInfo")
Info_Creator = ET.SubElement(GeneralInfo, "Info", name="Creator", value="Koza4ok (skTMVA)")
Info_AnalysisType = ET.SubElement(GeneralInfo, "Info", name="AnalysisType", value="Classification")
# <Options>
Options = ET.SubElement(MethodSetup, "Options")
Option_NodePurityLimit = ET.SubElement(Options, "Option", name="NodePurityLimit", modified="No").text = str(NodePurityLimit)
Option_BoostType = ET.SubElement(Options, "Option", name="BoostType", modified="Yes").text = "Grad"
# <Weights>
Weights = ET.SubElement(MethodSetup, "Weights", NTrees=str(NTrees), AnalysisType="1")
# We support only binary classification
# from http://scikit-learn.org/stable/modules/generated/sklearn.ensemble.GradientBoostingClassifier.html
# estimators_ : ndarray of DecisionTreeRegressor, shape = [n_estimators, loss_.K]
# where loss_.K is 1 for binary classification, otherwise n_classes.
for idx, dt in enumerate(clf.estimators_[:, 0]):
# <BinaryTree type="DecisionTree" boostWeight="9.2106320437773737e-01" itree="0">
BinaryTree = ET.SubElement(Weights, "BinaryTree", type="DecisionTree", boostWeight="1.0e+00", itree=str(idx))
build_xml_tree__Grad(dt, 0, "s", -1, BinaryTree)
# Create XML-tree structure and save it to file
tree = ET.ElementTree(MethodSetup)
tree.write(tmva_outfile_xml)
var_type = val[1]
Variable = ET.SubElement(Variables, "Variable", VarIndex=str(ind), Type=val[1],
Expression=name, Label=name, Title=name, Unit="", Internal=name,
Min="0.0e+00", Max="0.0e+00")
# <GeneralInfo>
GeneralInfo = ET.SubElement(MethodSetup, "GeneralInfo")
Info_Creator = ET.SubElement(GeneralInfo, "Info", name="Creator", value="Koza4ok (skTMVA)")
Info_AnalysisType = ET.SubElement(GeneralInfo, "Info", name="AnalysisType", value="Classification")
# <Options>
Options = ET.SubElement(MethodSetup, "Options")
Option_NodePurityLimit = ET.SubElement(Options, "Option", name="NodePurityLimit", modified="No").text = str(NodePurityLimit)
# <Weights>
Weights = ET.SubElement(MethodSetup, "Weights", NTrees=str(NTrees), AnalysisType="0")
for idx, dt in enumerate(clf.estimators_):
tree_weight = clf.estimator_weights_[idx]
# <BinaryTree type="DecisionTree" boostWeight="9.2106320437773737e-01" itree="0">
BinaryTree = ET.SubElement(Weights, "BinaryTree", type="DecisionTree", boostWeight=str(tree_weight), itree=str(idx))
build_xml_tree(dt, 0, "s", -1, BinaryTree)
tree = ET.ElementTree(MethodSetup)
tree.write("SKLearn_BDT_muons.weights.xml")
# Save to file fpr, tpr
#np.savez('output_fullsim_v3_electrons_fpr_tpr_10per.npz',
# fpr=fpr, tpr=tpr)
def convert_bdt__XGBClassifier(sklearn_bdt_clf, input_var_list,
tmva_outfile_xml):
# classificator
clf = sklearn_bdt_clf
#clf = pickle.dumps(sklearn_bdt_clf)
if clf.n_classes_ != 2:
sys.exit(
"Error: Number of classes in sklearn classifier is not equal 2.")
# Order of variables must be _exactly_ as in the training numpy array
# E.g.
# var_list = [
# ('m_el_pt', 'F'),
# ('m_el_eta', 'F'),
# ('m_el_sigd0PV', 'F'),
# ('m_el_z0SinTheta', 'F'),
# ('m_el_etcone20Dpt', 'F'),
# ('m_el_ptcone20Dpt', 'F')
# ]
var_list = input_var_list
# Run-time parameters
NTrees = clf.n_estimators
#<MethodSetup Method="BDT::BDT">
# <GeneralInfo>
# <Info name="Creator" value="musthero"/>
# <Info name="AnalysisType" value="Classification"/>
# <Options>
# <Option name="NodePurityLimit" modified="No">5.000000e-01</Option>
#<Weights NTrees="2" AnalysisType="0">
#-------------
# <MethodSetup>
MethodSetup = ET.Element("MethodSetup", Method="BDT::BDT")
#<Variables>
Variables = ET.SubElement(MethodSetup,
"Variables",
NVar=str(len(var_list)))
for ind, val in enumerate(var_list):
name = val[0]
var_type = val[1]
Variable = ET.SubElement(Variables,
"Variable",
VarIndex=str(ind),
Type=val[1],
Expression=name,
Label=name,
Title=name,
Unit="",
Internal=name,
Min="0.0e+00",
Max="0.0e+00")
# <GeneralInfo>
GeneralInfo = ET.SubElement(MethodSetup, "GeneralInfo")
Info_Creator = ET.SubElement(GeneralInfo,
"Info",
name="Creator",
value="Koza4ok (skTMVA)")
Info_AnalysisType = ET.SubElement(GeneralInfo,
"Info",
name="AnalysisType",
value="Classification")
# <Options>
Options = ET.SubElement(MethodSetup, "Options")
Option_NodePurityLimit = ET.SubElement(
Options, "Option", name="NodePurityLimit",
modified="No").text = str(NodePurityLimit)
Option_BoostType = ET.SubElement(Options,
"Option",
name="BoostType",
modified="Yes").text = "AdaBoost"
# <Weights>
Weights = ET.SubElement(MethodSetup,
"Weights",
NTrees=str(NTrees),
AnalysisType="0")
#for idx, dt in enumerate(clf.estimators_):
for idx in range(NTrees):
tree_weight = clf.evals_result()['validation_0']['logloss'][
idx] #clf.estimator_weights_[itree]
# <BinaryTree type="DecisionTree" boostWeight="9.2106320437773737e-01" itree="0">
BinaryTree = ET.SubElement(Weights,
"BinaryTree",
type="DecisionTree",
boostWeight=str(tree_weight),
itree=str(idx))
#BinaryTree = ET.SubElement(Weights, "BinaryTree", type="DecisionTree", boostWeight="1.0e+00", itree=str(idx))
build_xml_tree__XGBClassifier(sklearn_bdt_clf, 0, "s", -1, BinaryTree)
# Create XML-tree structure and save it to file
tree = ET.ElementTree(MethodSetup)
tree.write(tmva_outfile_xml)
def convert_bdt__Grad(sklearn_bdt_clf, input_var_list, tmva_outfile_xml):
# classificator
clf = sklearn_bdt_clf
if clf.loss_.K != 1:
sys.exit(
"Error: Only binary classification is supported for regression trees."
)
if clf.n_classes_ != 2:
sys.exit(
"Error: Number of classes in sklearn classifier is not equal 2.")
# Order of variables must be _exactly_ as in the training numpy array
# E.g.
# var_list = [
# ('m_el_pt', 'F'),
# ('m_el_eta', 'F'),
# ('m_el_sigd0PV', 'F'),
# ('m_el_z0SinTheta', 'F'),
# ('m_el_etcone20Dpt', 'F'),
# ('m_el_ptcone20Dpt', 'F')
# ]
var_list = input_var_list
# Run-time parameters
NTrees = clf.n_estimators
#<MethodSetup Method="BDT::BDT">
# <GeneralInfo>
# <Info name="Creator" value="musthero"/>
# <Info name="AnalysisType" value="Classification"/>
# <Options>
# <Option name="NodePurityLimit" modified="No">5.000000e-01</Option>
#<Weights NTrees="2" AnalysisType="0">
#-------------
# <MethodSetup>
MethodSetup = ET.Element("MethodSetup", Method="BDT::BDT")
#<Variables>
Variables = ET.SubElement(MethodSetup,
"Variables",
NVar=str(len(var_list)))
for ind, val in enumerate(var_list):
name = val[0]
var_type = val[1]
Variable = ET.SubElement(Variables,
"Variable",
VarIndex=str(ind),
Type=val[1],
Expression=name,
Label=name,
Title=name,
Unit="",
Internal=name,
Min="0.0e+00",
Max="0.0e+00")
# <GeneralInfo>
GeneralInfo = ET.SubElement(MethodSetup, "GeneralInfo")
Info_Creator = ET.SubElement(GeneralInfo,
"Info",
name="Creator",
value="Koza4ok (skTMVA)")
Info_AnalysisType = ET.SubElement(GeneralInfo,
"Info",
name="AnalysisType",
value="Classification")
# <Options>
Options = ET.SubElement(MethodSetup, "Options")
Option_NodePurityLimit = ET.SubElement(
Options, "Option", name="NodePurityLimit",
modified="No").text = str(NodePurityLimit)
Option_BoostType = ET.SubElement(Options,
"Option",
name="BoostType",
modified="Yes").text = "Grad"
# <Weights>
Weights = ET.SubElement(MethodSetup,
"Weights",
NTrees=str(NTrees),
AnalysisType="1")
# We support only binary classification
# from http://scikit-learn.org/stable/modules/generated/sklearn.ensemble.GradientBoostingClassifier.html
# estimators_ : ndarray of DecisionTreeRegressor, shape = [n_estimators, loss_.K]
# where loss_.K is 1 for binary classification, otherwise n_classes.
for idx, dt in enumerate(clf.estimators_[:, 0]):
# <BinaryTree type="DecisionTree" boostWeight="9.2106320437773737e-01" itree="0">
BinaryTree = ET.SubElement(Weights,
"BinaryTree",
type="DecisionTree",
boostWeight="1.0e+00",
itree=str(idx))
build_xml_tree__Grad(dt, 0, "s", -1, BinaryTree)
# Create XML-tree structure and save it to file
tree = ET.ElementTree(MethodSetup)
tree.write(tmva_outfile_xml)
build_xml_tree(children_right[node_id], "r", node_depth, node_elementTree)
else:
# leaf node
node_depth = parent_depth + 1
# node parameters
pos = "s" if node_id == 0 else node_pos
depth = node_depth
IVar = -1
global NodePurityLimit
sig = value[node_id][0][1]
bkg = value[node_id][0][0]
total = float(sig + bkg)
purity = float(sig)/total
nType = 1 if purity >= NodePurityLimit else -1
node_elementTree = ET.SubElement(parent_elementTree, "Node", pos=pos, depth=str(depth), NCoef="0", IVar=str(IVar),
Cut="0.0e+00", cType="1", res="0.0e+01", rms="0.0e+00", purity=str(purity), nType=str(nType))
BinaryTree = ET.Element("BinaryTree")
build_xml_tree(0, "s", -1, BinaryTree)
tree = ET.ElementTree(BinaryTree)
tree.write("filename.xml")
# Save to file fpr, tpr
#np.savez('output_fullsim_v3_electrons_fpr_tpr_10per.npz',
# fpr=fpr, tpr=tpr)
