Logo ROOT   6.16/01
Reference Guide
MethodPyGTB.h
Go to the documentation of this file.
1// @(#)root/tmva/pymva $Id$
2// Authors: Omar Zapata, Lorenzo Moneta, Sergei Gleyzer 2015
3
4/**********************************************************************************
5 * Project: TMVA - a Root-integrated toolkit for multivariate data analysis *
6 * Package: TMVA *
7 * Class : MethodPyGTB *
8 * Web : http://oproject.org *
9 * *
10 * Description: *
11 * scikit-learn Package GradientBoostingClassifier method based on python *
12 * *
13 **********************************************************************************/
14
15#ifndef ROOT_TMVA_MethodPyGTB
16#define ROOT_TMVA_MethodPyGTB
17
18//////////////////////////////////////////////////////////////////////////
19// //
20// MethodPyGTB //
21// //
22//////////////////////////////////////////////////////////////////////////
23
24#include "TMVA/PyMethodBase.h"
25
26namespace TMVA {
27
28 class Factory;
29 class Reader;
30 class DataSetManager;
31 class Types;
32 class MethodPyGTB : public PyMethodBase {
33
34 public :
35 MethodPyGTB(const TString &jobName,
36 const TString &methodTitle,
37 DataSetInfo &theData,
38 const TString &theOption = "");
40 const TString &theWeightFile);
41 ~MethodPyGTB(void);
42
43 void Train();
44 void Init();
45 void DeclareOptions();
46 void ProcessOptions();
47
48 const Ranking *CreateRanking();
49
50 Bool_t HasAnalysisType(Types::EAnalysisType type, UInt_t numberClasses, UInt_t numberTargets);
51
52 virtual void TestClassification();
53
54 Double_t GetMvaValue(Double_t *errLower = 0, Double_t *errUpper = 0);
55 std::vector<Double_t> GetMvaValues(Long64_t firstEvt = 0, Long64_t lastEvt = -1, Bool_t logProgress = false);
56 std::vector<Float_t>& GetMulticlassValues();
57
58 virtual void ReadModelFromFile();
59
61 // the actual "weights"
62 virtual void AddWeightsXMLTo(void * /* parent */ ) const {} // = 0;
63 virtual void ReadWeightsFromXML(void * /*wghtnode*/) {} // = 0;
64 virtual void ReadWeightsFromStream(std::istream &) {} //= 0; backward compatibility
65
66 private :
68 friend class Factory;
69 friend class Reader;
70
71 protected:
72 std::vector<Double_t> mvaValues;
73 std::vector<Float_t> classValues;
74
75 UInt_t fNvars; // number of variables
76 UInt_t fNoutputs; // number of outputs
77 TString fFilenameClassifier; // Path to serialized classifier (default in `weights` folder)
78
79 //GTB options
80
82 TString fLoss; // {'deviance', 'exponential'}, optional (default='deviance')
83 //loss function to be optimized. 'deviance' refers to
84 //deviance (= logistic regression) for classification
85 //with probabilistic outputs. For loss 'exponential' gradient
86 //boosting recovers the AdaBoost algorithm.
87
89 Double_t fLearningRate; //float, optional (default=0.1)
90 //learning rate shrinks the contribution of each tree by `learning_rate`.
91 //There is a trade-off between learning_rate and n_estimators.
92
94 Int_t fNestimators; //integer, optional (default=10)
95 //The number of trees in the forest.
96
98 Double_t fSubsample; //float, optional (default=1.0)
99 //The fraction of samples to be used for fitting the individual base
100 //learners. If smaller than 1.0 this results in Stochastic Gradient
101 //Boosting. `subsample` interacts with the parameter `n_estimators`.
102 //Choosing `subsample < 1.0` leads to a reduction of variance
103 //and an increase in bias.
104
106 Int_t fMinSamplesSplit; // integer, optional (default=2)
107 //The minimum number of samples required to split an internal node.
108
110 Int_t fMinSamplesLeaf; //integer, optional (default=1)
111 //The minimum number of samples required to be at a leaf node.
112
114 Double_t fMinWeightFractionLeaf; //float, optional (default=0.)
115 //The minimum weighted fraction of the input samples required to be at a leaf node.
116
118 Int_t fMaxDepth; //integer, optional (default=3)
119 //maximum depth of the individual regression estimators. The maximum
120 //depth limits the number of nodes in the tree. Tune this parameter
121 //for best performance; the best value depends on the interaction
122 //of the input variables.
123 //Ignored if ``max_leaf_nodes`` is not None.
124
126 TString fInit; //BaseEstimator, None, optional (default=None)
127 //An estimator object that is used to compute the initial
128 //predictions. ``init`` has to provide ``fit`` and ``predict``.
129 //If None it uses ``loss.init_estimator``.
130
132 TString fRandomState; //int, RandomState instance or None, optional (default=None)
133 //If int, random_state is the seed used by the random number generator;
134 //If RandomState instance, random_state is the random number generator;
135 //If None, the random number generator is the RandomState instance used
136 //by `np.random`.
137
139 TString fMaxFeatures; //int, float, string or None, optional (default="auto")
140 //The number of features to consider when looking for the best split:
141 //- If int, then consider `max_features` features at each split.
142 //- If float, then `max_features` is a percentage and
143 //`int(max_features * n_features)` features are considered at each split.
144 //- If "auto", then `max_features=sqrt(n_features)`.
145 //- If "sqrt", then `max_features=sqrt(n_features)`.
146 //- If "log2", then `max_features=log2(n_features)`.
147 //- If None, then `max_features=n_features`.
148 // Note: the search for a split does not stop until at least one
149 // valid partition of the node samples is found, even if it requires to
150 // effectively inspect more than ``max_features`` features.
151 // Note: this parameter is tree-specific.
152
154 Int_t fVerbose; //Controls the verbosity of the tree building process.
155
157 TString fMaxLeafNodes; //int or None, optional (default=None)
158 //Grow trees with ``max_leaf_nodes`` in best-first fashion.
159 //Best nodes are defined as relative reduction in impurity.
160 //If None then unlimited number of leaf nodes.
161 //If not None then ``max_depth`` will be ignored.
162
164 Bool_t fWarmStart; //bool, optional (default=False)
165 //When set to ``True``, reuse the solution of the previous call to fit
166 //and add more estimators to the ensemble, otherwise, just fit a whole
167 //new forest.
168
169 // get help message text
170 void GetHelpMessage() const;
171
173 };
174
175} // namespace TMVA
176
177#endif // ROOT_TMVA_PyMethodGTB
int Int_t
Definition: RtypesCore.h:41
unsigned int UInt_t
Definition: RtypesCore.h:42
bool Bool_t
Definition: RtypesCore.h:59
double Double_t
Definition: RtypesCore.h:55
long long Long64_t
Definition: RtypesCore.h:69
#define ClassDef(name, id)
Definition: Rtypes.h:324
int type
Definition: TGX11.cxx:120
_object PyObject
Definition: TPyArg.h:20
Class that contains all the data information.
Definition: DataSetInfo.h:60
Class that contains all the data information.
This is the main MVA steering class.
Definition: Factory.h:81
virtual void ReadWeightsFromStream(std::istream &)=0
Double_t fSubsample
Definition: MethodPyGTB.h:98
PyObject * pInit
Definition: MethodPyGTB.h:125
DataSetManager * fDataSetManager
Definition: MethodPyGTB.h:67
PyObject * pMinSamplesLeaf
Definition: MethodPyGTB.h:109
Double_t fMinWeightFractionLeaf
Definition: MethodPyGTB.h:114
std::vector< Double_t > mvaValues
Definition: MethodPyGTB.h:72
PyObject * pMaxFeatures
Definition: MethodPyGTB.h:138
std::vector< Double_t > GetMvaValues(Long64_t firstEvt=0, Long64_t lastEvt=-1, Bool_t logProgress=false)
get all the MVA values for the events of the current Data type
PyObject * pMaxDepth
Definition: MethodPyGTB.h:117
PyObject * pMaxLeafNodes
Definition: MethodPyGTB.h:156
std::vector< Float_t > classValues
Definition: MethodPyGTB.h:73
Bool_t HasAnalysisType(Types::EAnalysisType type, UInt_t numberClasses, UInt_t numberTargets)
Double_t fLearningRate
Definition: MethodPyGTB.h:89
void GetHelpMessage() const
TString fMaxFeatures
Definition: MethodPyGTB.h:139
MethodPyGTB(const TString &jobName, const TString &methodTitle, DataSetInfo &theData, const TString &theOption="")
Definition: MethodPyGTB.cxx:58
TString fRandomState
Definition: MethodPyGTB.h:132
PyObject * pLearningRate
Definition: MethodPyGTB.h:88
const Ranking * CreateRanking()
virtual void TestClassification()
initialization
std::vector< Float_t > & GetMulticlassValues()
virtual void ReadWeightsFromStream(std::istream &)
Definition: MethodPyGTB.h:64
virtual void ReadModelFromFile()
PyObject * pNestimators
Definition: MethodPyGTB.h:93
TString fMaxLeafNodes
Definition: MethodPyGTB.h:157
PyObject * pVerbose
Definition: MethodPyGTB.h:153
virtual void AddWeightsXMLTo(void *) const
Definition: MethodPyGTB.h:62
virtual void ReadWeightsFromXML(void *)
Definition: MethodPyGTB.h:63
TString fFilenameClassifier
Definition: MethodPyGTB.h:77
PyObject * pMinSamplesSplit
Definition: MethodPyGTB.h:105
PyObject * pLoss
Definition: MethodPyGTB.h:81
PyObject * pSubsample
Definition: MethodPyGTB.h:97
PyObject * pRandomState
Definition: MethodPyGTB.h:131
PyObject * pWarmStart
Definition: MethodPyGTB.h:163
PyObject * pMinWeightFractionLeaf
Definition: MethodPyGTB.h:113
Double_t GetMvaValue(Double_t *errLower=0, Double_t *errUpper=0)
Ranking for variables in method (implementation)
Definition: Ranking.h:48
The Reader class serves to use the MVAs in a specific analysis context.
Definition: Reader.h:63
EAnalysisType
Definition: Types.h:127
Basic string class.
Definition: TString.h:131
Abstract ClassifierFactory template that handles arbitrary types.