Belle II Software light-2406-ragdoll
tmva_nn.py
1#!/usr/bin/env python3
2
3
10
11import basf2_mva
12import basf2_mva_util
13import time
14
15if __name__ == "__main__":
16 from basf2 import conditions, find_file
17 # NOTE: do not use testing payloads in production! Any results obtained like this WILL NOT BE PUBLISHED
18 conditions.testing_payloads = [
19 'localdb/database.txt'
20 ]
21
22 train_file = find_file("mva/train_D0toKpipi.root", "examples")
23 test_file = find_file("mva/test_D0toKpipi.root", "examples")
24
25 training_data = basf2_mva.vector(train_file)
26 testing_data = basf2_mva.vector(test_file)
27
28 variables = ['M', 'p', 'pt', 'pz',
29 'daughter(0, p)', 'daughter(0, pz)', 'daughter(0, pt)',
30 'daughter(1, p)', 'daughter(1, pz)', 'daughter(1, pt)',
31 'daughter(2, p)', 'daughter(2, pz)', 'daughter(2, pt)',
32 'chiProb', 'dr', 'dz',
33 'daughter(0, dr)', 'daughter(1, dr)',
34 'daughter(0, dz)', 'daughter(1, dz)',
35 'daughter(0, chiProb)', 'daughter(1, chiProb)', 'daughter(2, chiProb)',
36 'daughter(0, kaonID)', 'daughter(0, pionID)',
37 'daughterInvM(0, 1)', 'daughterInvM(0, 2)', 'daughterInvM(1, 2)']
38
39 # Train a MVA method and directly upload it to the database
40 general_options = basf2_mva.GeneralOptions()
41 general_options.m_datafiles = training_data
42 general_options.m_treename = "tree"
43 general_options.m_identifier = "TMVA"
44 general_options.m_variables = basf2_mva.vector(*variables)
45 general_options.m_target_variable = "isSignal"
46
47 tmva_nn_options = basf2_mva.TMVAOptionsClassification()
48 tmva_nn_options.m_type = "MLP"
49 tmva_nn_options.m_method = "MLP"
50 tmva_nn_options.m_config = ("H:!V:CreateMVAPdfs:VarTransform=N:NCycles=10:HiddenLayers=N+1:TrainingMethod=BFGS")
51
52 training_start = time.time()
53 basf2_mva.teacher(general_options, tmva_nn_options)
54 training_stop = time.time()
55
56 training_time = training_stop - training_start
57 method = basf2_mva_util.Method(general_options.m_identifier)
58
59 inference_start = time.time()
60 p, t = method.apply_expert(testing_data, general_options.m_treename)
61 inference_stop = time.time()
62 inference_time = inference_stop - inference_start
63
65 print("TMVA", training_time, inference_time, auc)
def calculate_auc_efficiency_vs_background_retention(p, t, w=None)