Belle II Software  release-08-01-10
FastBDTClassifierTrainingModule.cc
1 /**************************************************************************
2  * basf2 (Belle II Analysis Software Framework) *
3  * Author: The Belle II Collaboration *
4  * *
5  * See git log for contributors and copyright holders. *
6  * This file is licensed under LGPL-3.0, see LICENSE.md. *
7  **************************************************************************/
8 
9 #include <tracking/modules/vxdtfRedesign/FastBDTClassifierTrainingModule.h>
10 #include <tracking/trackFindingVXD/filterTools/FBDTClassifier.h>
11 #include <tracking/trackFindingVXD/segmentNetwork/DirectedNodeNetwork.h>
12 #include <tracking/trackFindingVXD/segmentNetwork/TrackNode.h>
13 
14 #include <tracking/spacePointCreation/PurityCalculatorTools.h>
15 #include <tracking/spacePointCreation/MCVXDPurityInfo.h>
16 
17 #include <framework/logging/Logger.h>
18 
19 #include <fstream>
20 #include <sstream>
21 
22 using namespace Belle2;
23 
24 REG_MODULE(FastBDTClassifierTraining);
25 
27 {
28  setDescription("TODO");
29 
30  addParam("outputFileName",
32  "Output file name to which the trained FBDTClassifier will be stored",
33  std::string("FBDTClassifier.dat"));
34 
35  addParam("networkInputName", m_PARAMnetworkInputName,
36  "Name of the StoreObjPtr where the network container used in this module is stored", std::string(""));
37 
38  addParam("train", m_PARAMdoTrain, "Set if the module should train a classifier after collecting or not", true);
39  addParam("nTrees", m_PARAMnTrees, "Number of Trees used in the FastBDT", 100);
40  addParam("treeDepth", m_PARAMtreeDepth, "Tree depth of the trees used in the FastBDT", 3);
41  addParam("shrinkage", m_PARAMshrinkage, "Shrinkage parameter used in the FastBDT", 0.15);
42  addParam("randRatio", m_PARAMrandRatio, "ratio of randomly chosen samples for training of one tree", 0.5);
43  addParam("storeSamples", m_PARAMstoreSamples, "store the collected samples into a file", false);
44  addParam("samplesFileName", m_PARAMsamplesFileName, "the file name into which/from whicht the collected samples are stored/read",
45  std::string("FBDTClassifier_samples.dat"));
46  addParam("useSamples", m_PARAMuseSamples,
47  "use samples for training that have been collected previously and bypass the collection of samples", false);
48 }
49 
51 {
52 
54  B2ERROR("storeSamples and useSamples are both set to true. However, only one option can be set at a time");
55  }
56 
57  if (m_PARAMnTrees < 1) {
58  B2WARNING("nTrees was set to " << m_PARAMnTrees << ". Has to be at least 1. Setting to 1.");
59  m_PARAMnTrees = 1;
60  }
61 
62  if (m_PARAMtreeDepth < 0) {
63  B2WARNING("Trees have to be at least a stump, but treeDepth was set to " << m_PARAMtreeDepth << ". Setting to 3 (default).");
64  m_PARAMtreeDepth = 3;
65  }
66 
67  if (m_PARAMshrinkage < 0 || m_PARAMshrinkage > 1) { // TODO: check this
68  B2WARNING("shrinkage has to be in [0,1] but was " << m_PARAMrandRatio << ". Setting to 0.15 (default).");
69  m_PARAMshrinkage = .15;
70  }
71 
72  if (m_PARAMrandRatio < 0 || m_PARAMrandRatio > 1) {
73  B2WARNING("randRatio has to be in [0,1] but was " << m_PARAMrandRatio << ". Setting to 0.5 (default).");
74  m_PARAMrandRatio = 0.5;
75  }
76 
77  if (m_PARAMuseSamples) {
78  std::ifstream sampFile(m_PARAMsamplesFileName);
79  if (!sampFile.is_open() || !sampFile.good()) {
80  B2ERROR("Was not able to open the samples file: " << m_PARAMsamplesFileName);
81  }
82 
84  } else { // only if no samples are provided the collection from the DNN is necessary
86  }
87 }
88 
90 {
91  if (m_PARAMuseSamples) return; // don't collect anything during event if samples are provided
92 
94 
95  // B2DEBUG(20, "size of hitNetwork " << hitNetwork.getNodes().size());
96 
97  size_t samplePriorEvent = m_samples.size();
98 
99  // XXXHit is of type DirectedNode<TrackNode, VoidMetaInfo>
100  for (const auto& outerHit : hitNetwork.getNodes()) { // loop over all outer nodes
101  for (const auto& centerHit : outerHit->getInnerNodes()) { // loop over all center nodes attached to outer node
102  for (const auto& innerHit : centerHit->getInnerNodes()) { // loop over all inner nodes attached to center node
103  m_samples.push_back(makeTrainSample(outerHit->getEntry().m_spacePoint,
104  centerHit->getEntry().m_spacePoint,
105  innerHit->getEntry().m_spacePoint));
106  } // inner node loop
107  } // center node loop
108  } // outer node loop
109 
110  B2DEBUG(21, "collected " << m_samples.size() - samplePriorEvent << " training samples in this event");
111 
112 }
113 
115 {
116  if (m_PARAMstoreSamples) {
117  B2DEBUG(20, "Storing the collected samples to file: " << m_PARAMsamplesFileName);
118  std::ofstream sampStream(m_PARAMsamplesFileName);
119  sampStream.precision(16); // increase precision for sample writeout
120  writeSamplesToStream(sampStream, m_samples);
121  sampStream.close();
122  }
123  if (m_PARAMdoTrain) {
124  FBDTClassifier<9> classifier{};
125  B2DEBUG(20, "Training a FBDTClassifier with " << m_samples.size() << " input samples. Training Parameters: \n" <<
126  "nTrees: " << m_PARAMnTrees << "\n" <<
127  "treeDetph: " << m_PARAMtreeDepth << "\n" <<
128  "shrinkage: " << m_PARAMshrinkage << "\n" <<
129  "randRatio: " << m_PARAMrandRatio << "\n");
131 
132  std::ofstream ofs(m_PARAMfbdtOutFileName);
133  classifier.writeToStream(ofs);
134  ofs.close();
135  }
136 }
137 
140  const Belle2::SpacePoint* inner)
141 {
142  std::vector<MCVXDPurityInfo> purityInfos = createPurityInfosVec({outer, center, inner});
143  auto mcId = purityInfos[0].getPurity(); // there is at least one entry in this vector!
144  bool signal = (mcId.first >= 0 && mcId.second == 1); // only assign true for actual MCParticle and purity 1
145 
146  std::array<double, 9> coords {{
147  inner->X(), inner->Y(), inner->Z(),
148  center->X(), center->Y(), center->Z(),
149  outer->X(), outer->Y(), outer->Z()
150  }};
151 
152  TrainSample sample(coords, signal);
153 
154  if (LogSystem::Instance().isLevelEnabled(LogConfig::c_Debug, 499, PACKAGENAME())) {
155  std::stringstream coordOutput;
156  for (double d : sample.hits) coordOutput << d << " ";
157 
158  B2DEBUG(29, "Created TrainingsSample with coordinates: ( " << coordOutput.str() << " ) " << sample.signal);
159  }
160 
161  return sample;
162 }
163 
164 // void FastBDTClassifierTrainingModule::readSamplesFromStream(std::istream& is)
165 // {
166 // std::string line;
167 // while(!is.eof()) {
168 // getline(is, line);
169 // if(line.empty()) break;
170 // stringstream ss(line);
171 // std::array<double, 9> coords;
172 // for(double& c : coords) ss >> c;
173 // bool sig; ss >> sig;
174 
175 // m_samples.push_back(FBDTTrainSample<9>(coords, sig));
176 // }
177 
178 // B2INFO("Read in " << m_samples.size() << " training samples.");
179 // }
180 
181 // void FastBDTClassifierTrainingModule::writeSamplesToStream(std::ostream& os) const
182 // {
183 // for (const auto& event : m_samples) {
184 // for (const auto& val : event.hits) {
185 // os << val << " ";
186 // }
187 // os << event.signal << std::endl;
188 // }
189 // B2INFO("Wrote out " << m_samples.size() << " training samples.");
190 // }
DirectedNodeNetwork< Belle2::TrackNode, Belle2::VoidMetaInfo > & accessHitNetwork()
Returns reference to the HitNetwork stored in this container, intended for read and write access.
Network of directed nodes of the type EntryType.
std::vector< Node * > & getNodes()
Returns all nodes of the network.
double m_PARAMrandRatio
ratio of samples to be used for training one tree in the FastBDT.
std::string m_PARAMsamplesFileName
filename to be used to store / read collect samples
void initialize() override
initialize the module
Belle2::StoreObjPtr< Belle2::DirectedNodeNetworkContainer > m_network
StoreObjPtr to access the DNNs that are used in this module.
void event() override
collect all possible combinations and store them
std::vector< TrainSample > m_samples
vector in which all samples are collected on the fly in event.
const TrainSample makeTrainSample(const Belle2::SpacePoint *outerHit, const Belle2::SpacePoint *centerHit, const Belle2::SpacePoint *innerHit)
create a trainings sample from the three hit combination
void terminate() override
take the collected data and train a FBDTClassifier and store it in the given output file
bool m_PARAMdoTrain
actually train a classifier or only do collection
std::string m_PARAMfbdtOutFileName
output file name into which the FBDTClassifier is stored.
double m_PARAMshrinkage
shrinkage parameter of FastBDT.
bool m_PARAMstoreSamples
store the collected samples into a file
std::string m_PARAMnetworkInputName
name of the StoreObjPtr in which the network container is stored which contains the network that is u...
bool m_PARAMuseSamples
use pre-collected samples for training and bypass the collection step
int m_PARAMnTrees
number of trees in the FastBDT.
@ c_Debug
Debug: for code development.
Definition: LogConfig.h:26
static LogSystem & Instance()
Static method to get a reference to the LogSystem instance.
Definition: LogSystem.cc:31
Base class for Modules.
Definition: Module.h:72
void setDescription(const std::string &description)
Sets the description of the module.
Definition: Module.cc:214
SpacePoint typically is build from 1 PXDCluster or 1-2 SVDClusters.
Definition: SpacePoint.h:42
double Z() const
return the z-value of the global position of the SpacePoint
Definition: SpacePoint.h:129
double X() const
return the x-value of the global position of the SpacePoint
Definition: SpacePoint.h:123
double Y() const
return the y-value of the global position of the SpacePoint
Definition: SpacePoint.h:126
bool isRequired(const std::string &name="")
Ensure this array/object has been registered previously.
REG_MODULE(arichBtest)
Register the Module.
void addParam(const std::string &name, T &paramVariable, const std::string &description, const T &defaultValue)
Adds a new parameter to the module.
Definition: Module.h:560
static void readSamplesFromStream(std::istream &is, std::vector< FBDTTrainSample< Ndims > > &samples)
read samples from stream and append them to samples
static std::vector< Belle2::MCVXDPurityInfo > createPurityInfosVec(const std::vector< const Belle2::SpacePoint * > &spacePoints)
create a vector of MCVXDPurityInfos objects for a std::vector<Belle2::SpacePoints>.
static void writeSamplesToStream(std::ostream &os, const std::vector< FBDTTrainSample< Ndims > > &samples)
write all samples to stream
Abstract base class for different kinds of events.
bundle together the classifier input and the target value into one struct for easier passing around.