Darwin  1.10(beta)
 All Classes Namespaces Files Functions Variables Typedefs Enumerations Enumerator Friends Groups Pages
Public Member Functions | Static Public Attributes | Protected Attributes | List of all members
drwnCompositeClassifier Class Reference

Implements a multi-class classifier by combining binary classifiers. More...

Inheritance diagram for drwnCompositeClassifier:
drwnClassifier drwnStdObjIface drwnProperties drwnWriteable drwnCloneable drwnTypeable

Public Member Functions

 drwnCompositeClassifier ()
 default constructor
 
 drwnCompositeClassifier (unsigned n, unsigned k=2)
 construct a classifer with n features and k classes
 
 drwnCompositeClassifier (const drwnCompositeClassifier &c)
 copy constructor
 
virtual const char * type () const
 returns object type as a string (e.g., Foo::type() { return "Foo"; })
 
virtual drwnCompositeClassifierclone () const
 returns a copy of the class usually implemented as virtual Foo* clone() { return new Foo(*this); }
 
virtual void initialize (unsigned n, unsigned k=2)
 initialize the classifier object for n features and k classes
 
virtual bool save (drwnXMLNode &node) const
 write object to XML node (see also write)
 
virtual bool load (drwnXMLNode &node)
 read object from XML node (see also read)
 
virtual double train (const drwnClassifierDataset &dataset)
 train the parameters of the classifier from a drwnClassifierDataset object
 
virtual void getClassScores (const vector< double > &features, vector< double > &outputScores) const
 compute the unnormalized log-probability for a single feature vector
 
- Public Member Functions inherited from drwnClassifier
 drwnClassifier ()
 default constructor
 
 drwnClassifier (unsigned n, unsigned k=2)
 construct a classifer with n features and k classes
 
 drwnClassifier (const drwnClassifier &c)
 copy constructor
 
int numFeatures () const
 returns the number of features expected by the classifier object
 
int numClasses () const
 returns the number of classes predicted by the classifier object
 
virtual bool valid () const
 returns true if the classifier is valid (has been initialized and trained)
 
virtual double train (const vector< vector< double > > &features, const vector< int > &targets)
 train the parameters of the classifier from a set of features and corresponding labels
 
virtual double train (const vector< vector< double > > &features, const vector< int > &targets, const vector< double > &weights)
 train the parameters of the classifier from a weighted set of features and corresponding labels
 
virtual double train (const char *filename)
 train the parameters of the classifier from data stored in filename
 
virtual void getClassScores (const vector< vector< double > > &features, vector< vector< double > > &outputScores) const
 compute the unnormalized log-probability for a set of feature vectors
 
virtual void getClassMarginals (const vector< double > &features, vector< double > &outputMarginals) const
 compute the class marginal probabilities for a single feature vector
 
virtual void getClassMarginals (const vector< vector< double > > &features, vector< vector< double > > &outputMarginals) const
 compute the class marginal probabilities for a set of feature vectors
 
virtual int getClassification (const vector< double > &features) const
 return the most likely class label for a single feature vector
 
virtual void getClassifications (const vector< vector< double > > &features, vector< int > &outputLabels) const
 compute the most likely class labels for a set of feature vector
 
- Public Member Functions inherited from drwnWriteable
bool write (const char *filename) const
 write object to file (calls save)
 
bool read (const char *filename)
 read object from file (calls load)
 
void dump () const
 print object's current state to standard output (for debugging)
 
- Public Member Functions inherited from drwnProperties
unsigned numProperties () const
 
bool hasProperty (const string &name) const
 
bool hasProperty (const char *name) const
 
unsigned findProperty (const string &name) const
 
unsigned findProperty (const char *name) const
 
void setProperty (unsigned indx, bool value)
 
void setProperty (unsigned indx, int value)
 
void setProperty (unsigned indx, double value)
 
void setProperty (unsigned indx, const string &value)
 
void setProperty (unsigned indx, const char *value)
 
void setProperty (unsigned indx, const Eigen::VectorXd &value)
 
void setProperty (unsigned indx, const Eigen::MatrixXd &value)
 
void setProperty (const char *name, bool value)
 
void setProperty (const char *name, int value)
 
void setProperty (const char *name, double value)
 
void setProperty (const char *name, const string &value)
 
void setProperty (const char *name, const char *value)
 
void setProperty (const char *name, const Eigen::VectorXd &value)
 
void setProperty (const char *name, const Eigen::MatrixXd &value)
 
string getPropertyAsString (unsigned indx) const
 
drwnPropertyType getPropertyType (unsigned indx) const
 
bool isReadOnly (unsigned indx) const
 
const drwnPropertyInterfacegetProperty (unsigned indx) const
 
const drwnPropertyInterfacegetProperty (const char *name) const
 
bool getBoolProperty (unsigned indx) const
 
int getIntProperty (unsigned indx) const
 
double getDoubleProperty (unsigned indx) const
 
const string & getStringProperty (unsigned indx) const
 
const list< string > & getListProperty (unsigned indx) const
 
int getSelectionProperty (unsigned indx) const
 
const Eigen::VectorXd & getVectorProperty (unsigned indx) const
 
const Eigen::MatrixXd & getMatrixProperty (unsigned indx) const
 
const string & getPropertyName (unsigned indx) const
 
vector< string > getPropertyNames () const
 
void readProperties (drwnXMLNode &xml, const char *tag="property")
 
void writeProperties (drwnXMLNode &xml, const char *tag="property") const
 
void printProperties (ostream &os) const
 

Static Public Attributes

static string BASE_CLASSIFIER = string("drwnBoostedClassifier")
 the base classifier (e.g., drwnBoostedClassifier)
 
static
drwnCompositeClassifierMethod 
METHOD = DRWN_ONE_VS_ALL
 composition method
 

Protected Attributes

string _baseClassifier
 the base classifier (e.g., drwnBoostedClassifier)
 
drwnCompositeClassifierMethod _method
 composition method
 
vector< drwnClassifier * > _binaryClassifiers
 binary classifiers
 
drwnFeatureWhitener _featureWhitener
 feature whitener for output of binary classifiers
 
drwnTMultiClassLogistic
< drwnBiasJointFeatureMap
_calibrationWeights
 calibration weights
 
- Protected Attributes inherited from drwnClassifier
int _nFeatures
 number of features
 
int _nClasses
 number of classes
 
bool _bValid
 true if the classifier has been trained or loaded
 

Additional Inherited Members

- Protected Member Functions inherited from drwnProperties
void declareProperty (const string &name, drwnPropertyInterface *optif)
 
void undeclareProperty (const string &name)
 
void exposeProperties (drwnProperties *opts, const string &prefix=string(""), bool bSerializable=false)
 
virtual void propertyChanged (const string &name)
 

Detailed Description

Implements a multi-class classifier by combining binary classifiers.

A common approach to multi-class classification is to learn a bank of one-versus-one or one-versus-all classifiers and combine their output via multi-class logistic regression. This class provides this functionality where the type of classifier used by the one-versus-one or ones-versus-all bank is controlled by the BASE_CLASSIFIER parameters.

The following code snippet shows example learning a composite classifier (with boosted base classifier) on a training dataset and then testing it on a hold out evaluation dataset.

// load training dataset
dataset.read("training_data.bin");
// train the classifier
const int nFeatures = dataset.numFeatures();
const int nClasses = dataset.maxTarget() + 1;
drwnCompositeClassifier::BASE_CLASSIFIER = string("drwnBoostedClassifier");
drwnCompositeClassifier model(nFeatures, nClasses);
model.train(dataset);
// load evaluation set
dataset.read("testing_data.bin", false);
// predict labels
vector<int> predictions;
model.getClassifications(dataset.features, predictions);
See Also
drwnClassifier, drwnML Tutorial

The documentation for this class was generated from the following files: