BALL  1.4.2
 All Classes Namespaces Files Functions Variables Typedefs Enumerations Enumerator Properties Friends Macros Groups Pages
List of all members
BALL::QSAR::GPModel Class Reference

#include <BALL/QSAR/gpModel.h>

Inheritance diagram for BALL::QSAR::GPModel:
BALL::QSAR::KernelModel BALL::QSAR::NonLinearModel BALL::QSAR::RegressionModel BALL::QSAR::Model

Public Member Functions

Constructors and Destructors
 GPModel (const QSARData &q, int k_type, double p1, double p2=-1)
 
 GPModel (const QSARData &q, Vector< double > &w)
 
 GPModel (const QSARData &q, String s1, String s2)
 
 GPModel (const QSARData &q, const LinearModel &lm, int column)
 
 ~GPModel ()
 
Accessors
void train ()
 
Vector< doublepredict (const vector< double > &substance, bool transform=1)
 
double calculateStdErr ()
 
void setParameters (vector< double > &v)
 
vector< doublegetParameters () const
 
- Public Member Functions inherited from BALL::QSAR::KernelModel
 KernelModel (const QSARData &q, int k_type, double p1, double p2)
 
 KernelModel (const QSARData &q, String f, String g)
 
 KernelModel (const QSARData &q, Vector< double > &w)
 
 KernelModel (const QSARData &q, const LinearModel &lm, int column)
 
 ~KernelModel ()
 
virtual void saveToFile (string filename)
 
virtual void readFromFile (string filename)
 
void operator= (const Model &m)
 
- Public Member Functions inherited from BALL::QSAR::NonLinearModel
 NonLinearModel (const QSARData &q)
 
 ~NonLinearModel ()
 

Private Attributes

Attributes
Matrix< doubleL_
 
Vector< doubleinput_
 
Vector< doubleK_t_
 
double lambda_
 

Additional Inherited Members

- Public Attributes inherited from BALL::QSAR::KernelModel
Kernelkernel
 
- Public Attributes inherited from BALL::QSAR::RegressionModel
RegressionValidationvalidation
 
- Public Attributes inherited from BALL::QSAR::Model
const QSARDatadata
 
Validationmodel_val
 
- Protected Member Functions inherited from BALL::QSAR::KernelModel
void calculateOffsets ()
 
void readKernelParametersFromFile (std::ifstream &in)
 
void saveKernelParametersToFile (std::ofstream &out)
 
void saveTrainingResult (std::ofstream &out)
 
void readTrainingResult (std::ifstream &input, int no_substances, int no_y)
 
- Protected Member Functions inherited from BALL::QSAR::NonLinearModel
- Protected Member Functions inherited from BALL::QSAR::RegressionModel
void readDescriptorInformationFromFile (std::ifstream &in, int no_descriptors, bool transformation, int no_coefficients)
 
void saveDescriptorInformationToFile (std::ofstream &out)
 
- Protected Attributes inherited from BALL::QSAR::KernelModel
Matrix< doubleK_
 
- Protected Attributes inherited from BALL::QSAR::RegressionModel
BALL::Matrix< doubletraining_result_
 
Vector< doubleoffsets_
 
- Friends inherited from BALL::QSAR::KernelModel

Detailed Description

class for gaussian process regression

Definition at line 41 of file gpModel.h.

Constructor & Destructor Documentation

BALL::QSAR::GPModel::GPModel ( const QSARData q,
int  k_type,
double  p1,
double  p2 = -1 
)
BALL::QSAR::GPModel::GPModel ( const QSARData q,
Vector< double > &  w 
)
BALL::QSAR::GPModel::GPModel ( const QSARData q,
String  s1,
String  s2 
)

constructor that sets KernelModel.f to s1 and KernelModel.g to s2

BALL::QSAR::GPModel::GPModel ( const QSARData q,
const LinearModel lm,
int  column 
)
BALL::QSAR::GPModel::~GPModel ( )

Member Function Documentation

double BALL::QSAR::GPModel::calculateStdErr ( )
virtual

calculates standart error for the last prediction as $ \sqrt{k(x_*,x_*)-\sum_{i=1}^n \sum_{j=1}^n k(x_*,x_i)*k(x_*,x_j)-L_{ij} } $

Reimplemented from BALL::QSAR::Model.

vector<double> BALL::QSAR::GPModel::getParameters ( ) const
virtual

Reimplemented from BALL::QSAR::Model.

Vector<double> BALL::QSAR::GPModel::predict ( const vector< double > &  substance,
bool  transform = 1 
)
virtual
Predicts the activities of a given substance
Parameters
substancethe substance which activity is to be predicted in form of a vecor containing the values for all descriptors (if neccessary, relevant descriptors will be selected automatically)
transformdetermines whether the values for each descriptor of the given substance should be transformed before prediction of activity.
If (transform==1): each descriptor value is transformed according to the centering of the respective column of QSARData.descriptor_matrix used to train this model.
If the substance to be predicted is part of the same input data (e.g. same SD-file) as the training data (as is the case during cross validation), transform should therefore be set to 0.
Returns
a RowVector containing one value for each predicted activity

Reimplemented from BALL::QSAR::KernelModel.

void BALL::QSAR::GPModel::setParameters ( vector< double > &  )
virtual

sets the model parameters according to the given values.

Reimplemented from BALL::QSAR::Model.

void BALL::QSAR::GPModel::train ( )
virtual

Starts training the model.

Implements BALL::QSAR::Model.

Member Data Documentation

Vector<double> BALL::QSAR::GPModel::input_
private

the last predicted substance

Definition at line 81 of file gpModel.h.

Vector<double> BALL::QSAR::GPModel::K_t_
private

Definition at line 83 of file gpModel.h.

Matrix<double> BALL::QSAR::GPModel::L_
private

Definition at line 78 of file gpModel.h.

double BALL::QSAR::GPModel::lambda_
private

Definition at line 85 of file gpModel.h.