Cosan::CosanLinearModel< NumericType > Class Template Reference

CosanLinear Model. All linear model inherit this object. More...

#include <CosanLinearModel.h>

Inheritance diagram for Cosan::CosanLinearModel< NumericType >:

Public Member Functions
	CosanLinearModel ()=delete
	CosanLinearModel (bool Bias)
void	SetBias (const bool NewBias)
virtual void	SetBeta (CosanMatrix< NumericType > InitBeta)
const CosanMatrix< NumericType > &	GetBeta () const
bool	GetBias () const
virtual CosanMatrix< NumericType >	predict (const CosanMatrix< NumericType > &X)=0
virtual const std::string	GetName () const
	Get the name of the objects. More...
Public Member Functions inherited from Cosan::CosanModel
	CosanModel ()
virtual void	SetMaxTrainTime (double t)
virtual double	GetMaxTrainTime () const
virtual EModelType	GetModelType ()
virtual EProblemType	GetProblemType ()
virtual ESolverType	GetSolverType () const
void	SetSolverType (ESolverType sr)
Public Member Functions inherited from Cosan::CosanBO
	CosanBO ()
	Default constructor. More...

Protected Attributes
CosanMatrix< NumericType >	MBeta
bool	MBias
Protected Attributes inherited from Cosan::CosanModel
double	MaxTrainTime
ESolverType	SolverType

Private Member Functions
void	Init ()

Detailed Description

template<Numeric NumericType>
class Cosan::CosanLinearModel< NumericType >

CosanLinear Model. All linear model inherit this object.

To initialize model object, one can do

model(); // The default constructor where the model is without intercept. 

model(bool bias); 

model(NumericType param, bool bias);//for model with only one parameter. Ridge regression and PCR are the cases. 

model(std::vector<NumericType> params, bool bias);//for model with more than one parameter. Each parameter is the enry of params. PCR with L2 square norm is the case. 

To initialize model object and run model fitting automatically, one can do

model(CosanRawData<NumericType> RD,NumericType param ,bool bias = false);  

model(CosanData<NumericType> CD,NumericType param ,bool bias = false); 

model(CosanMatrix<NumericType> X,CosanMatrix<NumericType> Y,NumericType param,bool bias = false); 

For model with more than 1 parameter, one should replace NumericType param by std::vector<NumericType> params.

model.fit(CosanMatrix<NumericType> X,CosanMatrix<NumericType> Y); 

model.predict(CosanMatrix<NumericType> X); 

To set new params and get current params, one can do

model.SetParams(NumericType param); //for model with more than 1 parameter, input should be std::vecotr<NumericType> params 

model.GetParams(); 

To obtain the coefficient for the model one can use

model.GetBeta(); 

Definition at line 32 of file CosanLinearModel.h.

Constructor & Destructor Documentation

CosanLinearModel() [1/2]

template<Numeric NumericType>

Cosan::CosanLinearModel< NumericType >::CosanLinearModel ( )

delete

CosanLinearModel() [2/2]

template<Numeric NumericType>

Cosan::CosanLinearModel< NumericType >::CosanLinearModel ( bool  Bias )

inline

Definition at line 36 of file CosanLinearModel.h.

                                       : CosanModel(){
                Init();
                MBias=Bias;
            }

Member Function Documentation

GetBeta()

template<Numeric NumericType>

const CosanMatrix<NumericType>& Cosan::CosanLinearModel< NumericType >::GetBeta ( ) const

inline

Definition at line 44 of file CosanLinearModel.h.

{return MBeta;}

GetBias()

template<Numeric NumericType>

bool Cosan::CosanLinearModel< NumericType >::GetBias ( ) const

inline

Definition at line 45 of file CosanLinearModel.h.

{return MBias;}

GetName()

template<Numeric NumericType>

virtual const std::string Cosan::CosanLinearModel< NumericType >::GetName ( ) const

inlinevirtual

Get the name of the objects.

Description: Get the name of the object. It should return "Abstract Object"

Reimplemented from Cosan::CosanBO.

Reimplemented in Cosan::CosanRidgeRegression< NumericType >, Cosan::CosanPrincipalComponentRegression< NumericType >, Cosan::CosanPCRRidge< NumericType >, and Cosan::CosanLinearRegression< NumericType >.

Definition at line 54 of file CosanLinearModel.h.

{return "LinearModel";}

Init()

template<Numeric NumericType>

void Cosan::CosanLinearModel< NumericType >::Init ( )

inlineprivate

Definition at line 59 of file CosanLinearModel.h.

                           {
            }

predict()

template<Numeric NumericType>

virtual CosanMatrix<NumericType> Cosan::CosanLinearModel< NumericType >::predict ( const CosanMatrix< NumericType > &  X )

pure virtual

Implemented in Cosan::CosanPrincipalComponentRegression< NumericType >, and Cosan::CosanLinearRegression< NumericType >.

SetBeta()

template<Numeric NumericType>

virtual void Cosan::CosanLinearModel< NumericType >::SetBeta ( CosanMatrix< NumericType >  InitBeta )

inlinevirtual

Definition at line 42 of file CosanLinearModel.h.

{MBeta=InitBeta;}

SetBias()

template<Numeric NumericType>

void Cosan::CosanLinearModel< NumericType >::SetBias ( const bool  NewBias )

inline

Definition at line 41 of file CosanLinearModel.h.

{MBias = NewBias;}

Member Data Documentation

MBeta

template<Numeric NumericType>

CosanMatrix<NumericType> Cosan::CosanLinearModel< NumericType >::MBeta

protected

Definition at line 56 of file CosanLinearModel.h.

MBias

template<Numeric NumericType>

bool Cosan::CosanLinearModel< NumericType >::MBias

protected

Definition at line 57 of file CosanLinearModel.h.

---

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

• cosan/model/CosanLinearModel.h