LogRegSGD Class Reference

#include <LogRegSGD.h>

Inheritance diagram for LogRegSGD:

Public Member Functions
	LogRegSGD (double _eta=0.01, int iter=10, bool _shuffle=false, int _randomState=0)
void	fit (const Matrix< double > &X, const Matrix< double > &y) override
void	partial_fit (const Matrix< double > &X, const Matrix< double > &y)
Matrix< double >	netInput (const Matrix< double > &X) override
Matrix< double >	activation (const Matrix< double > &X) override
Matrix< double >	predict (const Matrix< double > &X) override
double	costFunction (const Matrix< double > &mat) override
Public Member Functions inherited from ANNClassifier
	ANNClassifier (double _eta, int _n_iter)
virtual Matrix< double >	activation (const Matrix< double > &)=0
virtual Matrix< double >	netInput (const Matrix< double > &)=0
virtual double	costFunction (const Matrix< double > &)=0
Public Member Functions inherited from Classifier
	Classifier ()
void	initialize_weights (size_t numRows, size_t numColumns=1)
void	update_weights (const Matrix< double > &update, const Matrix< double > &delta)
Matrix< double >	transform (const Matrix< double > &in) override
virtual void	fit (const Matrix< double > &X, const Matrix< double > &y)=0
virtual Matrix< double >	predict (const Matrix< double > &)=0
virtual Matrix< double >	transform (const Matrix< double > &)=0

Public Attributes
bool	shuffle
	signalizes whether given dataset should be shuffled while fitting More...
int	randomState
	initialize weights with random state More...
SGD *	sgd = nullptr
	algorithmic object to represent fitting algorithm More...
Public Attributes inherited from Classifier
Matrix< double >	weights
	Vector holding weights. More...
Matrix< double >	costs
	Vector holding classification error per epoch. More...

Private Member Functions
void	initialize_weights (size_t m)
double	update_weights (const Matrix< double > &xi, const Matrix< double > &target)

Static Private Member Functions
static Matrix< double >	Log (const Matrix< double > &in)

Additional Inherited Members
Protected Attributes inherited from ANNClassifier
double	eta
	Learning rate. More...
int	n_iter
	number epochs More...
Protected Attributes inherited from Classifier
bool	w_initialized = false
	flag to initialize weights only once More...

Detailed Description

Logistical regression model using statistic gradient decent method for model fitting.

Examples

ds/TestLogRegSGD.cpp.

Constructor & Destructor Documentation

LogRegSGD()

LogRegSGD::LogRegSGD ( double  _eta = 0.01, int  iter = 10, bool  _shuffle = false, int  _randomState = 0  )

inlineexplicit

default constructor

Parameters

_eta	learning rate
iter	number learning iterations
_shuffle	shuffle data in each iteration?
_randomState	seed for random state

Member Function Documentation

activation()

Matrix< double > LogRegSGD::activation ( const Matrix< double > &  X )

inlineoverridevirtual

activate given input $$x\mapsto \phi(x)$$

Parameters

X	input values

Returns

activated values

Implements ANNClassifier.

costFunction()

double LogRegSGD::costFunction ( const Matrix< double > &  mat )

inlineoverridevirtual

do not use!

Parameters

mat

Returns

Implements ANNClassifier.

fit()

void LogRegSGD::fit ( const Matrix< double > &  X, const Matrix< double > &  y  )

inlineoverridevirtual

fits the weights of the model for given input values

Parameters

X	array-like with the shape: [n_samples, n_features]
y	array-like with shape: [n_samples, 1]

Returns

this

Implements Predictor.

initialize_weights()

void LogRegSGD::initialize_weights ( size_t  m )

inlineprivate

helper to initialize weights matrix

Parameters

m

dimension

Log()

static Matrix< double > LogRegSGD::Log ( const Matrix< double > &  in )

inlinestaticprivate

Calculates log of each element. Implementation doesn't use Matrix::Apply

Parameters

in	input matrix

Returns

netInput()

Matrix< double > LogRegSGD::netInput ( const Matrix< double > &  X )

inlineoverridevirtual

calculate netinput

Parameters

X	input values

Returns

Implements ANNClassifier.

partial_fit()

void LogRegSGD::partial_fit ( const Matrix< double > &  X, const Matrix< double > &  y  )

inline

partially fit weights

Parameters

X	input values
y	target output values

predict()

Matrix< double > LogRegSGD::predict ( const Matrix< double > &  X )

inlineoverridevirtual

predict output for given input

Parameters

X	input values

Returns

predicted output for given input

Implements Predictor.

update_weights()

double LogRegSGD::update_weights ( const Matrix< double > &  xi, const Matrix< double > &  target  )

inlineprivate

logistical regression

Parameters

xi	input values
target	target output values

Returns

cost

Member Data Documentation

randomState

int LogRegSGD::randomState

initialize weights with random state

sgd

SGD* LogRegSGD::sgd = nullptr

algorithmic object to represent fitting algorithm

shuffle

bool LogRegSGD::shuffle

signalizes whether given dataset should be shuffled while fitting

---

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

• include/math/ds/LogRegSGD.h