CVwRegressor Class Reference

---

Detailed Description

Regressor used by VW.

Stores the weight vectors and loss object, and is used for calculating losses and updates.

The weight vector uses 'num_bits' number of bits, set in the environment object to store weights.

Definition at line 35 of file VwRegressor.h.

Inheritance diagram for CVwRegressor:

List of all members.

Public Member Functions
	CVwRegressor ()
	CVwRegressor (CVwEnvironment *env_to_use)
virtual	~CVwRegressor ()
float64_t	get_loss (float64_t prediction, float64_t label)
float64_t	get_update (float64_t prediction, float64_t label, float64_t eta_t, float64_t norm)
virtual void	dump_regressor (char *reg_name, bool as_text)
virtual void	load_regressor (char *file_name)
virtual const char *	get_name () const
virtual void	init (CVwEnvironment *env_to_use=NULL)
Public Attributes
float32_t **	weight_vectors
	Weight vectors, one array for each thread.
CLossFunction *	loss
	Loss function.
Protected Attributes
CVwEnvironment *	env
	Environment.

---

Constructor & Destructor Documentation

CVwRegressor

(

)

Default constructor

Definition at line 21 of file VwRegressor.cpp.

CVwRegressor

(

CVwEnvironment *

env_to_use

)

Constructor taking an environment object

Parameters:

env_to_use

environment

Definition at line 29 of file VwRegressor.cpp.

~CVwRegressor

(

)

[virtual]

Destructor

Definition at line 37 of file VwRegressor.cpp.

---

Member Function Documentation

void dump_regressor	(	char *	reg_name,
		bool	as_text
	)			[virtual]

Dump regressor in binary/text form

Parameters:

	reg_name	output file name
	as_text	whether to dump as text

Definition at line 81 of file VwRegressor.cpp.

float64_t get_loss	(	float64_t	prediction,
		float64_t	label
	)

Get loss for a label-prediction set

Parameters:

	prediction	prediction
	label	label

Returns:

loss

Definition at line 63 of file VwRegressor.h.

virtual const char* get_name

(

)

const [virtual]

Return name of the object

Returns:

VwRegressor

Implements CSGObject.

Definition at line 103 of file VwRegressor.h.

float64_t get_update	(	float64_t	prediction,
		float64_t	label,
		float64_t	eta_t,
		float64_t	norm
	)

Get weight update for a prediction-label set

Parameters:

	prediction	prediction
	label	label
	eta_t	learning rate
	norm	scaling norm

Returns:

update

Definition at line 78 of file VwRegressor.h.

void init

(

CVwEnvironment *

env_to_use = NULL

)

[virtual]

Initialize weight vectors

Parameters:

env_to_use

environment object

Definition at line 44 of file VwRegressor.cpp.

void load_regressor

(

char *

file_name

)

[virtual]

Load the regressor from a file

Parameters:

file_name

name of dumped regressor binary file

Definition at line 168 of file VwRegressor.cpp.

---

Member Data Documentation

CVwEnvironment* env [protected]

Environment.

Definition at line 120 of file VwRegressor.h.

CLossFunction* loss

Loss function.

Definition at line 116 of file VwRegressor.h.

float32_t** weight_vectors

Weight vectors, one array for each thread.

Definition at line 114 of file VwRegressor.h.

---

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

• VwRegressor.h
• VwRegressor.cpp