Public Member Functions | Public Attributes

CVwEnvironment Class Reference


Detailed Description

Class CVwEnvironment is the environment used by VW.

Contains global constants and settings which change the behaviour of Vowpal Wabbit.

It is used while parsing input, and also while learning.

One CVwEnvironment object should be bound to the CStreamingVwFile or CStreamingVwCacheFile, and the pointer to it propagated upwards to CStreamingVwFeatures and finally to CVowpalWabbit.

Definition at line 39 of file VwEnvironment.h.

Inheritance diagram for CVwEnvironment:
Inheritance graph
[legend]

List of all members.

Public Member Functions

 CVwEnvironment ()
virtual ~CVwEnvironment ()
void set_num_bits (vw_size_t bits)
vw_size_t get_num_bits ()
void set_mask (vw_size_t m)
vw_size_t get_mask ()
float64_t get_min_label ()
float64_t get_max_label ()
vw_size_t num_threads ()
vw_size_t length ()
void set_stride (vw_size_t new_stride)
virtual const char * get_name () const

Public Attributes

vw_size_t num_bits
 log_2 of the number of features
vw_size_t thread_bits
 log_2 of the number of threads
vw_size_t mask
 Mask used for hashing.
vw_size_t thread_mask
 Mask used by regressor for learning.
vw_size_t stride
 Number of elements in weight vector per feature.
float64_t min_label
 Smallest label seen.
float64_t max_label
 Largest label seen.
float32_t eta
 Learning rate.
float32_t eta_decay_rate
 Decay rate of eta per pass.
bool adaptive
 Whether adaptive learning is used.
bool exact_adaptive_norm
 Whether exact norm is used for adaptive learning.
float32_t l1_regularization
 Level of L1 regularization.
bool random_weights
 Whether to use random weights.
float32_t initial_weight
 Initial value of all elements in weight vector.
float32_t update_sum
 Sum of updates.
float32_t t
 Value of t.
float64_t initial_t
 Initial value of t.
float32_t power_t
 t power value while updating
int64_t example_number
 Example number.
float64_t weighted_examples
 Weighted examples.
float64_t weighted_unlabeled_examples
 Weighted unlabelled examples.
float64_t weighted_labels
 Weighted labels.
vw_size_t total_features
 Total number of features.
float64_t sum_loss
 Sum of losses.
vw_size_t passes_complete
 Number of passes complete.
vw_size_t num_passes
 Number of passes.
vw_size_t ngram
 ngrams to generate
vw_size_t skips
 Skips in ngrams.
bool ignore_some
 Whether some namespaces are ignored.
bool ignore [256]
 Which namespaces to ignore.
DynArray< char * > pairs
 Pairs of features to cross for quadratic updates.
const char * vw_version
 VW version.
vw_size_t v_length
 Length of version string.

Constructor & Destructor Documentation

CVwEnvironment (  ) 

Default constructor Should initialize with reasonable default values

Definition at line 19 of file VwEnvironment.cpp.

virtual ~CVwEnvironment (  )  [virtual]

Destructor

Definition at line 51 of file VwEnvironment.h.


Member Function Documentation

vw_size_t get_mask (  ) 

Return the mask used

Returns:
mask

Definition at line 75 of file VwEnvironment.h.

float64_t get_max_label (  ) 

Return maximum label encountered

Returns:
max label

Definition at line 87 of file VwEnvironment.h.

float64_t get_min_label (  ) 

Return minimum label encountered

Returns:
min label

Definition at line 81 of file VwEnvironment.h.

virtual const char* get_name (  )  const [virtual]

Return the name of the object

Returns:
VwEnvironment

Implements CSGObject.

Definition at line 114 of file VwEnvironment.h.

vw_size_t get_num_bits (  ) 

Return number of bits used for weight vector

Returns:
number of bits

Definition at line 63 of file VwEnvironment.h.

vw_size_t length (  ) 

Return length of weight vector

Returns:
length of weight vector

Definition at line 99 of file VwEnvironment.h.

vw_size_t num_threads (  ) 

Return number of threads used for learning

Returns:
number of threads

Definition at line 93 of file VwEnvironment.h.

void set_mask ( vw_size_t  m  ) 

Set mask used while accessing features

Parameters:
m mask

Definition at line 69 of file VwEnvironment.h.

void set_num_bits ( vw_size_t  bits  ) 

Set number of bits used for the weight vector

Parameters:
bits number of bits

Definition at line 57 of file VwEnvironment.h.

void set_stride ( vw_size_t  new_stride  ) 

Set a new stride value. Also changes thread_mask.

Parameters:
new_stride new value of stride

Definition at line 69 of file VwEnvironment.cpp.


Member Data Documentation

bool adaptive

Whether adaptive learning is used.

Definition at line 145 of file VwEnvironment.h.

Learning rate.

Definition at line 140 of file VwEnvironment.h.

Decay rate of eta per pass.

Definition at line 142 of file VwEnvironment.h.

Whether exact norm is used for adaptive learning.

Definition at line 147 of file VwEnvironment.h.

int64_t example_number

Example number.

Definition at line 167 of file VwEnvironment.h.

bool ignore[256]

Which namespaces to ignore.

Definition at line 191 of file VwEnvironment.h.

Whether some namespaces are ignored.

Definition at line 189 of file VwEnvironment.h.

Initial value of t.

Definition at line 162 of file VwEnvironment.h.

Initial value of all elements in weight vector.

Definition at line 154 of file VwEnvironment.h.

Level of L1 regularization.

Definition at line 149 of file VwEnvironment.h.

Mask used for hashing.

Definition at line 128 of file VwEnvironment.h.

Largest label seen.

Definition at line 137 of file VwEnvironment.h.

Smallest label seen.

Definition at line 135 of file VwEnvironment.h.

ngrams to generate

Definition at line 184 of file VwEnvironment.h.

log_2 of the number of features

Definition at line 124 of file VwEnvironment.h.

Number of passes.

Definition at line 181 of file VwEnvironment.h.

DynArray<char*> pairs

Pairs of features to cross for quadratic updates.

Definition at line 194 of file VwEnvironment.h.

Number of passes complete.

Definition at line 179 of file VwEnvironment.h.

t power value while updating

Definition at line 164 of file VwEnvironment.h.

Whether to use random weights.

Definition at line 152 of file VwEnvironment.h.

Skips in ngrams.

Definition at line 186 of file VwEnvironment.h.

Number of elements in weight vector per feature.

Definition at line 132 of file VwEnvironment.h.

Sum of losses.

Definition at line 177 of file VwEnvironment.h.

Value of t.

Definition at line 160 of file VwEnvironment.h.

log_2 of the number of threads

Definition at line 126 of file VwEnvironment.h.

Mask used by regressor for learning.

Definition at line 130 of file VwEnvironment.h.

Total number of features.

Definition at line 175 of file VwEnvironment.h.

Sum of updates.

Definition at line 157 of file VwEnvironment.h.

Length of version string.

Definition at line 199 of file VwEnvironment.h.

const char* vw_version

VW version.

Definition at line 197 of file VwEnvironment.h.

Weighted examples.

Definition at line 169 of file VwEnvironment.h.

Weighted labels.

Definition at line 173 of file VwEnvironment.h.

Weighted unlabelled examples.

Definition at line 171 of file VwEnvironment.h.


The documentation for this class was generated from the following files:
 All Classes Namespaces Files Functions Variables Typedefs Enumerations Enumerator Friends Defines

SHOGUN Machine Learning Toolbox - Documentation