Public Member Functions | Static Public Member Functions | Protected Member Functions | Protected Attributes

CWeightedDegreeStringKernel Class Reference

Detailed Description

The Weighted Degree String kernel.

The WD kernel of order d compares two sequences ${\bf x}$ and ${\bf x'}$ of length L by summing all contributions of k-mer matches of lengths $k\in\{1,\dots,d\}$, weighted by coefficients $\beta_k$. It is defined as

\[ k({\bf x},{\bf x'})=\sum_{k=1}^d\beta_k\sum_{l=1}^{L-k+1}I({\bf u}_{k,l}({\bf x})={\bf u}_{k,l}({\bf x'})). \]

Here, ${\bf u}_{k,l}({\bf x})$ is the string of length k starting at position l of the sequence ${\bf x}$ and $I(\cdot)$ is the indicator function which evaluates to 1 when its argument is true and to 0 otherwise.

Definition at line 53 of file WeightedDegreeStringKernel.h.

Inheritance diagram for CWeightedDegreeStringKernel:
Inheritance graph
[legend]

List of all members.

Public Member Functions

 CWeightedDegreeStringKernel ()
 CWeightedDegreeStringKernel (int32_t degree, EWDKernType type=E_WD)
 CWeightedDegreeStringKernel (float64_t *weights, int32_t degree)
 CWeightedDegreeStringKernel (CStringFeatures< char > *l, CStringFeatures< char > *r, int32_t degree)
virtual ~CWeightedDegreeStringKernel ()
virtual bool init (CFeatures *l, CFeatures *r)
virtual void cleanup ()
EWDKernType get_type () const
virtual EKernelType get_kernel_type ()
virtual const char * get_name () const
virtual bool init_optimization (int32_t count, int32_t *IDX, float64_t *alphas)
virtual bool init_optimization (int32_t count, int32_t *IDX, float64_t *alphas, int32_t tree_num)
virtual bool delete_optimization ()
virtual float64_t compute_optimized (int32_t idx)
virtual void compute_batch (int32_t num_vec, int32_t *vec_idx, float64_t *target, int32_t num_suppvec, int32_t *IDX, float64_t *alphas, float64_t factor=1.0)
virtual void clear_normal ()
virtual void add_to_normal (int32_t idx, float64_t weight)
virtual int32_t get_num_subkernels ()
void compute_by_subkernel (int32_t idx, float64_t *subkernel_contrib)
const float64_t * get_subkernel_weights (int32_t &num_weights)
void set_subkernel_weights (float64_t *weights2, int32_t num_weights2)
virtual bool set_normalizer (CKernelNormalizer *normalizer_)
float64_t * compute_abs_weights (int32_t &len)
void compute_by_tree (int32_t idx, float64_t *LevelContrib)
bool is_tree_initialized ()
float64_t * get_degree_weights (int32_t &d, int32_t &len)
float64_t * get_weights (int32_t &num_weights)
float64_t * get_position_weights (int32_t &len)
bool set_wd_weights_by_type (EWDKernType type)
void set_wd_weights (SGVector< float64_t > new_weights)
bool set_weights (SGMatrix< float64_t > new_weights)
bool set_position_weights (float64_t *pws, int32_t len)
bool init_block_weights ()
bool init_block_weights_from_wd ()
bool init_block_weights_from_wd_external ()
bool init_block_weights_const ()
bool init_block_weights_linear ()
bool init_block_weights_sqpoly ()
bool init_block_weights_cubicpoly ()
bool init_block_weights_exp ()
bool init_block_weights_log ()
bool delete_position_weights ()
bool set_max_mismatch (int32_t max)
int32_t get_max_mismatch () const
bool set_degree (int32_t deg)
int32_t get_degree () const
bool set_use_block_computation (bool block)
bool get_use_block_computation ()
bool set_mkl_stepsize (int32_t step)
int32_t get_mkl_stepsize ()
bool set_which_degree (int32_t which)
int32_t get_which_degree ()

Static Public Member Functions

static void * compute_batch_helper (void *p)

Protected Member Functions

void create_empty_tries ()
void add_example_to_tree (int32_t idx, float64_t weight)
void add_example_to_single_tree (int32_t idx, float64_t weight, int32_t tree_num)
void add_example_to_tree_mismatch (int32_t idx, float64_t weight)
void add_example_to_single_tree_mismatch (int32_t idx, float64_t weight, int32_t tree_num)
float64_t compute_by_tree (int32_t idx)
float64_t compute (int32_t idx_a, int32_t idx_b)
float64_t compute_with_mismatch (char *avec, int32_t alen, char *bvec, int32_t blen)
float64_t compute_without_mismatch (char *avec, int32_t alen, char *bvec, int32_t blen)
float64_t compute_without_mismatch_matrix (char *avec, int32_t alen, char *bvec, int32_t blen)
float64_t compute_using_block (char *avec, int32_t alen, char *bvec, int32_t blen)
virtual void remove_lhs ()

Protected Attributes

float64_t * weights
int32_t weights_degree
int32_t weights_length
float64_t * position_weights
int32_t position_weights_len
float64_t * weights_buffer
int32_t mkl_stepsize
int32_t degree
int32_t length
int32_t max_mismatch
int32_t seq_length
bool initialized
bool block_computation
float64_t * block_weights
EWDKernType type
int32_t which_degree
CTrie< DNATrie > * tries
bool tree_initialized
CAlphabet * alphabet

Constructor & Destructor Documentation

CWeightedDegreeStringKernel (  ) 

default constructor

Definition at line 48 of file WeightedDegreeStringKernel.cpp.

CWeightedDegreeStringKernel ( int32_t  degree,
EWDKernType  type = E_WD 
)

constructor

Parameters:
degree degree
type weighted degree kernel type

Definition at line 55 of file WeightedDegreeStringKernel.cpp.

CWeightedDegreeStringKernel ( float64_t *  weights,
int32_t  degree 
)

constructor

Parameters:
weights kernel's weights
degree degree

Definition at line 68 of file WeightedDegreeStringKernel.cpp.

CWeightedDegreeStringKernel ( CStringFeatures< char > *  l,
CStringFeatures< char > *  r,
int32_t  degree 
)

constructor

Parameters:
l features of left-hand side
r features of right-hand side
degree degree

Definition at line 85 of file WeightedDegreeStringKernel.cpp.

~CWeightedDegreeStringKernel (  )  [virtual]

Definition at line 97 of file WeightedDegreeStringKernel.cpp.

Member Function Documentation

void add_example_to_single_tree ( int32_t  idx,
float64_t  weight,
int32_t  tree_num 
) [protected]

add example to single tree

Parameters:
idx index
weight weight
tree_num which tree

Definition at line 436 of file WeightedDegreeStringKernel.cpp.

void add_example_to_single_tree_mismatch ( int32_t  idx,
float64_t  weight,
int32_t  tree_num 
) [protected]

add example to single tree mismatch

Parameters:
idx index
weight weight
tree_num which tree

Definition at line 487 of file WeightedDegreeStringKernel.cpp.

void add_example_to_tree ( int32_t  idx,
float64_t  weight 
) [protected]

add example to tree

Parameters:
idx index
weight weight

Definition at line 390 of file WeightedDegreeStringKernel.cpp.

void add_example_to_tree_mismatch ( int32_t  idx,
float64_t  weight 
) [protected]

add example to tree mismatch

Parameters:
idx index
weight weight

Definition at line 461 of file WeightedDegreeStringKernel.cpp.

virtual void add_to_normal ( int32_t  idx,
float64_t  weight 
) [virtual]

add to normal

Parameters:
idx where to add
weight what to add

Reimplemented from CKernel.

Definition at line 213 of file WeightedDegreeStringKernel.h.

void cleanup (  )  [virtual]

clean up kernel

Reimplemented from CKernel.

Definition at line 183 of file WeightedDegreeStringKernel.cpp.

virtual void clear_normal (  )  [virtual]

clear normal subkernel functionality

Reimplemented from CKernel.

Definition at line 195 of file WeightedDegreeStringKernel.h.

float64_t compute ( int32_t  idx_a,
int32_t  idx_b 
) [protected, virtual]

compute kernel function for features a and b idx_{a,b} denote the index of the feature vectors in the corresponding feature object

Parameters:
idx_a index a
idx_b index b
Returns:
computed kernel function at indices a,b

Implements CKernel.

Definition at line 364 of file WeightedDegreeStringKernel.cpp.

float64_t * compute_abs_weights ( int32_t &  len  ) 

compute abs weights

Parameters:
len len
Returns:
computed abs weights

Definition at line 566 of file WeightedDegreeStringKernel.cpp.

void compute_batch ( int32_t  num_vec,
int32_t *  vec_idx,
float64_t *  target,
int32_t  num_suppvec,
int32_t *  IDX,
float64_t *  alphas,
float64_t  factor = 1.0 
) [virtual]

compute batch

Parameters:
num_vec number of vectors
vec_idx vector index
target target
num_suppvec number of support vectors
IDX IDX
alphas alphas
factor factor

Reimplemented from CKernel.

Definition at line 863 of file WeightedDegreeStringKernel.cpp.

void * compute_batch_helper ( void *  p  )  [static]

helper for compute batch

Parameters:
p thread parameter

Definition at line 827 of file WeightedDegreeStringKernel.cpp.

void compute_by_subkernel ( int32_t  idx,
float64_t *  subkernel_contrib 
) [virtual]

compute by subkernel

Parameters:
idx index
subkernel_contrib subkernel contribution

Reimplemented from CKernel.

Definition at line 247 of file WeightedDegreeStringKernel.h.

void compute_by_tree ( int32_t  idx,
float64_t *  LevelContrib 
)

compute by tree

Parameters:
idx index
LevelContrib level contribution
Returns:
computed value

Definition at line 539 of file WeightedDegreeStringKernel.cpp.

float64_t compute_by_tree ( int32_t  idx  )  [protected]

compute by tree

Parameters:
idx index
Returns:
computed value

Definition at line 515 of file WeightedDegreeStringKernel.cpp.

virtual float64_t compute_optimized ( int32_t  idx  )  [virtual]

compute optimized

Parameters:
idx index to compute
Returns:
optimized value at given index

Reimplemented from CKernel.

Definition at line 162 of file WeightedDegreeStringKernel.h.

float64_t compute_using_block ( char *  avec,
int32_t  alen,
char *  bvec,
int32_t  blen 
) [protected]

compute using block

Parameters:
avec vector a
alen length of vector a
bvec vector b
blen length of vector b
Returns:
computed value

Definition at line 291 of file WeightedDegreeStringKernel.cpp.

float64_t compute_with_mismatch ( char *  avec,
int32_t  alen,
char *  bvec,
int32_t  blen 
) [protected]

compute with mismatch

Parameters:
avec vector a
alen length of vector a
bvec vector b
blen length of vector b
Returns:
computed value

Definition at line 263 of file WeightedDegreeStringKernel.cpp.

float64_t compute_without_mismatch ( char *  avec,
int32_t  alen,
char *  bvec,
int32_t  blen 
) [protected]

compute without mismatch

Parameters:
avec vector a
alen length of vector a
bvec vector b
blen length of vector b
Returns:
computed value

Definition at line 317 of file WeightedDegreeStringKernel.cpp.

float64_t compute_without_mismatch_matrix ( char *  avec,
int32_t  alen,
char *  bvec,
int32_t  blen 
) [protected]

compute without mismatch matrix

Parameters:
avec vector a
alen length of vector a
bvec vector b
blen length of vector b
Returns:
computed value

Definition at line 340 of file WeightedDegreeStringKernel.cpp.

void create_empty_tries (  )  [protected]

create emtpy tries

Definition at line 128 of file WeightedDegreeStringKernel.cpp.

bool delete_optimization (  )  [virtual]

delete optimization

Returns:
if deleting was successful

Reimplemented from CKernel.

Definition at line 249 of file WeightedDegreeStringKernel.cpp.

bool delete_position_weights (  ) 

delete position weights

Returns:
if deleting was successful

Definition at line 518 of file WeightedDegreeStringKernel.h.

int32_t get_degree (  )  const

get degree

Returns:
degree

Definition at line 549 of file WeightedDegreeStringKernel.h.

float64_t* get_degree_weights ( int32_t &  d,
int32_t &  len 
)

get degree weights

Parameters:
d degree weights will be stored here
len number of degree weights will be stored here

Definition at line 389 of file WeightedDegreeStringKernel.h.

virtual EKernelType get_kernel_type (  )  [virtual]

return what type of kernel we are

Returns:
kernel type WEIGHTEDDEGREE

Implements CStringKernel< char >.

Definition at line 115 of file WeightedDegreeStringKernel.h.

int32_t get_max_mismatch (  )  const

get maximum mismatch

Returns:
maximum mismatch

Definition at line 536 of file WeightedDegreeStringKernel.h.

int32_t get_mkl_stepsize (  ) 

get MKL step size

Returns:
MKL step size

Definition at line 585 of file WeightedDegreeStringKernel.h.

virtual const char* get_name ( void   )  const [virtual]

return the kernel's name

Returns:
name WeightedDegree

Reimplemented from CStringKernel< char >.

Definition at line 121 of file WeightedDegreeStringKernel.h.

virtual int32_t get_num_subkernels (  )  [virtual]

get number of subkernels

Returns:
number of subkernels

Reimplemented from CKernel.

Definition at line 231 of file WeightedDegreeStringKernel.h.

float64_t* get_position_weights ( int32_t &  len  ) 

get position weights

Parameters:
len number of position weights will be stored here
Returns:
position weights

Definition at line 424 of file WeightedDegreeStringKernel.h.

const float64_t* get_subkernel_weights ( int32_t &  num_weights  )  [virtual]

get subkernel weights

Parameters:
num_weights number of weights will be stored here
Returns:
subkernel weights

Reimplemented from CKernel.

Definition at line 269 of file WeightedDegreeStringKernel.h.

EWDKernType get_type (  )  const

get WD kernel weighting type

Returns:
weighting type
See also:
EWDKernType

Definition at line 106 of file WeightedDegreeStringKernel.h.

bool get_use_block_computation (  ) 

check if block computation is performed

Returns:
if block computation is performed

Definition at line 566 of file WeightedDegreeStringKernel.h.

float64_t* get_weights ( int32_t &  num_weights  ) 

get weights

Parameters:
num_weights number of weights will be stored here
Returns:
weights

Definition at line 401 of file WeightedDegreeStringKernel.h.

int32_t get_which_degree (  ) 

get which degree

Returns:
which degree

Definition at line 602 of file WeightedDegreeStringKernel.h.

bool init ( CFeatures *  l,
CFeatures *  r 
) [virtual]

initialize kernel

Parameters:
l features of left-hand side
r features of right-hand side
Returns:
if initializing was successful

Reimplemented from CStringKernel< char >.

Definition at line 141 of file WeightedDegreeStringKernel.cpp.

bool init_block_weights (  ) 

initialize block weights

Returns:
if initialization was successful

Definition at line 802 of file WeightedDegreeStringKernel.cpp.

bool init_block_weights_const (  ) 

initialize block weights constant

Returns:
if initialization was successful

Definition at line 726 of file WeightedDegreeStringKernel.cpp.

bool init_block_weights_cubicpoly (  ) 

initialize block weights cubic polynomial

Returns:
if initialization was successful

Definition at line 761 of file WeightedDegreeStringKernel.cpp.

bool init_block_weights_exp (  ) 

initialize block weights exponential

Returns:
if initialization was successful

Definition at line 774 of file WeightedDegreeStringKernel.cpp.

bool init_block_weights_from_wd (  ) 

initialize block weights from weighted degree

Returns:
if initialization was successful

Definition at line 685 of file WeightedDegreeStringKernel.cpp.

bool init_block_weights_from_wd_external (  ) 

initialize block weights from external weighted degree

Returns:
if initialization was successful

Definition at line 702 of file WeightedDegreeStringKernel.cpp.

bool init_block_weights_linear (  ) 

initialize block weights linear

Returns:
if initialization was successful

Definition at line 736 of file WeightedDegreeStringKernel.cpp.

bool init_block_weights_log (  ) 

initialize block weights logarithmic

Returns:
if initialization was successful

Definition at line 788 of file WeightedDegreeStringKernel.cpp.

bool init_block_weights_sqpoly (  ) 

initialize block weights squared polynomial

Returns:
if initialization was successful

Definition at line 747 of file WeightedDegreeStringKernel.cpp.

bool init_optimization ( int32_t  count,
int32_t *  IDX,
float64_t *  alphas,
int32_t  tree_num 
) [virtual]

initialize optimization do initialization for tree_num up to upto_tree, use tree_num=-1 to construct all trees

Parameters:
count count
IDX IDX
alphas alphas
tree_num which tree
Returns:
if initializing was successful

Definition at line 207 of file WeightedDegreeStringKernel.cpp.

virtual bool init_optimization ( int32_t  count,
int32_t *  IDX,
float64_t *  alphas 
) [virtual]

initialize optimization

Parameters:
count count
IDX index
alphas alphas
Returns:
if initializing was successful

Reimplemented from CKernel.

Definition at line 132 of file WeightedDegreeStringKernel.h.

bool is_tree_initialized (  ) 

check if tree is initialized

Returns:
if tree is initialized

Definition at line 382 of file WeightedDegreeStringKernel.h.

void remove_lhs (  )  [protected, virtual]

remove lhs from kernel

Reimplemented from CKernel.

Definition at line 117 of file WeightedDegreeStringKernel.cpp.

bool set_degree ( int32_t  deg  ) 

set degree

Parameters:
deg new degree
Returns:
if setting was successful

Definition at line 543 of file WeightedDegreeStringKernel.h.

bool set_max_mismatch ( int32_t  max  ) 

set maximum mismatch

Parameters:
max new maximum mismatch
Returns:
if setting was successful

Definition at line 967 of file WeightedDegreeStringKernel.cpp.

bool set_mkl_stepsize ( int32_t  step  ) 

set MKL steps ize

Parameters:
step new step size
Returns:
if setting was successful

Definition at line 573 of file WeightedDegreeStringKernel.h.

virtual bool set_normalizer ( CKernelNormalizer *  normalizer_  )  [virtual]

set the current kernel normalizer

Returns:
if successful

Reimplemented from CKernel.

Definition at line 345 of file WeightedDegreeStringKernel.h.

bool set_position_weights ( float64_t *  pws,
int32_t  len 
)

set position weights

Parameters:
pws new position weights
len number of position weights
Returns:
if setting was successful

Definition at line 655 of file WeightedDegreeStringKernel.cpp.

void set_subkernel_weights ( float64_t *  weights2,
int32_t  num_weights2 
) [virtual]

set subkernel weights

Parameters:
weights2 weights
num_weights2 number of weights

Reimplemented from CKernel.

Definition at line 295 of file WeightedDegreeStringKernel.h.

bool set_use_block_computation ( bool  block  ) 

set if block computation shall be performed

Parameters:
block if block computation shall be performed
Returns:
if setting was successful

Definition at line 556 of file WeightedDegreeStringKernel.h.

void set_wd_weights ( SGVector< float64_t >  new_weights  ) 

set wd weights

Parameters:
new_weights new weights

Definition at line 441 of file WeightedDegreeStringKernel.h.

bool set_wd_weights_by_type ( EWDKernType  type  ) 

set wd weights

Parameters:
type weighted degree kernel type
Returns:
if setting was successful

if we know a better weighting later on do a switch

Definition at line 572 of file WeightedDegreeStringKernel.cpp.

bool set_weights ( SGMatrix< float64_t >  new_weights  ) 

set weights

Parameters:
new_weights new weights

Definition at line 625 of file WeightedDegreeStringKernel.cpp.

bool set_which_degree ( int32_t  which  ) 

set which degree

Parameters:
which which degree
Returns:
if setting was successful

Definition at line 592 of file WeightedDegreeStringKernel.h.

Member Data Documentation

CAlphabet* alphabet [protected]

alphabet of features

Definition at line 758 of file WeightedDegreeStringKernel.h.

bool block_computation [protected]

if block computation is used

Definition at line 742 of file WeightedDegreeStringKernel.h.

float64_t* block_weights [protected]

(internal) block weights

Definition at line 745 of file WeightedDegreeStringKernel.h.

int32_t degree [protected]

degree

Definition at line 729 of file WeightedDegreeStringKernel.h.

bool initialized [protected]

if kernel is initialized

Definition at line 739 of file WeightedDegreeStringKernel.h.

int32_t length [protected]

length

Definition at line 731 of file WeightedDegreeStringKernel.h.

int32_t max_mismatch [protected]

maximum mismatch

Definition at line 734 of file WeightedDegreeStringKernel.h.

int32_t mkl_stepsize [protected]

MKL step size

Definition at line 727 of file WeightedDegreeStringKernel.h.

float64_t* position_weights [protected]

position weights

Definition at line 721 of file WeightedDegreeStringKernel.h.

int32_t position_weights_len [protected]

position weights

Definition at line 723 of file WeightedDegreeStringKernel.h.

int32_t seq_length [protected]

sequence length

Definition at line 736 of file WeightedDegreeStringKernel.h.

bool tree_initialized [protected]

if tree is initialized

Definition at line 755 of file WeightedDegreeStringKernel.h.

CTrie<DNATrie>* tries [protected]

tries

Definition at line 752 of file WeightedDegreeStringKernel.h.

EWDKernType type [protected]

WeightedDegree kernel type

Definition at line 747 of file WeightedDegreeStringKernel.h.

float64_t* weights [protected]

degree*length weights length must match seq_length if != 0

Definition at line 713 of file WeightedDegreeStringKernel.h.

float64_t* weights_buffer [protected]

weights buffer

Definition at line 725 of file WeightedDegreeStringKernel.h.

int32_t weights_degree [protected]

degree

Definition at line 715 of file WeightedDegreeStringKernel.h.

int32_t weights_length [protected]

length

Definition at line 717 of file WeightedDegreeStringKernel.h.

int32_t which_degree [protected]

which degree

Definition at line 749 of file WeightedDegreeStringKernel.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