Detailed Description

The Weighted Degree Position String kernel (Weighted Degree kernel with shifts).

The WD-shift 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 $\beta_k$ allowing for a positional tolerance of up to shift s.

It is formally defined as

$\begin{eqnarray*} &&\!\!\!\!\!\!\!k({\bf x}_i,{\bf x}_j)=\sum_{k=1}^d\beta_k\sum_{l=1}^{\!\!\!\!L-k+1\!\!\!\!}\gamma_l\sum_{\begin{array}{c}s=0\\ \!\!\!\!s+l\leq L\!\!\!\!\end{array}}^{S(l)} \delta_s\;\mu_{k,l,s,{{\bf x}_i},{{\bf x}_j}},\\ &&\!\!\!\!\!\!\!\!\!\! {\footnotesize \mu_{k,l,s,{{\bf x}_i},{{\bf x}_j}}\!\!\! =\!\! I({\bf u}_{k,l+s}({\bf x}_i)\! =\!{\bf u}_{k,l}({\bf x}_j))\! +\!I({\bf u}_{k,l}({\bf x}_i)\! =\!{\bf u}_{k,l+s}({\bf x}_j))},\nonumber \end{eqnarray*}$

where $\beta_j$ are the weighting coefficients of the j-mers, $\gamma_l$ is a weighting over the position in the sequence, $\delta_s=1/(2(s+1))$ is the weight assigned to shifts (in either direction) of extent s, and S(l) determines the shift range at position l.

Definition at line 48 of file WeightedDegreePositionStringKernel.h.

Inheritance diagram for CWeightedDegreePositionStringKernel:

[legend]

List of all members.

Public Member Functions
	CWeightedDegreePositionStringKernel (void)
	CWeightedDegreePositionStringKernel (int32_t size, int32_t degree, int32_t max_mismatch=0, int32_t mkl_stepsize=1)
	CWeightedDegreePositionStringKernel (int32_t size, float64_t weights, int32_t degree, int32_t max_mismatch, int32_t shift, int32_t shift_len, int32_t mkl_stepsize=1)
	CWeightedDegreePositionStringKernel (CStringFeatures< char > l, CStringFeatures< char > r, int32_t degree)
virtual	~CWeightedDegreePositionStringKernel ()
virtual bool	init (CFeatures l, CFeatures r)
virtual void	cleanup ()
virtual EKernelType	get_kernel_type ()
virtual const char *	get_name () const
virtual bool	init_optimization (int32_t p_count, int32_t IDX, float64_t alphas)
virtual bool	init_optimization (int32_t count, int32_t IDX, float64_t alphas, int32_t tree_num, int32_t upto_tree=-1)
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)
float64_t *	compute_abs_weights (int32_t &len)
bool	is_tree_initialized ()
int32_t	get_max_mismatch ()
int32_t	get_degree ()
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_shifts (int32_t *shifts, int32_t len)
virtual bool	set_weights (float64_t *weights, int32_t d, int32_t len=0)
virtual bool	set_wd_weights ()
virtual bool	set_position_weights (float64_t *pws, int32_t len)
bool	set_position_weights_lhs (float64_t *pws, int32_t len, int32_t num)
bool	set_position_weights_rhs (float64_t *pws, int32_t len, int32_t num)
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	delete_position_weights_lhs ()
bool	delete_position_weights_rhs ()
virtual float64_t	compute_by_tree (int32_t idx)
virtual void	compute_by_tree (int32_t idx, float64_t *LevelContrib)
float64_t *	compute_scoring (int32_t max_degree, int32_t &num_feat, int32_t &num_sym, float64_t target, int32_t num_suppvec, int32_t IDX, float64_t *weights)
char *	compute_consensus (int32_t &num_feat, int32_t num_suppvec, int32_t IDX, float64_t alphas)
float64_t *	extract_w (int32_t max_degree, int32_t &num_feat, int32_t &num_sym, float64_t w_result, int32_t num_suppvec, int32_t IDX, float64_t *alphas)
float64_t *	compute_POIM (int32_t max_degree, int32_t &num_feat, int32_t &num_sym, float64_t poim_result, int32_t num_suppvec, int32_t IDX, float64_t alphas, float64_t distrib)
void	prepare_POIM2 (float64_t *distrib, int32_t num_sym, int32_t num_feat)
void	compute_POIM2 (int32_t max_degree, CSVM *svm)
void	get_POIM2 (float64_t *poim, int32_t result_len)
void	cleanup_POIM2 ()
	cleanup POIM2
Static Public Member Functions
static void *	compute_batch_helper (void *p)
Protected Member Functions
void	create_empty_tries ()
virtual 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)
virtual 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_without_mismatch_position_weights (char avec, float64_t posweights_lhs, int32_t alen, char bvec, float64_t posweights_rhs, int32_t blen)
virtual void	remove_lhs ()
virtual void	load_serializable_post (void) throw (ShogunException)
Protected Attributes
float64_t *	weights
int32_t	weights_degree
int32_t	weights_length
float64_t *	position_weights
int32_t	position_weights_len
float64_t *	position_weights_lhs
int32_t	position_weights_lhs_len
float64_t *	position_weights_rhs
int32_t	position_weights_rhs_len
bool *	position_mask
float64_t *	weights_buffer
int32_t	mkl_stepsize
int32_t	degree
int32_t	length
int32_t	max_mismatch
int32_t	seq_length
int32_t *	shift
int32_t	shift_len
int32_t	max_shift
bool	block_computation
float64_t *	block_weights
EWDKernType	type
int32_t	which_degree
CTrie< DNATrie >	tries
CTrie< POIMTrie >	poim_tries
bool	tree_initialized
bool	use_poim_tries
float64_t *	m_poim_distrib
float64_t *	m_poim
int32_t	m_poim_num_sym
int32_t	m_poim_num_feat
int32_t	m_poim_result_len
CAlphabet *	alphabet

Constructor & Destructor Documentation

CWeightedDegreePositionStringKernel ( void )

default constructor

Definition at line 52 of file WeightedDegreePositionStringKernel.cpp.

CWeightedDegreePositionStringKernel	(	int32_t	size,
		int32_t	degree,
		int32_t	max_mismatch = `0`,
		int32_t	mkl_stepsize = `1`
	)

constructor

Parameters:

size	cache size
degree	degree
max_mismatch	maximum mismatch
mkl_stepsize	MKL stepsize

Definition at line 59 of file WeightedDegreePositionStringKernel.cpp.

CWeightedDegreePositionStringKernel	(	int32_t	size,
		float64_t *	weights,
		int32_t	degree,
		int32_t	max_mismatch,
		int32_t *	shift,
		int32_t	shift_len,
		int32_t	mkl_stepsize = `1`
	)

constructor

Parameters:

size	cache size
weights	weights
degree	degree
max_mismatch	maximum mismatch
shift	position shifts
shift_len	number of shifts
mkl_stepsize	MKL stepsize

Definition at line 76 of file WeightedDegreePositionStringKernel.cpp.

CWeightedDegreePositionStringKernel	(	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 100 of file WeightedDegreePositionStringKernel.cpp.

~CWeightedDegreePositionStringKernel ( ) [virtual]

Definition at line 119 of file WeightedDegreePositionStringKernel.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 695 of file WeightedDegreePositionStringKernel.cpp.

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

add example to tree

Parameters:

idx	index
weight	weight

Definition at line 644 of file WeightedDegreePositionStringKernel.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 211 of file WeightedDegreePositionStringKernel.h.

void cleanup ( ) [virtual]

clean up kernel

Reimplemented from CKernel.

Definition at line 227 of file WeightedDegreePositionStringKernel.cpp.

void cleanup_POIM2 ( )

cleanup POIM2

Definition at line 1880 of file WeightedDegreePositionStringKernel.cpp.

virtual void clear_normal ( ) [virtual]

clear normal subkernel functionality

Reimplemented from CKernel.

Definition at line 185 of file WeightedDegreePositionStringKernel.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 608 of file WeightedDegreePositionStringKernel.cpp.

float64_t * compute_abs_weights ( int32_t & len )

compute abs weights

Parameters:

len len

Returns:: computed abs weights

Definition at line 831 of file WeightedDegreePositionStringKernel.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 1239 of file WeightedDegreePositionStringKernel.cpp.

void * compute_batch_helper ( void * p ) [static]

helper for compute batch

Parameters:

p	thread parameter

Definition at line 1181 of file WeightedDegreePositionStringKernel.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 235 of file WeightedDegreePositionStringKernel.h.

float64_t compute_by_tree ( int32_t idx ) [virtual]

compute by tree

Parameters:

idx index

Returns:: computed value

Definition at line 749 of file WeightedDegreePositionStringKernel.cpp.

void compute_by_tree	(	int32_t	idx,
		float64_t *	LevelContrib
	)		`[virtual]`

compute by tree

Parameters:

idx	index
LevelContrib	level contribution

Definition at line 788 of file WeightedDegreePositionStringKernel.cpp.

char * compute_consensus	(	int32_t &	num_feat,
		int32_t	num_suppvec,
		int32_t *	IDX,
		float64_t *	alphas
	)

compute consensus string

Parameters:

num_feat	number of features
num_suppvec	number of support vectors
IDX	IDX
alphas	alphas

Returns:: consensus string

Definition at line 1508 of file WeightedDegreePositionStringKernel.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 153 of file WeightedDegreePositionStringKernel.h.

float64_t * compute_POIM	(	int32_t	max_degree,
		int32_t &	num_feat,
		int32_t &	num_sym,
		float64_t *	poim_result,
		int32_t	num_suppvec,
		int32_t *	IDX,
		float64_t *	alphas,
		float64_t *	distrib
	)

compute POIM

Parameters:

max_degree	maximum degree
num_feat	number of features
num_sym	number of symbols
poim_result	poim
num_suppvec	number of support vectors
IDX	IDX
alphas	alphas
distrib	distribution

Returns:: computed POIMs

Definition at line 1688 of file WeightedDegreePositionStringKernel.cpp.

void compute_POIM2	(	int32_t	max_degree,
		CSVM *	svm
	)

compute POIM2

Parameters:

max_degree	maximum degree
svm	SVM

Definition at line 1836 of file WeightedDegreePositionStringKernel.cpp.

float64_t * compute_scoring	(	int32_t	max_degree,
		int32_t &	num_feat,
		int32_t &	num_sym,
		float64_t *	target,
		int32_t	num_suppvec,
		int32_t *	IDX,
		float64_t *	weights
	)

compute positional scoring function, which assigns a weight per position, per symbol in the sequence

Parameters:

max_degree	maximum degree
num_feat	number of features
num_sym	number of symbols
target	target
num_suppvec	number of support vectors
IDX	IDX
weights	weights

Returns:: computed scores

Definition at line 1352 of file WeightedDegreePositionStringKernel.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 318 of file WeightedDegreePositionStringKernel.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 398 of file WeightedDegreePositionStringKernel.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 464 of file WeightedDegreePositionStringKernel.cpp.

float64_t compute_without_mismatch_position_weights	(	char *	avec,
		float64_t *	posweights_lhs,
		int32_t	alen,
		char *	bvec,
		float64_t *	posweights_rhs,
		int32_t	blen
	)		`[protected]`

compute without mismatch position weights

Parameters:

avec	vector a
posweights_lhs	position weights left-hand side
alen	length of vector a
bvec	vector b
posweights_rhs	position weights right-hand side
blen	length of vector b

Returns:: computed value

Definition at line 528 of file WeightedDegreePositionStringKernel.cpp.

void create_empty_tries ( ) [protected]

create emtpy tries

Definition at line 159 of file WeightedDegreePositionStringKernel.cpp.

bool delete_optimization ( ) [virtual]

delete optimization

Returns:: if deleting was successful

Reimplemented from CKernel.

Definition at line 293 of file WeightedDegreePositionStringKernel.cpp.

bool delete_position_weights ( )

delete position weights

Returns:: if deleting was successful

Definition at line 477 of file WeightedDegreePositionStringKernel.h.

bool delete_position_weights_lhs ( )

delete position weights left-hand side

Returns:: if deleting was successful

Definition at line 488 of file WeightedDegreePositionStringKernel.h.

bool delete_position_weights_rhs ( )

delete position weights right-hand side

Returns:: if deleting was successful

Definition at line 499 of file WeightedDegreePositionStringKernel.h.

float64_t * extract_w	(	int32_t	max_degree,
		int32_t &	num_feat,
		int32_t &	num_sym,
		float64_t *	w_result,
		int32_t	num_suppvec,
		int32_t *	IDX,
		float64_t *	alphas
	)

extract w

Parameters:

max_degree	maximum degree
num_feat	number of features
num_sym	number of symbols
w_result	w
num_suppvec	number of support vectors
IDX	IDX
alphas	alphas

Returns:: w

Definition at line 1625 of file WeightedDegreePositionStringKernel.cpp.

int32_t get_degree ( )

get degree

Returns:: the degree

Definition at line 328 of file WeightedDegreePositionStringKernel.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 335 of file WeightedDegreePositionStringKernel.h.

virtual EKernelType get_kernel_type ( ) [virtual]

return what type of kernel we are

Returns:: kernel type WEIGHTEDDEGREEPOS

Implements CStringKernel< char >.

Definition at line 106 of file WeightedDegreePositionStringKernel.h.

int32_t get_max_mismatch ( )

get maximum mismatch

Returns:: maximum mismatch

Definition at line 322 of file WeightedDegreePositionStringKernel.h.

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

return the kernel's name

Returns:: name WeightedDegreePos

Reimplemented from CStringKernel< char >.

Definition at line 112 of file WeightedDegreePositionStringKernel.h.

virtual int32_t get_num_subkernels ( ) [virtual]

get number of subkernels

Returns:: number of subkernels

Reimplemented from CKernel.

Definition at line 221 of file WeightedDegreePositionStringKernel.h.

void get_POIM2	(	float64_t **	poim,
		int32_t *	result_len
	)

get POIM2

Parameters:

poim	POIMs (returned)
result_len	(returned)

Definition at line 1871 of file WeightedDegreePositionStringKernel.cpp.

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 366 of file WeightedDegreePositionStringKernel.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 252 of file WeightedDegreePositionStringKernel.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 347 of file WeightedDegreePositionStringKernel.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 178 of file WeightedDegreePositionStringKernel.cpp.

bool init_block_weights ( )

initialize block weights

Returns:: if initialization was successful

Definition at line 1155 of file WeightedDegreePositionStringKernel.cpp.

bool init_block_weights_const ( )

initialize block weights constant

Returns:: if initialization was successful

Definition at line 1059 of file WeightedDegreePositionStringKernel.cpp.

bool init_block_weights_cubicpoly ( )

initialize block weights cubic polynomial

Returns:: if initialization was successful

Definition at line 1104 of file WeightedDegreePositionStringKernel.cpp.

bool init_block_weights_exp ( )

initialize block weights exponential

Returns:: if initialization was successful

Definition at line 1121 of file WeightedDegreePositionStringKernel.cpp.

bool init_block_weights_from_wd ( )

initialize block weights from weighted degree

Returns:: if initialization was successful

Definition at line 1011 of file WeightedDegreePositionStringKernel.cpp.

bool init_block_weights_from_wd_external ( )

initialize block weights from external weighted degree

Returns:: if initialization was successful

Definition at line 1031 of file WeightedDegreePositionStringKernel.cpp.

bool init_block_weights_linear ( )

initialize block weights linear

Returns:: if initialization was successful

Definition at line 1073 of file WeightedDegreePositionStringKernel.cpp.

bool init_block_weights_log ( )

initialize block weights logarithmic

Returns:: if initialization was successful

Definition at line 1138 of file WeightedDegreePositionStringKernel.cpp.

bool init_block_weights_sqpoly ( )

initialize block weights squared polynomial

Returns:: if initialization was successful

Definition at line 1087 of file WeightedDegreePositionStringKernel.cpp.

bool init_optimization	(	int32_t	count,
		int32_t *	IDX,
		float64_t *	alphas,
		int32_t	tree_num,
		int32_t	upto_tree = `-1`
	)		`[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
upto_tree	up to this tree

Returns:: if initializing was successful

Definition at line 247 of file WeightedDegreePositionStringKernel.cpp.

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

initialize optimization

Parameters:

p_count	count
IDX	index
alphas	alphas

Returns:: if initializing was successful

Reimplemented from CKernel.

Definition at line 121 of file WeightedDegreePositionStringKernel.h.

bool is_tree_initialized ( )

check if tree is initialized

Returns:: if tree is initialized

Definition at line 316 of file WeightedDegreePositionStringKernel.h.

void load_serializable_post ( void ) throw (ShogunException) [protected, virtual]

Can (optionally) be overridden to post-initialize some member variables which are not PARAMETER::ADD'ed. Make sure that at first the overridden method BASE_CLASS::LOAD_SERIALIZABLE_POST is called.

Exceptions:

ShogunException Will be thrown if an error occurres.

Reimplemented from CKernel.

Definition at line 1891 of file WeightedDegreePositionStringKernel.cpp.