Generates local scores from discrete data using AD trees. More...

#include <stdio.h>
#include <stdlib.h>
#include <assert.h>
#include <math.h>
#include <limits.h>
#include <string.h>
#include "scoring.h"
#include "utils.h"
#include "versiongit.h"

Include dependency graph for scoring.c:

Data Structures
union	adcontab
	A contingency table. More...

struct	adtree
	represents data for a 'query', e.g. X1=1,X3=2 More...

struct	scored_parentset
	Parent set with its score Note that the child variable is not represented. More...

struct	subset_tree
	stores a set of downward closed subsets of the variables in a tree More...

struct	varynode
	for splitting data on values of a particular variable (variable not stored in varynode) More...

Macros
#define	BLOCKSIZE 10000

#define	MAXARITY UCHAR_MAX

#define	min(A, B) ((A) < (B) ? (A) : (B))

Typedefs
typedef union adcontab	ADCONTAB
	A contingency table.

typedef struct adtree	ADTREE
	An AD tree.

typedef unsigned char	ARITY
	Arity of a variable in the data.

typedef unsigned int	COUNT
	Count, typically of datapoints.

typedef unsigned int	ROW
	Index of a row (i.e. datapoint) in the data.

typedef double	SCORE
	A local score. More...

typedef double	SCOREARG
	An argument for a local score function. More...

typedef struct scored_parentset	SCORED_PARENTSET
	parent set with its local score

typedef struct subset_tree	SUBSET_TREE
	subset tree

typedef unsigned char	VALUE
	Value of a variable in the data.

typedef unsigned int	VARIABLE
	Variable in the data.

typedef struct varynode	VARYNODE
	A varynode (in an AD tree)

Functions
static void	add_subset (SUBSET_TREE tree, SUBSET_TREE bottom_ptr, VARIABLE *vars, int num_vars, VARIABLE offset, const VARIABLE nvars)
	Add a subset to the tree (if not already there ) More...

static SCIP_RETCODE	addGeneralDAGConstraints (SCIP scip, ParentSetData psd, int **is_parent)
	Reads in structural constraints to restrict which parent sets are considered. More...

static SCIP_RETCODE	addParentSets (SCIP scip, VARIABLE child, SCORED_PARENTSET parent_sets, int num_parent_sets, ParentSetData psd, SCIP_Real ***scores)
	Record local scores for parent sets of a particular child. More...

static void	build_adtree (ADTREE adtree, const VARIABLE variable, ROW theserows, COUNT count, const int rmin, const int depth, int n_nodes, const int adtreedepthlim, const int adtreenodeslim, const VARIABLE nvars, VALUE data, ARITY arity)
	Build an AD tree from (a subset of) the data. More...

static void	build_varynode (VARYNODE varynode, VARIABLE variable, ROW theserows, COUNT count, const int rmin, int depth, int n_nodes, const int adtreedepthlim, const int adtreenodeslim, VARIABLE nvars, VALUE data, ARITY arity)
	Build an vary node from (a subset of) the data. More...

static SCIP_Bool	check_subset (SUBSET_TREE tree, SUBSET_TREE bottom_ptr, VARIABLE *vars, int num_vars, VARIABLE offset)
	Check whether a particular subset of variables is stored in a tree of subsets. More...

static SCIP_RETCODE	ci_constraint (SCIP scip, ParentSetData psd, const char a_str, const char b_str, const char s_str, int *is_parent)
	Use a conditional independence (ci) constraint $A \perp B \| S$ to rule out parents. More...

static void	delete_adtree (ADTREE adtree, const VARIABLE variable, VARIABLE nvars, ARITY arity)
	Delete an AD tree. More...

static void	delete_contab (ADCONTAB adcontab, VARIABLE variables, int nvariables, const ARITY arity)
	Delete a contingency table. More...

static void	delete_tree (SUBSET_TREE tree, SUBSET_TREE bottom_ptr, VARIABLE length)
	Delete a tree of subsets. More...

static SCORE	log_likelihood (const ADCONTAB adcontab, VARIABLE variables, int nvariables, const SCOREARG aijk, const SCORE lgaijk, COUNT npos_cells_ptr, const ARITY arity, SCIP_Bool fast)
	Compute the (marginal) log-likelihood for a subset of the variables (modulo a constant) More...

static SCORE	log_likelihood_cache (SCIP scip, const ADTREE adtree, VARIABLE variables, int nvariables, COUNT npos_cells_ptr, const int nvarscachelimit, const int cachesizelimit, const int cacheblocksize, int cachesize_ptr, SCORE llh_cache_ptr, COUNT pos_cells_cache_ptr, const double alpha, const ARITY arity, SCIP_Bool fast, VALUE *data, int nsubsets)
	Compute the (marginal) log-likelihood for a subset of the variables and store in cache, or retrieve from cache if already computed. More...

static int	lookup (char *names, ARITY length, ARITY used, char *name)
	Find the integer encoding for (the string representation of) a value of a variable. More...

static void	makecontab (const ADTREE adtree, VARIABLE offset, VARIABLE variables, int nvariables, ADCONTAB adcontab, VALUE data, const ARITY arity)
	Construct a contingency table from an adtree. More...

static void	makecontableaf (const ROW leaflist, const COUNT count, VARIABLE variables, int nvariables, ADCONTAB adcontab, VALUE data, const ARITY arity)
	Construct a contingency table from a leaflist ( contingency table must be for at least one variable ) More...

static SCIP_RETCODE	parseSet (SCIP scip, ParentSetData psd, const char str, int set, int n_set, SCIP_Bool success)
	Parses a string to extract a set from it. More...

static SCIP_RETCODE	process_constraint (SCIP scip, ParentSetData psd, const char line, int *is_parent)
	Processes a constraint on the DAG structure. More...

static int	rank_subset (VARIABLE variables, int nvariables, int nsubsets)
	Map a subset of (BN) variables to a unique index. More...

static SCIP_RETCODE	readProblemFromFile (SCIP scip, const char filename, int num_delims, char delims, SCIP_Bool merge_delims, char nodeNames, ARITY arities, VALUE **items, int num_vars, int num_rows, char *labels, int num_observed)
	Read discrete data from a file. More...

SCIP_RETCODE	SC_addScoringParameters (SCIP *scip)
	Add scoring parameters. More...

SCIP_RETCODE	SC_readProblemInDataFormat (SCIP scip, const char filename, int num_delims, char delims, SCIP_Bool merge_delims, ParentSetData psd, SCIP_Real **scores, PropertyData prop)
	Read data in from a file, generate and store local scores. More...

static int	sps_sort (const void p1, const void p2)
	Compare two scored parent sets according to score. More...

static void	subtract_contab (ADCONTAB adcontab1, const ADCONTAB adcontab2, VARIABLE variables, int nvariables, const ARITY arity)
	Subtract one contingency table from another. More...

Detailed Description

Generates local scores from discrete data using AD trees.

Author: James Cussens; Mark Bartlett

Typedef Documentation

◆ SCORE

typedef double SCORE

A local score.

Currently only BDeu implemented

◆ SCOREARG

typedef double SCOREARG

An argument for a local score function.

Function Documentation

◆ add_subset()

static void add_subset	(	SUBSET_TREE *	tree,
		SUBSET_TREE *	bottom_ptr,
		VARIABLE *	vars,
		int	num_vars,
		VARIABLE	offset,
		const VARIABLE	nvars
	)

static

Add a subset to the tree (if not already there )

Parameters

tree	Set of subsets of variables
bottom_ptr	Indicates non-null leaf nodes
vars	The subset of variables to add
num_vars	The number of variables in the subset
offset	Offset for this node
nvars	Number of variables in the data

References build_varynode(), adtree::count, and scored_parentset::nvars.

Here is the call graph for this function:

◆ addGeneralDAGConstraints()

static SCIP_RETCODE addGeneralDAGConstraints	(	SCIP *	scip,
		ParentSetData *	psd,
		int **	is_parent
	)

static

Reads in structural constraints to restrict which parent sets are considered.

(Altered version of function of same name in probdata_bn.c)

Parameters

scip	The SCIP instance to add the constraint to.
psd	The parent set data relating to this problem, only nodeNames is set.
is_parent	is_parent[i][j] = 1 (-1) if j must (not) be a parent of i

Returns: SCIP_OKAY if the constraints could be successfully added.

References process_constraint().

Here is the call graph for this function:

◆ addParentSets()

static SCIP_RETCODE addParentSets	(	SCIP *	scip,
		VARIABLE	child,
		SCORED_PARENTSET **	parent_sets,
		int	num_parent_sets,
		ParentSetData *	psd,
		SCIP_Real ***	scores
	)

static

Record local scores for parent sets of a particular child.

Parameters

scip	SCIP data structure
child	Child variable whose scored parent sets are being added
parent_sets	Scored parent sets for child
num_parent_sets	The number of scored parent sets for child
psd	Parent set data to populate
scores	(*scores)[child][j] will be the score for the jth parent set of child

References ParentSetData::nParents, ParentSetData::nParentSets, scored_parentset::nvars, ParentSetData::ParentSets, scored_parentset::score, and scored_parentset::vars.

◆ build_adtree()

static void build_adtree	(	ADTREE *	adtree,
		const VARIABLE	variable,
		ROW *	theserows,
		COUNT	count,
		const int	rmin,
		const int	depth,
		int *	n_nodes,
		const int	adtreedepthlim,
		const int	adtreenodeslim,
		const VARIABLE	nvars,
		VALUE **	data,
		ARITY *	arity
	)

static

Build an AD tree from (a subset of) the data.

Parameters

adtree	pointer to ADTREE being built
variable	first variable to specialise further on, if variable=nvars then there is none
theserows	datapoint indices for this tree
count	number of datapoints for this tree
rmin	if count below this then create a leaflist
depth	the depth of this node
n_nodes	(pointer to) the number of nodes in the ADTREE
adtreedepthlim	limit on the depth of the ADTREE
adtreenodeslim	limit on the number of nodes in the ADTREE
nvars	Number of variables in the data
data	data[i][j] is the value of variable i in row j
arity	arity[i] is the arity of variable i,

References build_varynode(), adtree::children, adtree::count, adtree::leaflist, and scored_parentset::nvars.

Referenced by build_varynode().

Here is the call graph for this function:

◆ build_varynode()

static void build_varynode	(	VARYNODE *	varynode,
		VARIABLE	variable,
		ROW *	theserows,
		COUNT	count,
		int	rmin,
		const int	depth,
		int *	n_nodes,
		const int	adtreedepthlim,
		const int	adtreenodeslim,
		VARIABLE	nvars,
		VALUE **	data,
		ARITY *	arity
	)

static

Build an vary node from (a subset of) the data.

Parameters

varynode	varynode being built
variable	which variable is being split
theserows	datapoint indices to divide between values of variable
count	number of datapoints for this tree
depth	the depth of this node
n_nodes	(pointer to) the number of nodes in the ADTREE
adtreedepthlim	limit on the depth of the ADTREE
adtreenodeslim	limit on the number of nodes in the ADTREE
nvars	Number of variables in the data
data	data[i][j] is the value of variable i in row j
arity	arity[i] is the arity of variable i,

References build_adtree(), varynode::children, adtree::count, and varynode::mcv.

Referenced by add_subset(), and build_adtree().

Here is the call graph for this function:

◆ check_subset()

static SCIP_Bool check_subset	(	SUBSET_TREE *	tree,
		SUBSET_TREE *	bottom_ptr,
		VARIABLE *	vars,
		int	num_vars,
		VARIABLE	offset
	)

static

Check whether a particular subset of variables is stored in a tree of subsets.

Returns: True iff the subset is stored in the tree of subsets

Parameters

tree	Set of subsets of variables
bottom_ptr	Indicates non-null leaf nodes
vars	The subset of variables to look for
num_vars	The number of variables in the subset
offset	Offset for this node

◆ ci_constraint()

static SCIP_RETCODE ci_constraint	(	SCIP *	scip,
		ParentSetData *	psd,
		const char *	a_str,
		const char *	b_str,
		const char *	s_str,
		int **	is_parent
	)

static

Use a conditional independence (ci) constraint $A \perp B | S$ to rule out parents.

$ A, B$ and $ S $ are given as comma separated variable names

Parameters

scip	The SCIP instance in which to add the constraint.
psd	The parent set data for the problem.
a_str	A set
b_str	B set
s_str	S (separator) set. ( not used )
is_parent	is_parent[i][j] is set to 1 (-1) if j can(not) be a parent of i

Returns: SCIP_OKAY if the constraint was added successfully or an error code otherwise.

References parseSet().

Referenced by process_constraint().

Here is the call graph for this function:

◆ delete_adtree()

static void delete_adtree	(	ADTREE *	adtree,
		const VARIABLE	variable,
		VARIABLE	nvars,
		ARITY *	arity
	)

static

Delete an AD tree.

Parameters

adtree	pointer to ADTREE being deleted
variable	first variable to specialise further on, if variable=nvars then there is none
nvars	Number of variables in the data
arity	arity[i] is the arity of variable i,

References adtree::children, varynode::children, adtree::leaflist, and scored_parentset::nvars.

◆ delete_contab()

static void delete_contab	(	ADCONTAB	adcontab,
		VARIABLE *	variables,
		int	nvariables,
		const ARITY *	arity
	)

static

Delete a contingency table.

The contingency table must have at least one variable (i.e. it cannot be a count), but this only checked for in DEBUG mode.

Parameters

adcontab	Contingency table to delete
variables	Variables in the contingency table
nvariables	Number of variables in the contingency table (must be positive)
arity	arity[i] is the arity of variable i,

References adcontab::children.

Referenced by log_likelihood_cache().

◆ delete_tree()

static void delete_tree	(	SUBSET_TREE *	tree,
		SUBSET_TREE *	bottom_ptr,
		VARIABLE	length
	)

static

Delete a tree of subsets.

Parameters

tree	Tree of subsets to delete
bottom_ptr	Indicates non-null leaf nodes
length	Number of children of this node

◆ log_likelihood()

static SCORE log_likelihood	(	const ADCONTAB *	adcontab,
		VARIABLE *	variables,
		int	nvariables,
		const SCOREARG	aijk,
		const SCORE	lgaijk,
		COUNT *	npos_cells_ptr,
		const ARITY *	arity,
		SCIP_Bool	fast
	)

static

Compute the (marginal) log-likelihood for a subset of the variables (modulo a constant)

To get the true log-likelihood one would need to add lgamma(alpha) - log(alpha+N) to the result of this function, where N is the number of samples. Since local scores are always computed as this_function(family)-this_function(parents), there is no need to compute the constant term

Parameters

adcontab	contingency table for variables
variables	the subset of the variables
nvariables	the size of the subset of the variables
aijk	effective samples size (ESS) divided by the number of joint instantiations of the subset of variables
lgaijk	log(Gamma(aijk))
npos_cells_ptr	pointer to number of cells in adcontab with a positive count
arity	arity[i] is the arity of variable i,
fast	If true then C's lgamma function is used for computing scores, otherwise a slower, more accurate sum of logs

References adcontab::children, and adcontab::count.

Referenced by log_likelihood_cache().

◆ log_likelihood_cache()

static SCORE log_likelihood_cache	(	SCIP *	scip,
		const ADTREE *	adtree,
		VARIABLE *	variables,
		int	nvariables,
		COUNT *	npos_cells_ptr,
		const int	nvarscachelimit,
		const int	cachesizelimit,
		const int	cacheblocksize,
		int *	cachesize_ptr,
		SCORE **	llh_cache_ptr,
		COUNT **	pos_cells_cache_ptr,
		const double	alpha,
		const ARITY *	arity,
		SCIP_Bool	fast,
		VALUE **	data,
		int *	nsubsets
	)

static

Compute the (marginal) log-likelihood for a subset of the variables and store in cache, or retrieve from cache if already computed.

Parameters

scip	SCIP instance
adtree	the data
variables	the subset of the variables
nvariables	the size of the subset of the variables
npos_cells_ptr	pointer to number of cells in adcontab with a positive count
nvarscachelimit	subsets must have size below this to be cached.
cachesizelimit	the maximum number of log-likelihoods and pos_cell_cache values to cache (limit is common to both).
cacheblocksize	how much to increase the size of the cache for log-likelihoods and pos_cell_cache values when it is too small (common to both).
cachesize_ptr	(pointer to) the current size of the cache for log-likelihoods and pos_cell_cache values (common to both)
llh_cache_ptr	(pointer to) llh_cache[r] is the log-likelihood for (the data projected onto ) the unique subset of variables with rank r
pos_cells_cache_ptr	(pointer to) pos_cell_cache[r] is the number of non-zero counts in the contingency table for the unique subset of variables with rank r. A value of zero indicates that neither the correct count nor the the associated llh_cache[r] value has yet been computed
alpha	The 'effective sample size' governing the BDeu score
arity	arity[i] is the arity of variable i,
fast	If true then C's lgamma function is used for computing scores, otherwise a slower, more accurate sum of logs
data	data[i][j] is the value of variable i in row j
nsubsets	nsubsets[nvariables] is how many susbsets have size strictly less than nvariables

References delete_contab(), log_likelihood(), makecontab(), and rank_subset().

Here is the call graph for this function:

◆ lookup()

static int lookup	(	char **	names,
		ARITY	length,
		ARITY *	used,
		char *	name
	)

static

Find the integer encoding for (the string representation of) a value of a variable.

If the input string name does not occurs in names, then it will be added to it (as long as doing so would not exceed the arity), and (*used) will be incremented.

Returns: the integer encoding the value, or -1 not possible due to exceeding the given variable arity length.

Parameters

names	`names`[r] is (the string representation of) the rth value of the variable
length	arity of the variable (number of values the variable has)
used	the number of strings stored in `names`
name	(string representation of) a value of a variable

References BgeMatrixCreate(), and UT_readFileAndSplit().

Here is the call graph for this function:

◆ makecontab()

static void makecontab	(	const ADTREE *	adtree,
		VARIABLE	offset,
		VARIABLE *	variables,
		int	nvariables,
		ADCONTAB *	adcontab,
		VALUE **	data,
		const ARITY *	arity
	)

static

Construct a contingency table from an adtree.

Parameters

adtree	(Pointer to) the ADTREE or NULL
offset	Offset for first variable (to identify correct vary nodes)
variables	Variables in the sought contingency table (sorted)
nvariables	Number of variables in the contingency table
adcontab	Returned contingency table
data	data[i][j] is the value of variable i in row j
arity	arity[i] is the arity of variable i,

References adtree::children, varynode::children, adcontab::children, adtree::count, adcontab::count, adtree::leaflist, makecontableaf(), varynode::mcv, and subtract_contab().

Referenced by log_likelihood_cache().

Here is the call graph for this function:

◆ makecontableaf()

static void makecontableaf	(	const ROW *	leaflist,
		const COUNT	count,
		VARIABLE *	variables,
		int	nvariables,
		ADCONTAB *	adcontab,
		VALUE **	data,
		const ARITY *	arity
	)

static

Construct a contingency table from a leaflist ( contingency table must be for at least one variable )

Parameters

leaflist	datapoints for this query
count	number of datapoints (in leaflist)
variables	variables in the contingency table (sorted)
nvariables	number of variables in the contigency table
adcontab	(pointer to ) returned contingency table
data	data[i][j] is the value of variable i in row j
arity	arity[i] is the arity of variable i,

References adtree::children, adcontab::children, adtree::count, and adcontab::count.

Referenced by makecontab().

◆ parseSet()

static SCIP_RETCODE parseSet	(	SCIP *	scip,
		ParentSetData *	psd,
		const char *	str,
		int *	set,
		int *	n_set,
		SCIP_Bool *	success
	)

static

Parses a string to extract a set from it.

Copied from probdata_bn

Parameters

scip	The SCIP instance that this belongs to.
psd	The data set that the set refers to.
str	string to parse
set	result set
n_set	length of set
success	success flag

References get_index().

Referenced by ci_constraint().

Here is the call graph for this function:

◆ process_constraint()

static SCIP_RETCODE process_constraint	(	SCIP *	scip,
		ParentSetData *	psd,
		const char *	line,
		int **	is_parent
	)

static

Processes a constraint on the DAG structure.

Parameters

scip	The SCIP instance in which to add the constraint.
psd	The parent set data relating to this problem, only nodeNames is set.
line	The description of the constraint to add.
is_parent	is_parent[i][j] is set to 1 (-1) if j can(not) be a parent of i

Returns: SCIP_OKAY if the constraint was added or an error otherwise.

References ci_constraint(), get_index(), and ParentSetData::n.

Referenced by addGeneralDAGConstraints().

Here is the call graph for this function:

◆ rank_subset()

static int rank_subset	(	VARIABLE *	variables,
		int	nvariables,
		int *	nsubsets
	)

static

Map a subset of (BN) variables to a unique index.

If A is a subset of B then rank(A) < rank(B)

Returns: The rank of the subset

Parameters

variables	the subset of the variables
nvariables	the size of the subset of the variables
nsubsets	nsubsets[nvariables] is how many susbsets have size strictly less than nvariables

Referenced by log_likelihood_cache().

◆ readProblemFromFile()

static SCIP_RETCODE readProblemFromFile	(	SCIP *	scip,
		const char *	filename,
		int	num_delims,
		char *	delims,
		SCIP_Bool	merge_delims,
		char ***	nodeNames,
		ARITY **	arities,
		VALUE ***	items,
		int *	num_vars,
		int *	num_rows,
		char ****	labels,
		int **	num_observed
	)

static

Read discrete data from a file.

The last 7 arguments are outputs for which we use pointers. This function allocates space for these outputs.

Parameters

scip	SCIP data structure
filename	File containing the data
num_delims	The number of field delimiters to use
delims	The field delimiters to use
merge_delims	Whether multiple field delimiters should be merged in to one
nodeNames	(Pointer to) node (i.e. variable) names
arities	(Pointer to) node variable arities
items	(Pointer to) the data. &c (*items)[i][j] is the integer-encoded value of the ith variable in the jth datapoint
num_vars	(Pointer to) the number of variables
num_rows	(Pointer to) the number of rows (i.e. datapoints) in the data
labels	(Pointer to) a mapping from the integer encoding of a value to the corresponding string (*labels)[i][r] is the string corresponding to the rth value of the ith variable
num_observed	(Pointer to) the number of distinct values observed in the data for each variable

References UT_readFileAndSplit().

Here is the call graph for this function:

◆ SC_addScoringParameters()

SCIP_RETCODE SC_addScoringParameters ( SCIP * scip )

Add scoring parameters.

Parameters

scip	SCIP data structure

References UT_addBoolParam(), UT_addIntParam(), and UT_addRealParam().

Referenced by BN_addParameters().

Here is the call graph for this function:

◆ SC_readProblemInDataFormat()

SCIP_RETCODE SC_readProblemInDataFormat	(	SCIP *	scip,
		const char *	filename,
		int	num_delims,
		char *	delims,
		SCIP_Bool	merge_delims,
		ParentSetData *	psd,
		SCIP_Real ***	scores,
		PropertyData *	prop
	)

Read data in from a file, generate and store local scores.

This function sets the global variables alpha, palim, fast, cachesizelimit, cacheblocksize, nvarscachelimit, nvars, arity, nrows and data.

See also: alpha, palim, fast, cachesizelimit, cacheblocksize, nvarscachelimit, nvars, arity, nrows and data Detailed information about e.g. scoring parameters as well as about the data is stored in a PropertyData data structure

< The 'effective sample size' governing the BDeu score

< Maximum allowed size for a parent set

< If true then C's lgamma function is used for computing scores, otherwise a slower, more accurate sum of logs

< the maximum number of log-likelihoods and pos_cell_cache values to cache (limit is common to both).

< how much to increase the size of the cache for log-likelihoods and pos_cell_cache values when it is too small (common to both).

< subsets must have size below this to be cached.

< arity[i] is the arity of variable i, (discrete)

< data[i][j] is the value of variable i in row j (discrete)

< Number of variables in the data

< Number of rows (i.e. datapoints) in the data

< the current size of the cache for log-likelihoods and pos_cell_cache values (common to both)

< llh_cache[r] is the log-likelihood for (the data projected onto ) the unique subset of variables with rank r

< pos_cell_cache[r] is the number of non-zero counts in the contingency table for the unique subset of variables with rank r.

< nsubsets[nvariables] is how many susbsets have size strictly less than nvariables

< Subset tree used to indicate leaf nodes. Its value is never set or used; it only exists for bottom_ptr to have something to point to.

See also: bottom_ptr

< Pointer to bottom.

See also: bottom

Parameters

scip	SCIP data structure
filename	File containing the data
num_delims	The number of field delimiters to use
delims	The field delimiters to use
merge_delims	Whether multiple field delimiters should be merged in to one
psd	Parent set data to populate
scores	(*scores)[child][j] will be the score for the jth parent set of child
prop	Property data structure

References scored_parentset::nvars.

Referenced by readProblemInNonCIPFormat().

◆ sps_sort()

static int sps_sort	(	const void *	p1,
		const void *	p2
	)

static

Compare two scored parent sets according to score.

For use with qsort, so arguments have type void*

Returns: -1 if first has higher score, else 1

Parameters

p1	First scored parent set
p2	Second scored parent set

◆ subtract_contab()

static void subtract_contab	(	ADCONTAB *	adcontab1,
		const ADCONTAB *	adcontab2,
		VARIABLE *	variables,
		int	nvariables,
		const ARITY *	arity
	)

static

Subtract one contingency table from another.

The two contingency tables must be over the same variables

Parameters

adcontab1	Contingency table from which adcontab2 will be subtracted
adcontab2	Contingency table which will be subtracted from adcontab1
variables	Variables common to both adcontab1 and adcontab2
nvariables	Number of variables
arity	arity[i] is the arity of variable i,

References adcontab::children, adtree::count, and adcontab::count.

Referenced by makecontab().

Data Structures

Macros

Typedefs

Functions

Detailed Description

Typedef Documentation

◆ SCORE

◆ SCOREARG

Function Documentation

◆ add_subset()

◆ addGeneralDAGConstraints()

◆ addParentSets()

◆ build_adtree()

◆ build_varynode()

◆ check_subset()

◆ ci_constraint()

◆ delete_adtree()

◆ delete_contab()

◆ delete_tree()

◆ log_likelihood()

◆ log_likelihood_cache()

◆ lookup()

◆ makecontab()

◆ makecontableaf()

◆ parseSet()

◆ process_constraint()

◆ rank_subset()

◆ readProblemFromFile()

◆ SC_addScoringParameters()

◆ SC_readProblemInDataFormat()

◆ sps_sort()

◆ subtract_contab()