ired::AssignmentIterator< FeatureNetwork > Class Template Reference

Iterate over the assignments of a subset of variables. More...

#include <assignment.hpp>

Public Types
using	assignment_t = Assignment
using	var_idx_t = assignment_t::var_idx_t
using	var_value_t = assignment_t::var_value_t
using	ep = typename FeatureNetwork::evaluation_policy_t
using	fun_value_t = typename ep::fun_value_t
using	function_t = typename ep::function_t
using	constraint_t = typename ep::constraint_t

Public Member Functions
	AssignmentIterator (assignment_t &a, const std::vector< var_idx_t > &vars, const FeatureNetwork &cn, const std::vector< const constraint_t * > &constraints, const std::vector< const function_t * > &functions, const fun_value_t &initial_value)
	constructor
void	reset ()
	Reset iterator.
auto	value ()
	get evaluation according to functions
void	register_finish_stage2_hook (int stage1_size, const std::function< void()> &hook)
	register hook to run after finishing stage 2 enumeration
const auto &	operator++ ()
	Next valid partial assignment.
bool	finished ()
	Check for termination.

Detailed Description

template<class FeatureNetwork>
class ired::AssignmentIterator< FeatureNetwork >

Iterate over the assignments of a subset of variables.

Note

the assignment iterator works destructively on parent assignment!

The job of the assignment iterator goes beyond pure enumeration of the sub-assignments that are valid due to the domains of a subset of variables. It also checks validity due to constraints and functions and evaluates the functions.

Constraints and functions are evaluated as soon as all their variables are determined. For this purpose, the iterator precomputes tables ('boards'). Function values are combined via EvaluationPolicy::mul; partial products are tracked on the enumeration stack.

Member Typedef Documentation

assignment_t

template<class FeatureNetwork >

using ired::AssignmentIterator< FeatureNetwork >::assignment_t = Assignment

constraint_t

template<class FeatureNetwork >

using ired::AssignmentIterator< FeatureNetwork >::constraint_t = typename ep::constraint_t

ep

template<class FeatureNetwork >

using ired::AssignmentIterator< FeatureNetwork >::ep = typename FeatureNetwork::evaluation_policy_t

fun_value_t

template<class FeatureNetwork >

using ired::AssignmentIterator< FeatureNetwork >::fun_value_t = typename ep::fun_value_t

function_t

template<class FeatureNetwork >

using ired::AssignmentIterator< FeatureNetwork >::function_t = typename ep::function_t

var_idx_t

template<class FeatureNetwork >

using ired::AssignmentIterator< FeatureNetwork >::var_idx_t = assignment_t::var_idx_t

var_value_t

template<class FeatureNetwork >

using ired::AssignmentIterator< FeatureNetwork >::var_value_t = assignment_t::var_value_t

Constructor & Destructor Documentation

AssignmentIterator()

template<class FeatureNetwork >

ired::AssignmentIterator< FeatureNetwork >::AssignmentIterator ( assignment_t &  a, const std::vector< var_idx_t > &  vars, const FeatureNetwork &  cn, const std::vector< const constraint_t * > &  constraints, const std::vector< const function_t * > &  functions, const fun_value_t &  initial_value  )

inline

constructor

Parameters

vars	vector of indices of the to-be-enumerated variables
cn	constraint network, where everything takes place
constraints	check assignments
functions	check and evaluate assignments

During the iteration, constraints and functions are evaluated. We evaluate only those constraints/functions that have at least on undetermined dependency on a variable in vars, where all dependencies outside of vars are already determined.

After construction with non-empty vars, the assignment is either set to the first valid assignment or finished() is true. After construction with empty vars, assignment is unchanged and finished() is false, but will be true immediately after a call to operator ++()!

Member Function Documentation

finished()

template<class FeatureNetwork >

bool ired::AssignmentIterator< FeatureNetwork >::finished ( )

inline

Check for termination.

operator++()

template<class FeatureNetwork >

const auto & ired::AssignmentIterator< FeatureNetwork >::operator++ ( )

inline

Next valid partial assignment.

Sets the assignment to the next valid assignment of vars_

register_finish_stage2_hook()

template<class FeatureNetwork >

void ired::AssignmentIterator< FeatureNetwork >::register_finish_stage2_hook ( int  stage1_size, const std::function< void()> &  hook  )

inline

register hook to run after finishing stage 2 enumeration

Parameters

stage1_size	number of variables in stage 1
hook	function to be called after finishing stage 2 enumeration

This is used in two-stage enumeration over two disjoint sets of variables; in the evaluation of a CN, this allows to marginalize over the second set and store the results as value of a function (the 'message') of the variables in the first set.

reset()

template<class FeatureNetwork >

void ired::AssignmentIterator< FeatureNetwork >::reset ( )

inline

Reset iterator.

value()

template<class FeatureNetwork >

auto ired::AssignmentIterator< FeatureNetwork >::value ( )

inline

get evaluation according to functions

after initialization or a call of operator++, this yields the product of all functions on enumerated variables, where at least one variable is determined only by the enumeration

---

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

• src/infrared/assignment.hpp