syn::regression Class Reference

#include <synthetic.hpp>

Inheritance diagram for syn::regression:

Public Types
enum	dist_type : char { one, dist_type::squared, dist_type::inf }

Public Member Functions
	regression (rdims dims, double *e_mat, double *centres, double *coefs, const std::string &n, double fmin, double *minloc)
virtual void	kernel (double *xs, double *ret)=0
double	evaluate (inst::set x) override
Public Member Functions inherited from syn::synthetic
	synthetic (const std::string &n, u_int dims, double lb, double ub, double opt)
	synthetic (const std::string &n, u_int dims, double opt)
	~synthetic ()
sspace::sspace_t *	get_search_space ()
void	set_properties (std::vector< properties > p)
std::vector< properties >	get_properties ()
inst::set	get_opt_param ()
double	get_opt ()
void	update_opt (double opt)
u_int	get_dims ()
void	validate_param_set (inst::set x)
Public Member Functions inherited from optk::benchmark
	benchmark (const std::string &name)
std::string	get_name ()

Protected Member Functions
void	dist_sq_1 (double *xs, double *ret)
void	dist_sq_2 (double *xs, double *ret)
void	dist_sq_inf (double *xs, double *ret)
Protected Member Functions inherited from syn::synthetic
void	set_opt_param (inst::set op)

Private Attributes
double *	m_e_mat
double *	m_centres
double *	m_coefs
rdims	m_dims

Additional Inherited Members
Protected Attributes inherited from syn::synthetic
u_int	m_dims
double	m_lb
double	m_ub
double	m_opt
std::vector< properties >	m_properties
inst::set	opt_params
sspace::sspace_t	m_sspace
Protected Attributes inherited from optk::benchmark
std::string	m_name

Detailed Description

This is the base class for implementing a set of functions devised by Michael McCourt working at SigOpt, (a commercial function optimisation service). The derived functions all fit into the framework of a linear combination of functions, many of which are positive definite kernels.

Member Enumeration Documentation

dist_type

enum syn::regression::dist_type : char

strong

Enumerator
squared	For calculating the l1 norm
inf	For calculating the l2 norm For claculating the infinity norm

Constructor & Destructor Documentation

regression()

syn::regression::regression	(	rdims	dims,
		double *	e_mat,
		double *	centres,
		double *	coefs,
		const std::string &	n,
		double	fmin,
		double *	minloc
	)

The constructor for the regression-type problems.

Parameters

dims	The number of dimensions of the problem
e_mat	TODO document this
centers	TODO document this
coefs	The regression coefficients
n	The benchmark name
fmin	The known global minimum.
minloc	The position of a/the global minimum

Member Function Documentation

dist_sq_1()

void syn::regression::dist_sq_1 ( double *  xs, double *  ret  )

protected

The dist_sq(1) function

Parameters

xs	The input parameters
ret	A double[dims.coef][dims.prob] array to return values.

dist_sq_2()

void syn::regression::dist_sq_2 ( double *  xs, double *  ret  )

protected

The dist_sq(2) function

Parameters

xs	The input parameters
ret	A double[dims.coef][dims.prob] array to return values.

dist_sq_inf()

void syn::regression::dist_sq_inf ( double *  xs, double *  ret  )

protected

The dist_sq(inf) function

Parameters

xs	The input parameters
ret	A double[dims.coef][dims.prob] array to return values.

evaluate()

double syn::regression::evaluate ( inst::set  x )

overridevirtual

Evaluate the benchmark (i.e. calculate the value of the objective fuinction) on the provided values.

Parameters

x	A vector of double-precision values to evaluate the benchmark at.

Todo:

Verify that a vector of doubles does not constrict the range of benchmark functions which may be evaluated (i.e. as opposed to categorical variables such as strings).

Implements optk::benchmark.

---

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

• includes/benchmarks/synthetic.hpp
• src/benchmarks/synthetic.cpp