CSA::Solver< Scalar_x, Scalar_fx > Class Template Reference

Public Member Functions
	Solver ()
int	minimize (int n, Scalar_x *x, Scalar_fx(*fx)(void *, Scalar_x *), void(*step)(void *, Scalar_x *y, const Scalar_x *, float tgen), void(*progress)(void *, Scalar_fx cost, float tgen, float tacc, int opt_id, int iter), void *instance)

Public Attributes
int	m = 4
int	max_iterations = 1000000
float	tgen_initial = 0.01
float	tgen_schedule = 0.99999
float	tacc_initial = 0.9
float	tacc_schedule = 0.01
float	desired_variance = 0.99

Constructor & Destructor Documentation

Solver()

template<typename Scalar_x , typename Scalar_fx >

CSA::Solver< Scalar_x, Scalar_fx >::Solver ( )

inline

Default constructor.

Member Function Documentation

minimize()

template<typename Scalar_x , typename Scalar_fx >

int CSA::Solver< Scalar_x, Scalar_fx >::minimize ( int  n, Scalar_x *  x, Scalar_fx(*)(void *, Scalar_x *)  fx, void(*)(void *, Scalar_x *y, const Scalar_x *, float tgen)  step, void(*)(void *, Scalar_fx cost, float tgen, float tacc, int opt_id, int iter)  progress, void *  instance  )

inline

Run the CSA process to minimize the target function.

Parameters

n	The size of the input array x.
x	The input array, representing an initial guess of the solution. This array will be modified with the best solution found.
fx	The function to minimize. The first argument to the function is a pointer (possibly NULL) to the callback object instance.
step	The step function. Should populate the array y with a a random step based on the current position x and the generation schedule tgen. Like fx the first argument to the function should be a pointer to the callback object instance.
progress	An optional function that receives updates when a new best solution is found. Like fx and step, the first argument is a pointer to the callback object instance. The next arguments, in order, are the current cost, the current generation temperature, the current acceptance temperature, the thread ID, and the iteration number.
instance	An optional pointer to a callback object.

Warning

By default, the iterations of the main loop are not processed monotonically by the threads, so the iter parameter that the progress function receives will appear to be random. This behavior can be changed by setting the macro CSA_ITER_MONO to a non-zero value. However, this may result in a significant slow-down.

Member Data Documentation

desired_variance

template<typename Scalar_x , typename Scalar_fx >

float CSA::Solver< Scalar_x, Scalar_fx >::desired_variance = 0.99

The desired variance of the acceptance probabilities.

m

template<typename Scalar_x , typename Scalar_fx >

int CSA::Solver< Scalar_x, Scalar_fx >::m = 4

The number of threads and coupled annealing processes.

max_iterations

template<typename Scalar_x , typename Scalar_fx >

int CSA::Solver< Scalar_x, Scalar_fx >::max_iterations = 1000000

The maximum number of iterations/steps.

tacc_initial

template<typename Scalar_x , typename Scalar_fx >

float CSA::Solver< Scalar_x, Scalar_fx >::tacc_initial = 0.9

The initial value of the acceptance temperature.

tacc_schedule

template<typename Scalar_x , typename Scalar_fx >

float CSA::Solver< Scalar_x, Scalar_fx >::tacc_schedule = 0.01

Determines the factor by which tacc is increased or decreased during each update.

tgen_initial

template<typename Scalar_x , typename Scalar_fx >

float CSA::Solver< Scalar_x, Scalar_fx >::tgen_initial = 0.01

The initial value of the generation temperature.

tgen_schedule

template<typename Scalar_x , typename Scalar_fx >

float CSA::Solver< Scalar_x, Scalar_fx >::tgen_schedule = 0.99999

Determines the factor that tgen is multiplied by during each update.

---

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

• csa.hpp