galib::cuda::StandardGACUDA< GeneType > Class Template Reference

Generational GA with the same interface as StandardGA, built for CUDA-enabled builds. More...

#include <StandardGACUDA.h>

Inheritance diagram for galib::cuda::StandardGACUDA< GeneType >:

Public Member Functions
	StandardGACUDA (FitnessFunction< GeneType > &ff, std::unique_ptr< Selection< GeneType > > sel, std::unique_ptr< Mutation< GeneType > > mu, std::unique_ptr< Crossover< GeneType > > cs, double m_rate, double c_rate, std::size_t max_gen, bool elitism=true)
	Constructs the algorithm with all required operators and parameters.
void	run (Population< GeneType > &population) override
	Runs the evolutionary optimisation loop.
Public Member Functions inherited from galib::Algorithm< GeneType >
virtual	~Algorithm ()=default
virtual void	enableConsoleLogging (const std::size_t interval)
	Enables logging of progress to the standard output.
virtual void	enableFileLogging (const std::string &path, std::size_t interval)
	Enables logging of population snapshots to a CSV file.
virtual void	enableLogging (const std::string &)
	Placeholder for generic logging configuration.

Additional Inherited Members
Protected Member Functions inherited from galib::Algorithm< GeneType >
void	notifyLoggers (std::size_t gen, const Population< GeneType > &pop)
	Notifies all active loggers to record the current state.
Protected Attributes inherited from galib::Algorithm< GeneType >
std::unique_ptr< utils::ConsoleLogger< GeneType > >	console_logger_m
std::unique_ptr< utils::FileLogger< GeneType > >	file_logger_m

Detailed Description

template<typename GeneType = double>
class galib::cuda::StandardGACUDA< GeneType >

Generational GA with the same interface as StandardGA, built for CUDA-enabled builds.

Drop-in replacement for galib::StandardGA. The evolution loop (selection, crossover, mutation) runs on the CPU using OpenMP, making it fully compatible with any pluggable operator. When the library is compiled with CUDA support, the CUDA runtime is linked and GPU-accelerated fitness evaluation can be added in future iterations.

The algorithm follows a generational model:

1. Evaluate the initial population.
2. For each generation: optionally copy the elite individual, then fill the rest of the population via selection → crossover → mutation.
3. Evaluate the new population and repeat.

Template Parameters

GeneType

Numeric type of each gene (default: double).

Note

Prefer constructing through galib::utils::AlgorithmBuilder with algorithm.standard.use_cuda: true rather than instantiating directly.

Fitness is minimised — lower values are considered better.

Definition at line 45 of file StandardGACUDA.h.

Constructor & Destructor Documentation

StandardGACUDA()

template<typename GeneType = double>

galib::cuda::StandardGACUDA< GeneType >::StandardGACUDA ( FitnessFunction< GeneType > &  ff, std::unique_ptr< Selection< GeneType > >  sel, std::unique_ptr< Mutation< GeneType > >  mu, std::unique_ptr< Crossover< GeneType > >  cs, double  m_rate, double  c_rate, std::size_t  max_gen, bool  elitism = true  )

inline

Constructs the algorithm with all required operators and parameters.

Parameters

ff	Fitness function used to evaluate individuals. Must outlive this object.
sel	Selection operator (e.g. TournamentSelection). Ownership is transferred.
mu	Mutation operator (e.g. GaussianMutation). Ownership is transferred.
cs	Crossover operator (e.g. SinglePointCrossover). Ownership is transferred.
m_rate	Probability [0, 1] that any individual gene is mutated.
c_rate	Probability [0, 1] that two parents produce children via crossover; otherwise parents are copied unchanged.
max_gen	Number of generations to run.
elitism	If true, the best individual is always copied to the next generation.

Definition at line 60 of file StandardGACUDA.h.

Member Function Documentation

run()

template<typename GeneType = double>

void galib::cuda::StandardGACUDA< GeneType >::run ( Population< GeneType > &  population )

inlineoverridevirtual

Runs the evolutionary optimisation loop.

Evolves population in-place for max_gen generations. The population must be initialised with valid gene values before calling this method.

Progress is printed to stdout every 50 generations unless a ConsoleLogger is attached via enableConsoleLogging(), in which case that logger is used instead.

Parameters

population

In/out: the population to evolve. Modified in place.

Note

Population must not be empty.

Implements galib::Algorithm< GeneType >.

Definition at line 90 of file StandardGACUDA.h.

References galib::cuda::internal::CUDARunParams::crossover_rate, galib::cuda::internal::CUDARunParams::dims, galib::Population< GeneType >::empty(), galib::cuda::internal::CUDARunParams::fitness_callback, galib::Population< GeneType >::getNumGenes(), galib::cuda::internal::CUDARunParams::max_generations, galib::cuda::internal::CUDARunParams::mutation_rate, galib::cuda::internal::CUDARunParams::pop_size, galib::cuda::internal::CUDARunParams::problem_name, galib::cuda::internal::CUDARunParams::progress_callback, galib::cuda::internal::runCUDAEvolution(), galib::cuda::internal::CUDARunParams::sigma, galib::Population< GeneType >::size(), galib::cuda::internal::CUDARunParams::tournament_size, and galib::cuda::internal::CUDARunParams::use_elitism.

---

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

• include/algorithms/standard/StandardGACUDA.h