Class representing the variation operators in Brush. More...

#include <variation.h>

Collaboration diagram for Brush::Var::Variation< T >:

[legend]

Public Member Functions
	Variation ()=default
	Default constructor.

	Variation (Parameters &params, SearchSpace &ss, Dataset &d)
	Constructor that initializes the Variation object with parameters and search space.

	~Variation ()
	Destructor.

void	init ()
	Initializes the Variation object with parameters and search space.

std::optional< Individual< T > >	cross (const Individual< T > &mom, const Individual< T > &dad)
	Performs croearch_spaceover operation on two individuals.

std::optional< Individual< T > >	mutate (const Individual< T > &parent, string choice="")
	Performs mutation operation on an individual.

void	vary (Population< T > &pop, int island, const vector< size_t > &parents)
	Handles variation of a population.

void	update_ss ()

void	vary_and_update (Population< T > &pop, int island, const vector< size_t > &parents, const Dataset &data, Evaluation< T > &evaluator, bool do_simplification)
	Varies a population and updates the selection strategy based on rewards.

std::optional< Node >	bandit_sample_terminal (DataType R)

std::optional< Node >	bandit_get_node_like (Node node)

std::optional< Node >	bandit_sample_op_with_arg (DataType ret, DataType arg, int max_args=0)

std::optional< Node >	bandit_sample_op (DataType ret)

void	log_simplification_table (std::ofstream &log)

Public Attributes
SearchSpace	search_space

Dataset &	data

Parameters	parameters

Private Attributes
Bandit	variation_bandit

map< DataType, Bandit >	terminal_bandits

map< DataType, map< size_t, Bandit > >	op_bandits

Constants_simplifier	constants_simplifier

Inexact_simplifier	inexact_simplifier

Detailed Description

template<ProgramType T>
class Brush::Var::Variation< T >

Class representing the variation operators in Brush.

The Variation class is responsible for performing individual-level variations and handling the variation of a population in Brush. It contains methods for crossing individuals, mutating individuals, and varying a population.

Definition at line 44 of file variation.h.

Constructor & Destructor Documentation

◆ Variation() [1/2]

template<ProgramType T>

Brush::Var::Variation< T >::Variation ( )

default

Default constructor.

◆ Variation() [2/2]

template<ProgramType T>

Brush::Var::Variation< T >::Variation	(	Parameters &	params,
		SearchSpace &	ss,
		Dataset &	d )

inline

Constructor that initializes the Variation object with parameters and search space.

Parameters

params	The parameters for the variation operator.
ss	The search space for the variation operator.

Definition at line 57 of file variation.h.

Here is the call graph for this function:

◆ ~Variation()

template<ProgramType T>

Brush::Var::Variation< T >::~Variation ( )

inline

Destructor.

Definition at line 68 of file variation.h.

Member Function Documentation

◆ bandit_get_node_like()

template<ProgramType T>

std::optional< Node > Brush::Var::Variation< T >::bandit_get_node_like ( Node node )

inline

Definition at line 483 of file variation.h.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ bandit_sample_op()

template<ProgramType T>

std::optional< Node > Brush::Var::Variation< T >::bandit_sample_op ( DataType ret )

inline

Definition at line 559 of file variation.h.

◆ bandit_sample_op_with_arg()

template<ProgramType T>

std::optional< Node > Brush::Var::Variation< T >::bandit_sample_op_with_arg	(	DataType	ret,
		DataType	arg,
		int	max_args = 0 )

inline

Definition at line 516 of file variation.h.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ bandit_sample_terminal()

template<ProgramType T>

std::optional< Node > Brush::Var::Variation< T >::bandit_sample_terminal ( DataType R )

inline

Definition at line 458 of file variation.h.

Here is the caller graph for this function:

◆ cross()

template<Brush::ProgramType T>

std::optional< Individual< T > > Brush::Var::Variation< T >::cross	(	const Individual< T > &	mom,
		const Individual< T > &	dad )

Performs croearch_spaceover operation on two individuals.

Stochastically swaps subtrees between root and other, returning a new program.

Parameters

mom	The first parent individual.
dad	The second parent individual.

Returns: An optional containing the offspring individual if the crossover is successful, or an empty optional otherwise.

The spot where the cross will take place in the root parent is sampled based on attribute get_prob_change of each node in the tree. After selecting the cross spot, the program will iterate through the other parent searching for all compatible sub-trees to replace.

Due to the stochastic behavior, it may come to a case where there is no candidate to replace the spot node. In this case, the method returns std::nullopt (and has overloads so it can be used in a boolean context).

If the cross succeeds, the child program can be accessed through the .value() attribute of the std::optional. TODO: update this documentation (it doesnt take the program but the individual. also update mutation documentation) This means that, if you use the cross as auto opt = mutate(parent, SS), either opt==false or opt.value() contains the child.

Template Parameters

T	the program type

Parameters

root	the root parent
other	the donating parent

Returns: std::optional that may contain the child program of type T

Definition at line 416 of file variation.cpp.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ init()

template<ProgramType T>

void Brush::Var::Variation< T >::init ( )

inline

Initializes the Variation object with parameters and search space.

Parameters

params	The parameters for the variation operator.
ss	The search space for the variation operator.

Definition at line 76 of file variation.h.

Here is the caller graph for this function:

◆ log_simplification_table()

template<ProgramType T>

void Brush::Var::Variation< T >::log_simplification_table ( std::ofstream & log )

inline

Definition at line 581 of file variation.h.

◆ mutate()

template<Brush::ProgramType T>

std::optional< Individual< T > > Brush::Var::Variation< T >::mutate	(	const Individual< T > &	parent,
		string	choice = "" )

Performs mutation operation on an individual.

Stochastically mutate a program.

Parameters

parent The parent individual.

Returns: An optional containing the mutated individual if the mutation is successful, or an empty optional otherwise.

Types of mutation:

point mutation changes a single node.
insertion mutation inserts a node as the parent of an existing node, and fills in the other arguments.
deletion mutation deletes a node.
subtree mutation inserts a new subtree into the program.
toggle_weight_on mutation turns a node's weight ON.
toggle_weight_off mutation turns a node's weight OFF.

Every mutation has a probability (weight) based on global parameters. The spot where the mutation will take place is sampled based on attribute get_prob_change of each node in the tree. Inside each type of mutation, when a new node is inserted, it is sampled based on terminal_weights.

Due to the stochastic behavior, and the several sampling steps, it may come to a case where the search space does not hold any possible modification to do in the program. In this case, the method returns std::nullopt (and has overloads so it can be used in a boolean context).

If the mutation succeeds, the mutated program can be accessed through the .value() attribute of the std::optional. This method will try to generate a child that is within size and depth constraints. If ater three attempts it failed to generate a solution under these conditions, then the last model is returned regardless of it's size and depth. This loosens the size and depth constraints by a small factor (the final program will have at most 3*max_arity extra nodes), and it is better than the alternatives of just replicating the parent or generating a new random solution to add to the population.

This means that, if you use the mutation as auto opt = mutate(parent, SS), either opt==false or opt.value() contains the child program.

Template Parameters

T	program type

Parameters

parent	the program to be mutated
SS	a search space

Returns: std::optional that may contain the child program of type T

Definition at line 560 of file variation.cpp.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ update_ss()

template<Brush::ProgramType T>

void Brush::Var::Variation< T >::update_ss ( )

Updates the probability distribution sampling for variation and nodes based on the given rewards.

Parameters

pop	The population to update the selection strategy for.
rewards	The rewards obtained from the evaluation of individuals.

Definition at line 760 of file variation.cpp.

Here is the caller graph for this function:

◆ vary()

template<Brush::ProgramType T>

void Brush::Var::Variation< T >::vary	(	Population< T > &	pop,
		int	island,
		const vector< size_t > &	parents )

Handles variation of a population.

Parameters

pop	The population to be varied.
island	The island index.
parents	The indices of the parent individuals.
p	The parameters for the variation operator.

Definition at line 674 of file variation.cpp.

Here is the call graph for this function:

◆ vary_and_update()

template<ProgramType T>

void Brush::Var::Variation< T >::vary_and_update	(	Population< T > &	pop,
		int	island,
		const vector< size_t > &	parents,
		const Dataset &	data,
		Evaluation< T > &	evaluator,
		bool	do_simplification )

inline

Varies a population and updates the selection strategy based on rewards.

This function performs variation on a population, calculates rewards, and updates the selection strategy based on the obtained rewards.

Parameters

pop	The population to be varied and updated.
island	The island index.
parents	The indices of the parent individuals.

Definition at line 220 of file variation.h.

Here is the call graph for this function:

Here is the caller graph for this function:

Member Data Documentation

◆ constants_simplifier

template<ProgramType T>

Constants_simplifier Brush::Var::Variation< T >::constants_simplifier

private

Definition at line 603 of file variation.h.

◆ data

template<ProgramType T>

Dataset& Brush::Var::Variation< T >::data

Definition at line 592 of file variation.h.

◆ inexact_simplifier

template<ProgramType T>

Inexact_simplifier Brush::Var::Variation< T >::inexact_simplifier

private

Definition at line 604 of file variation.h.

◆ op_bandits

template<ProgramType T>

map<DataType, map<size_t, Bandit> > Brush::Var::Variation< T >::op_bandits

private

Definition at line 600 of file variation.h.

◆ parameters

template<ProgramType T>

Parameters Brush::Var::Variation< T >::parameters

Definition at line 593 of file variation.h.

◆ search_space

template<ProgramType T>

SearchSpace Brush::Var::Variation< T >::search_space

Definition at line 591 of file variation.h.

◆ terminal_bandits

template<ProgramType T>

map<DataType, Bandit> Brush::Var::Variation< T >::terminal_bandits

private

Definition at line 599 of file variation.h.

◆ variation_bandit

template<ProgramType T>

Bandit Brush::Var::Variation< T >::variation_bandit

private

Definition at line 598 of file variation.h.

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

Public Member Functions

Public Attributes

Private Attributes

Detailed Description

Constructor & Destructor Documentation

◆ Variation() [1/2]

◆ Variation() [2/2]

◆ ~Variation()

Member Function Documentation

◆ bandit_get_node_like()

◆ bandit_sample_op()

◆ bandit_sample_op_with_arg()

◆ bandit_sample_terminal()

◆ cross()

◆ init()

◆ log_simplification_table()

◆ mutate()

◆ update_ss()

◆ vary()

◆ vary_and_update()

Member Data Documentation

◆ constants_simplifier

◆ data

◆ inexact_simplifier

◆ op_bandits

◆ parameters

◆ search_space

◆ terminal_bandits

◆ variation_bandit