Learn::LearningParameters Struct Reference

Structure for simplifying the transmission of LearningParameters to functions. More...

#include <learningParameters.h>

Public Attributes
Mutator::MutationParameters	mutation
Selector::SelectionParameters	selection
size_t	archiveSize = 50
	Number of recordings held in the Archive.
double	archivingProbability = 0.05
	Probability of archiving the result of each Program execution.
uint64_t	nbIterationsPerPolicyEvaluation = 5
	Number of evaluation of each policy per generation.
uint64_t	nbIterationsPerPolicyValidation = 5
	Number of evaluation of each policy per generation for validation.
uint64_t	maxNbActionsPerEval = 1000
uint64_t	nbGenerations = 500
	Number of generations of the training.
size_t	maxNbEvaluationPerPolicy = 1000
size_t	nbRegisters = 8
	Number of registers for the Program execution.
size_t	nbProgramConstant = 0
	Number of Constants available in a program.
size_t	nbThreads = std::thread::hardware_concurrency()
	Number of threads (ParallelLearningAgent only)
bool	detailedTiming = false
	enable detailed timing during learning
std::string	activationFunction = "none"
bool	doValidation = false
uint64_t	stepValidation = 1
	Value to indicate how many generation between each validation step.

Static Public Attributes
static const std::string	archiveSizeComment
	JSon comment.
static const std::string	archivingProbabilityComment
	JSon comment.
static const std::string	nbIterationsPerPolicyEvaluationComment
	JSon comment.
static const std::string	nbIterationsPerPolicyValidationComment
	JSon comment.
static const std::string	maxNbActionsPerEvalComment
	JSon comment.
static const std::string	nbGenerationsComment
	JSon comment.
static const std::string	maxNbEvaluationPerPolicyComment
	JSon comment.
static const std::string	nbRegistersComment
	JSon comment.
static const std::string	nbProgramConstantComment
	JSon comment.
static const std::string	nbThreadsComment
	JSon comment.
static const std::string	detailedTimingComment
	JSon comment.
static const std::string	activationFunctionComment
	JSon comment.
static const std::string	doValidationComment
	JSon comment.
static const std::string	stepValidationComment
	JSon comment.

Detailed Description

Structure for simplifying the transmission of LearningParameters to functions.

When modifying this structure and its comments, you must also update the functions in the ParameterParser namespace.

Member Data Documentation

activationFunction

std::string Learn::LearningParameters::activationFunction = "none"

string that indicate the activation function used for continuous actions

activationFunctionComment

const std::string Learn::LearningParameters::activationFunctionComment

inlinestatic

Initial value:

=
            "// string that indicate the activation function used for "
            "continuous actions \n"
            "// \"activationFunction\" : 'none', // Default value"

JSon comment.

archiveSizeComment

const std::string Learn::LearningParameters::archiveSizeComment

inlinestatic

Initial value:

=
            "// Number of recordings held in the Archive.\n"
            "// \"archiveSize\" : 50, // Default value"

JSon comment.

archivingProbabilityComment

const std::string Learn::LearningParameters::archivingProbabilityComment

inlinestatic

Initial value:

=
            "// Probability of archiving the result of each Program "
            "execution.\n"
            "// \"archivingProbability\" : 0.05, // Default value"

JSon comment.

detailedTimingComment

const std::string Learn::LearningParameters::detailedTimingComment

inlinestatic

Initial value:

=
            "// Boolean used to activate detailed timing during learning.\n"
            "// \"detailedTiming\" : false, // Default value"

JSon comment.

doValidation

bool Learn::LearningParameters::doValidation = false

Boolean set to true if the user wants a validation after each training, and false otherwise

doValidationComment

const std::string Learn::LearningParameters::doValidationComment

inlinestatic

Initial value:

=
            "// Boolean used to activate an evaluation of the surviving roots "
            "in validation\n"
            "// mode after the training at each generation.\n"
            "// \"doValidation\" : false, // Default value"

JSon comment.

maxNbActionsPerEval

uint64_t Learn::LearningParameters::maxNbActionsPerEval = 1000

Maximum number of actions performed on the learning environment during the each evaluation of a root

maxNbActionsPerEvalComment

const std::string Learn::LearningParameters::maxNbActionsPerEvalComment

inlinestatic

Initial value:

=
            "// Maximum number of actions performed on the learning "
            "environment during the\n"
            "// each evaluation of a root.\n"
            "// \"maxNbActionsPerEval\" : 1000, // Default value"

JSon comment.

maxNbEvaluationPerPolicy

size_t Learn::LearningParameters::maxNbEvaluationPerPolicy = 1000

Maximum number of times a given policy (i.e. a root TPGVertex) is evaluated.

maxNbEvaluationPerPolicyComment

const std::string Learn::LearningParameters::maxNbEvaluationPerPolicyComment

inlinestatic

Initial value:

=
            "// Maximum number of times a given root is evaluated."
            "After this number is\n"
            "// reached, possibly after several generations, the score of the "
            "root will be\n"
            "// fixed, and no further evaluation will be done.\n"
            "// \"maxNbEvaluationPerPolicy\" : 1000, // Default value"

JSon comment.

mutation

Mutator::MutationParameters Learn::LearningParameters::mutation

MutationParameters for controlling stochastic aspects of the learning process.

nbGenerationsComment

const std::string Learn::LearningParameters::nbGenerationsComment

inlinestatic

Initial value:

=
            "// Number of generations of the training.\n"
            "// \"nbGenerations\" : 500, // Default value"

JSon comment.

nbIterationsPerPolicyEvaluation

uint64_t Learn::LearningParameters::nbIterationsPerPolicyEvaluation = 5

Number of evaluation of each policy per generation.

In LearningAgent and ParallelLearningAgent it is just the number of times the evaluations are repeated (that can produce a more representative result in non-deterministic environments).

nbIterationsPerPolicyEvaluationComment

const std::string Learn::LearningParameters::nbIterationsPerPolicyEvaluationComment

inlinestatic

Initial value:

=
            "// Number of evaluation of each root per generation.\n"
            "// \"nbIterationsPerPolicyEvaluation\" : 5, // Default value"

JSon comment.

nbIterationsPerPolicyValidationComment

const std::string Learn::LearningParameters::nbIterationsPerPolicyValidationComment

inlinestatic

Initial value:

=
            "// Number of evaluation of each root per generation.\n"
            "// \"nbIterationsPerPolicyValidation\" : 5, // Default value"

JSon comment.

nbProgramConstantComment

const std::string Learn::LearningParameters::nbProgramConstantComment

inlinestatic

Initial value:

=
            "// Number of Constant available in each Program.\n"
            "// \"nbProgramConstant\" : 0, // Default value"

JSon comment.

nbRegistersComment

const std::string Learn::LearningParameters::nbRegistersComment

inlinestatic

Initial value:

=
            "// Number of registers for the Program execution.\n"
            "// \"nbRegisters\" : 8, // Default value"

JSon comment.

nbThreads

size_t Learn::LearningParameters::nbThreads = std::thread::hardware_concurrency()

Number of threads (ParallelLearningAgent only)

Integer parameter controlling the number of threads used for parallel execution. Possible values are:

• default : Let the runtime decide using std::thread::hardware_concurrency().
• 0 or 1: Do not use parallelism.
• n > 1: Set the number of threads explicitly.

nbThreadsComment

const std::string Learn::LearningParameters::nbThreadsComment

inlinestatic

Initial value:

=
            "// [Only used in ParallelLearningAgent and child classes.]\n"
            "// Number of threads used for the training process.\n"
            "// When undefined in the json file, this parameter is "
            "automatically set to the\n"
            "// number of cores of the CPU.\n"
            "// /* \"nbThreads\" : 0,*/ // Commented by default"

JSon comment.

selection

Selector::SelectionParameters Learn::LearningParameters::selection

SelectionParameters for controlling selection aspects of the learning process.

stepValidationComment

const std::string Learn::LearningParameters::stepValidationComment

inlinestatic

Initial value:

=
            "// Value to indicate how many generation between each validation "
            "step\n"
            "// \"stepValidation\" : 1, // Default value"

JSon comment.

---

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

• D:/a/gegelati/gegelati/gegelatilib/include/learn/learningParameters.h