cuberl::rl::policies::EpsilonGreedyPolicy Class Reference

The EpsilonGreedyPolicy class. More...

#include <epsilon_greedy_policy.h>

Public Types
typedef uint_t	output_type
	The type returned when calling this->operator()

Public Member Functions
	EpsilonGreedyPolicy (real_t eps)
	Constructor. Creates an epsilon-greedy tabular policy.
	EpsilonGreedyPolicy (real_t eps, uint_t seed)
	Constructor. Creates an epsilon-greedy tabular policy.
	EpsilonGreedyPolicy (real_t eps, uint_t seed, EpsilonDecayOption decay_op, real_t min_eps=MIN_EPS, real_t max_eps=MAX_EPS, real_t epsilon_decay=EPSILON_DECAY_FACTOR)
	Constructor Creates an epsilon greedy policy with an epsilon decay strategy.
template<typename MapType >
output_type	operator() (const MapType &q_map, uint_t state) const
	operator() Select action for the given state
template<typename VecType >
output_type	operator() (const VecType &vec) const
	Get an action i.e. index from the given values.
template<typename MatType >
output_type	get_action (const MatType &q_map, uint_t state_idx)
	get_action. Given a
template<typename VecTp >
output_type	get_action (const VecTp &q_map)
	get_action. Given a vector always returns the position of the maximum occuring element. If the given vector is empty returns CubeAIConsts::invalid_size_type
void	on_episode (uint_t episode_idx) noexcept
	any actions the policy should perform on the given episode index
void	reset () noexcept
	Reset the policy.
real_t	eps_value () const noexcept
	Returns the value of the epsilon.
void	set_eps_value (real_t eps)
	Set the epsilon value.
EpsilonDecayOption	decay_option () const noexcept
	Returns the decay option.

Static Public Attributes
static constexpr real_t	MIN_EPS = 0.01
static constexpr real_t	MAX_EPS = 1.0
static constexpr real_t	EPSILON_DECAY_FACTOR = 0.01

Detailed Description

The EpsilonGreedyPolicy class.

Member Typedef Documentation

output_type

typedef uint_t cuberl::rl::policies::EpsilonGreedyPolicy::output_type

The type returned when calling this->operator()

Constructor & Destructor Documentation

EpsilonGreedyPolicy() [1/3]

cuberl::rl::policies::EpsilonGreedyPolicy::EpsilonGreedyPolicy ( real_t  eps )

inline

Constructor. Creates an epsilon-greedy tabular policy.

EpsilonGreedyPolicy() [2/3]

cuberl::rl::policies::EpsilonGreedyPolicy::EpsilonGreedyPolicy ( real_t  eps, uint_t  seed  )

inlineexplicit

Constructor. Creates an epsilon-greedy tabular policy.

EpsilonGreedyPolicy() [3/3]

cuberl::rl::policies::EpsilonGreedyPolicy::EpsilonGreedyPolicy ( real_t  eps, uint_t  seed, EpsilonDecayOption  decay_op, real_t  min_eps = MIN_EPS, real_t  max_eps = MAX_EPS, real_t  epsilon_decay = EPSILON_DECAY_FACTOR  )

inlineexplicit

Constructor Creates an epsilon greedy policy with an epsilon decay strategy.

Member Function Documentation

decay_option()

EpsilonDecayOption cuberl::rl::policies::EpsilonGreedyPolicy::decay_option ( ) const

inlinenoexcept

Returns the decay option.

eps_value()

real_t cuberl::rl::policies::EpsilonGreedyPolicy::eps_value ( ) const

inlinenoexcept

Returns the value of the epsilon.

get_action() [1/2]

template<typename MatType >

output_type cuberl::rl::policies::EpsilonGreedyPolicy::get_action	(	const MatType &	q_map,
		uint_t	state_idx
	)

get_action. Given a

get_action() [2/2]

template<typename VecTp >

EpsilonGreedyPolicy::output_type cuberl::rl::policies::EpsilonGreedyPolicy::get_action

(

const VecTp &

q_map

)

get_action. Given a vector always returns the position of the maximum occuring element. If the given vector is empty returns CubeAIConsts::invalid_size_type

on_episode()

void cuberl::rl::policies::EpsilonGreedyPolicy::on_episode ( uint_t  episode_idx )

noexcept

any actions the policy should perform on the given episode index

operator()() [1/2]

template<typename MapType >

output_type cuberl::rl::policies::EpsilonGreedyPolicy::operator()	(	const MapType &	q_map,
		uint_t	state
	)		const

operator() Select action for the given state

operator()() [2/2]

template<typename VecType >

EpsilonGreedyPolicy::output_type cuberl::rl::policies::EpsilonGreedyPolicy::operator()

(

const VecType &

vec

)

const

Get an action i.e. index from the given values.

reset()

void cuberl::rl::policies::EpsilonGreedyPolicy::reset ( )

inlinenoexcept

Reset the policy.

set_eps_value()

void cuberl::rl::policies::EpsilonGreedyPolicy::set_eps_value

(

real_t

eps

)

Set the epsilon value.

Parameters

eps

Member Data Documentation

EPSILON_DECAY_FACTOR

constexpr real_t cuberl::rl::policies::EpsilonGreedyPolicy::EPSILON_DECAY_FACTOR = 0.01

staticconstexpr

MAX_EPS

constexpr real_t cuberl::rl::policies::EpsilonGreedyPolicy::MAX_EPS = 1.0

staticconstexpr

MIN_EPS

constexpr real_t cuberl::rl::policies::EpsilonGreedyPolicy::MIN_EPS = 0.01

staticconstexpr

---

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

• libs/cuberl/include/cuberl/rl/policies/epsilon_greedy_policy.h