cuberl::rl::algos Namespace Reference

Namespaces
namespace	dp
namespace	mc
namespace	pg
namespace	td

Classes
struct	RLAlgoConfig
	The RLAlgoConfig struct. More...
class	RLSolverBase
	RLAlgoBase. Base class for RL algorithms. More...

Functions
template<typename WorldTp >
auto	state_actions_from_v (const WorldTp &env, const DynVec< real_t > &v, real_t gamma, uint_t state) -> DynVec< real_t >
	Given the state index returns the list of actions under the provided value functions.
template<typename T >
std::vector< T >	create_discounts_array (T base, uint_t npoints)
	create_discounts_array
template<typename T >
std::vector< T >	calculate_discounted_return_vector (const std::vector< T > &rewards, T gamma)
	Create a vector where element i is the product $$\gamma^i * rewards[i]$$.
template<typename TimeStepType , typename T >
std::vector< T >	calculate_discounted_return_vector (const std::vector< TimeStepType > &trajectory, T gamma)
	calculate_discounted_return_vector. Creates the discounted return vector for the given trajectory
template<typename T >
T	calculate_discounted_return (const std::vector< T > &rewards, T gamma)
	calculate_discounted_return. Calculates the sum of the discounted rewards for the given rewards array using the given gamma
template<typename T >
T	calculate_mean_discounted_return (const std::vector< T > &rewards, T gamma)
	calculate_mean_discounted_return. Same as calculate_discounted_return but the result is weighted by 1/N where N is the size of the given rewards array
template<typename TimeStepType , typename T >
T	calculate_discounted_return (const std::vector< TimeStepType > &trajectory, T gamma)
	Calculate the discounted return from the given trajectory.
template<typename TimeStepType , typename T >
T	calculate_mean_discounted_return (const std::vector< TimeStepType > &trajectory, T gamma)
template<typename T >
std::vector< T >	calculate_step_discounted_return (const std::vector< T > &rewards, T gamma)
	Given an array of rewards, for each entry calculate the following: $$G = \sum_{k=t+1}^T \gamma^{k-t-1}R_k$$.

Function Documentation

calculate_discounted_return() [1/2]

template<typename T >

T cuberl::rl::algos::calculate_discounted_return	(	const std::vector< T > &	rewards,
		T	gamma
	)

calculate_discounted_return. Calculates the sum of the discounted rewards for the given rewards array using the given gamma

Parameters

rewards
gamma

Returns

calculate_discounted_return() [2/2]

template<typename TimeStepType , typename T >

T cuberl::rl::algos::calculate_discounted_return	(	const std::vector< TimeStepType > &	trajectory,
		T	gamma
	)

Calculate the discounted return from the given trajectory.

calculate_discounted_return_vector() [1/2]

template<typename T >

std::vector< T > cuberl::rl::algos::calculate_discounted_return_vector	(	const std::vector< T > &	rewards,
		T	gamma
	)

Create a vector where element i is the product $$\gamma^i * rewards[i]$$.

calculate_discounted_return_vector() [2/2]

template<typename TimeStepType , typename T >

std::vector< T > cuberl::rl::algos::calculate_discounted_return_vector	(	const std::vector< TimeStepType > &	trajectory,
		T	gamma
	)

calculate_discounted_return_vector. Creates the discounted return vector for the given trajectory

calculate_mean_discounted_return() [1/2]

template<typename T >

T cuberl::rl::algos::calculate_mean_discounted_return	(	const std::vector< T > &	rewards,
		T	gamma
	)

calculate_mean_discounted_return. Same as calculate_discounted_return but the result is weighted by 1/N where N is the size of the given rewards array

Parameters

rewards
gamma

Returns

calculate_mean_discounted_return() [2/2]

template<typename TimeStepType , typename T >

T cuberl::rl::algos::calculate_mean_discounted_return	(	const std::vector< TimeStepType > &	trajectory,
		T	gamma
	)

calculate_step_discounted_return()

template<typename T >

std::vector< T > cuberl::rl::algos::calculate_step_discounted_return	(	const std::vector< T > &	rewards,
		T	gamma
	)

Given an array of rewards, for each entry calculate the following: $$G = \sum_{k=t+1}^T \gamma^{k-t-1}R_k$$.

create_discounts_array()

template<typename T >

std::vector< T > cuberl::rl::algos::create_discounts_array	(	T	base,
		uint_t	npoints
	)

create_discounts_array

Parameters

end
base
start
endpoint

Returns

state_actions_from_v()

template<typename WorldTp >

auto cuberl::rl::algos::state_actions_from_v	(	const WorldTp &	env,
		const DynVec< real_t > &	v,
		real_t	gamma,
		uint_t	state
	)		-> DynVec<real_t>

Given the state index returns the list of actions under the provided value functions.