cuberl::rl::RLSerialAgentTrainer< EnvType, AgentType > Class Template Reference

#include <rl_serial_agent_trainer.h>

Inheritance diagram for cuberl::rl::RLSerialAgentTrainer< EnvType, AgentType >:

Collaboration diagram for cuberl::rl::RLSerialAgentTrainer< EnvType, AgentType >:

Public Types
typedef EnvType	env_type
typedef AgentType	agent_type

Public Member Functions
	RLSerialAgentTrainer (const RLSerialTrainerConfig &config, agent_type &agent)
	RLSerialAgentTrainer.
virtual bitrl::utils::IterativeAlgorithmResult	train (env_type &env)
	train Iterate to train the agent on the given environment
virtual void	actions_before_training_begins (env_type &)
	actions_before_training_begins. Execute any actions the algorithm needs before starting the episode
virtual void	actions_before_episode_begins (env_type &, uint_t)
	actions_before_episode_begins. Execute any actions the algorithm needs before starting the episode
virtual void	actions_after_episode_ends (env_type &, uint_t, const EpisodeInfo &einfo)
	actions_after_episode_ends. Execute any actions the algorithm needs after ending the episode
virtual void	actions_after_training_ends (env_type &)
	actions_after_training_ends. Execute any actions the algorithm needs after the iterations are finished
const std::vector< real_t > &	episodes_total_rewards () const noexcept
	episodes_total_rewards
const std::vector< uint_t > &	n_itrs_per_episode () const noexcept
	n_itrs_per_episode

Protected Attributes
uint_t	output_msg_frequency_
bitrl::utils::IterativeAlgorithmController	itr_ctrl_
	itr_ctrl_ Handles the iteration over the episodes
agent_type &	agent_
	agent_
std::vector< real_t >	total_reward_per_episode_
	total_reward_per_episode_
std::vector< uint_t >	n_itrs_per_episode_
	n_itrs_per_episode_ Holds the number of iterations performed per training episode

Detailed Description

template<typename EnvType, typename AgentType>
class cuberl::rl::RLSerialAgentTrainer< EnvType, AgentType >

\detailed The RLSerialAgentTrainer class handles the training for serial reinforcement learning agents

Member Typedef Documentation

agent_type

template<typename EnvType , typename AgentType >

typedef AgentType cuberl::rl::RLSerialAgentTrainer< EnvType, AgentType >::agent_type

env_type

template<typename EnvType , typename AgentType >

typedef EnvType cuberl::rl::RLSerialAgentTrainer< EnvType, AgentType >::env_type

Constructor & Destructor Documentation

RLSerialAgentTrainer()

template<typename EnvType , typename AgentType >

cuberl::rl::RLSerialAgentTrainer< EnvType, AgentType >::RLSerialAgentTrainer	(	const RLSerialTrainerConfig &	config,
		agent_type &	agent
	)

RLSerialAgentTrainer.

Parameters

config
agent

Member Function Documentation

actions_after_episode_ends()

template<typename EnvType , typename AgentType >

void cuberl::rl::RLSerialAgentTrainer< EnvType, AgentType >::actions_after_episode_ends ( env_type &  env, uint_t  episode_idx, const EpisodeInfo &  einfo  )

virtual

actions_after_episode_ends. Execute any actions the algorithm needs after ending the episode

actions_after_training_ends()

template<typename EnvType , typename AgentType >

void cuberl::rl::RLSerialAgentTrainer< EnvType, AgentType >::actions_after_training_ends ( env_type &  env )

virtual

actions_after_training_ends. Execute any actions the algorithm needs after the iterations are finished

actions_before_episode_begins()

template<typename EnvType , typename AgentType >

void cuberl::rl::RLSerialAgentTrainer< EnvType, AgentType >::actions_before_episode_begins ( env_type &  env, uint_t  episode_idx  )

virtual

actions_before_episode_begins. Execute any actions the algorithm needs before starting the episode

actions_before_training_begins()

template<typename EnvType , typename AgentType >

void cuberl::rl::RLSerialAgentTrainer< EnvType, AgentType >::actions_before_training_begins ( env_type &  env )

virtual

actions_before_training_begins. Execute any actions the algorithm needs before starting the episode

episodes_total_rewards()

template<typename EnvType , typename AgentType >

const std::vector< real_t > & cuberl::rl::RLSerialAgentTrainer< EnvType, AgentType >::episodes_total_rewards ( ) const

inlinenoexcept

episodes_total_rewards

Returns

n_itrs_per_episode()

template<typename EnvType , typename AgentType >

const std::vector< uint_t > & cuberl::rl::RLSerialAgentTrainer< EnvType, AgentType >::n_itrs_per_episode ( ) const

inlinenoexcept

n_itrs_per_episode

Returns

train()

template<typename EnvType , typename AgentType >

bitrl::utils::IterativeAlgorithmResult cuberl::rl::RLSerialAgentTrainer< EnvType, AgentType >::train ( env_type &  env )

virtual

train Iterate to train the agent on the given environment

Member Data Documentation

agent_

template<typename EnvType , typename AgentType >

agent_type& cuberl::rl::RLSerialAgentTrainer< EnvType, AgentType >::agent_

protected

agent_

itr_ctrl_

template<typename EnvType , typename AgentType >

bitrl::utils::IterativeAlgorithmController cuberl::rl::RLSerialAgentTrainer< EnvType, AgentType >::itr_ctrl_

protected

itr_ctrl_ Handles the iteration over the episodes

n_itrs_per_episode_

template<typename EnvType , typename AgentType >

std::vector<uint_t> cuberl::rl::RLSerialAgentTrainer< EnvType, AgentType >::n_itrs_per_episode_

protected

n_itrs_per_episode_ Holds the number of iterations performed per training episode

output_msg_frequency_

template<typename EnvType , typename AgentType >

uint_t cuberl::rl::RLSerialAgentTrainer< EnvType, AgentType >::output_msg_frequency_

protected

total_reward_per_episode_

template<typename EnvType , typename AgentType >

std::vector<real_t> cuberl::rl::RLSerialAgentTrainer< EnvType, AgentType >::total_reward_per_episode_

protected

total_reward_per_episode_

---

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

• libs/cuberl/include/cuberl/rl/trainers/rl_serial_agent_trainer.h