mlpack::rl::GreedyPolicy< EnvironmentType > Class Template Reference

Implementation for epsilon greedy policy. More...

#include <greedy_policy.hpp>

Public Types
using	ActionType = typename EnvironmentType::Action
	Convenient typedef for action.

Public Member Functions
	GreedyPolicy (const double initialEpsilon, const size_t annealInterval, const double minEpsilon, const double decayRate=1.0)
	Constructor for epsilon greedy policy class. More...
ActionType	Sample (const arma::colvec &actionValue, bool deterministic=false, const bool isNoisy=false)
	Sample an action based on given action values. More...
void	Anneal ()
	Exploration probability will anneal at each step.
const double &	Epsilon () const

Detailed Description

template<typename EnvironmentType>
class mlpack::rl::GreedyPolicy< EnvironmentType >

Implementation for epsilon greedy policy.

In general we will select an action greedily based on the action value, however sometimes we will also randomly select an action to encourage exploration.

Template Parameters

EnvironmentType

The reinforcement learning task.

Constructor & Destructor Documentation

GreedyPolicy()

template<typename EnvironmentType >

mlpack::rl::GreedyPolicy< EnvironmentType >::GreedyPolicy ( const double  initialEpsilon, const size_t  annealInterval, const double  minEpsilon, const double  decayRate = 1.0  )

inline

Constructor for epsilon greedy policy class.

Parameters

initialEpsilon	The initial probability to explore (select a random action).
annealInterval	The steps during which the probability to explore will anneal.
minEpsilon	Epsilon will never be less than this value.
decayRate	How much to change the model in response to the estimated error each time the model weights are updated.

Member Function Documentation

Epsilon()

template<typename EnvironmentType >

const double& mlpack::rl::GreedyPolicy< EnvironmentType >::Epsilon ( ) const

inline

Returns

Current possibility to explore.

Sample()

template<typename EnvironmentType >

ActionType mlpack::rl::GreedyPolicy< EnvironmentType >::Sample ( const arma::colvec &  actionValue, bool  deterministic = false, const bool  isNoisy = false  )

inline

Sample an action based on given action values.

Parameters

actionValue	Values for each action.
deterministic	Always select the action greedily.
isNoisy	Specifies whether the network used is noisy.

Returns

Sampled action.

---

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

• src/mlpack/methods/reinforcement_learning/policy/greedy_policy.hpp