RewardClipping< EnvironmentType > Class Template Reference

Interface for clipping the reward to some value between the specified maximum and minimum value (Clipping here is implemented as $ g_{\text{clipped}} = \max(g_{\text{min}}, \min(g_{\text{min}}, g))) $.) More...

Public Types

using Action = typename EnvironmentType::Action
 Convenient typedef for action. More...

 
using State = typename EnvironmentType::State
 Convenient typedef for state. More...

 

Public Member Functions

 RewardClipping (EnvironmentType &environment, const double minReward=-1.0, const double maxReward=1.0)
 Constructor for creating a RewardClipping instance. More...

 
EnvironmentType & Environment () const
 Get the environment. More...

 
EnvironmentType & Environment ()
 Modify the environment. More...

 
State InitialSample ()
 The InitialSample method is called by the environment to initialize the starting state. More...

 
bool IsTerminal (const State &state) const
 Checks whether given state is a terminal state. More...

 
double MaxReward () const
 Get the maximum reward value. More...

 
double & MaxReward ()
 Modify the maximum reward value. More...

 
double MinReward () const
 Get the minimum reward value. More...

 
double & MinReward ()
 Modify the minimum reward value. More...

 
double Sample (const State &state, const Action &action, State &nextState)
 Dynamics of Environment. More...

 
double Sample (const State &state, const Action &action)
 Dynamics of Environment. More...

 

Detailed Description


template
<
typename
EnvironmentType
>

class mlpack::rl::RewardClipping< EnvironmentType >

Interface for clipping the reward to some value between the specified maximum and minimum value (Clipping here is implemented as $ g_{\text{clipped}} = \max(g_{\text{min}}, \min(g_{\text{min}}, g))) $.)

Template Parameters
EnvironmentTypeA type of Environment that is being wrapped.

Definition at line 29 of file reward_clipping.hpp.

Member Typedef Documentation

◆ Action

using Action = typename EnvironmentType::Action

Convenient typedef for action.

Definition at line 36 of file reward_clipping.hpp.

◆ State

using State = typename EnvironmentType::State

Convenient typedef for state.

Definition at line 33 of file reward_clipping.hpp.

Constructor & Destructor Documentation

◆ RewardClipping()

RewardClipping ( EnvironmentType &  environment,
const double  minReward = -1.0,
const double  maxReward = 1.0 
)
inline

Constructor for creating a RewardClipping instance.

Parameters
minRewardMinimum possible value of clipped reward.
maxRewardMaximum possible value of clipped reward.
environmentAn instance of the environment used for actual simulations.

Definition at line 46 of file reward_clipping.hpp.

Member Function Documentation

◆ Environment() [1/2]

EnvironmentType& Environment ( ) const
inline

Get the environment.

Definition at line 112 of file reward_clipping.hpp.

◆ Environment() [2/2]

EnvironmentType& Environment ( )
inline

Modify the environment.

Definition at line 114 of file reward_clipping.hpp.

◆ InitialSample()

State InitialSample ( )
inline

The InitialSample method is called by the environment to initialize the starting state.

Returns whatever Initial Sample is returned by the environment.

Definition at line 61 of file reward_clipping.hpp.

◆ IsTerminal()

bool IsTerminal ( const State state) const
inline

Checks whether given state is a terminal state.

Returns the value by calling the environment method.

Parameters
statedesired state.
Returns
true if state is a terminal state, otherwise false.

Definition at line 73 of file reward_clipping.hpp.

◆ MaxReward() [1/2]

double MaxReward ( ) const
inline

Get the maximum reward value.

Definition at line 122 of file reward_clipping.hpp.

◆ MaxReward() [2/2]

double& MaxReward ( )
inline

Modify the maximum reward value.

Definition at line 124 of file reward_clipping.hpp.

◆ MinReward() [1/2]

double MinReward ( ) const
inline

Get the minimum reward value.

Definition at line 117 of file reward_clipping.hpp.

◆ MinReward() [2/2]

double& MinReward ( )
inline

Modify the minimum reward value.

Definition at line 119 of file reward_clipping.hpp.

◆ Sample() [1/2]

double Sample ( const State state,
const Action action,
State nextState 
)
inline

Dynamics of Environment.

The rewards returned from the base environment are clipped according the maximum and minimum values specified.

Parameters
stateThe current state.
actionThe current action.
nextStateThe next state.
Returns
clippedReward, Reward clipped between [minReward, maxReward].

Definition at line 87 of file reward_clipping.hpp.

Referenced by RewardClipping< EnvironmentType >::Sample().

◆ Sample() [2/2]

double Sample ( const State state,
const Action action 
)
inline

Dynamics of Environment.

The rewards returned from the base environment are clipped according the maximum and minimum values specified.

Parameters
stateThe current state.
actionThe current action.
Returns
clippedReward, Reward clipped between [minReward, maxReward].

Definition at line 105 of file reward_clipping.hpp.

References RewardClipping< EnvironmentType >::Sample().


The documentation for this class was generated from the following file:
  • /home/jenkins-mlpack/mlpack.org/_src/mlpack-3.2.1/src/mlpack/methods/reinforcement_learning/environment/reward_clipping.hpp