C# Class Accord.MachineLearning.Sarsa

Sarsa learning algorithm.

The class provides implementation of Sarsa algorithm, known as on-policy Temporal Difference control.

显示文件 Open project: accord-net/framework Class Usage Examples

Public Methods

Method	Description
GetAction ( int state ) : int	Get next action from the specified state. The method returns an action according to current exploration policy.
Sarsa ( int states, int actions, IExplorationPolicy explorationPolicy ) : System	Initializes a new instance of the Sarsa class. Action estimates are randomized in the case of this constructor is used.
Sarsa ( int states, int actions, IExplorationPolicy explorationPolicy, bool randomize ) : System	Initializes a new instance of the Sarsa class. The randomize parameter specifies if initial action estimates should be randomized with small values or not. Randomization of action values may be useful, when greedy exploration policies are used. In this case randomization ensures that actions of the same type are not chosen always.
UpdateState ( int previousState, int previousAction, double reward ) : void	Update Q-function's value for the previous state-action pair. Updates Q-function's value for the previous state-action pair in the case if the next state is terminal.
UpdateState ( int previousState, int previousAction, double reward, int nextState, int nextAction ) : void	Update Q-function's value for the previous state-action pair. Updates Q-function's value for the previous state-action pair in the case if the next state is non terminal.

Method Details

GetAction() public method

Get next action from the specified state.

The method returns an action according to current exploration policy.

public GetAction ( int state ) : int
state	int	Current state to get an action for.
return	int

Sarsa() public method

Initializes a new instance of the Sarsa class.

Action estimates are randomized in the case of this constructor is used.

public Sarsa ( int states, int actions, IExplorationPolicy explorationPolicy ) : System
states	int	Amount of possible states.
actions	int	Amount of possible actions.
explorationPolicy	IExplorationPolicy	Exploration policy.
return	System

Sarsa() public method

Initializes a new instance of the Sarsa class.

The randomize parameter specifies if initial action estimates should be randomized with small values or not. Randomization of action values may be useful, when greedy exploration policies are used. In this case randomization ensures that actions of the same type are not chosen always.

public Sarsa ( int states, int actions, IExplorationPolicy explorationPolicy, bool randomize ) : System
states	int	Amount of possible states.
actions	int	Amount of possible actions.
explorationPolicy	IExplorationPolicy	Exploration policy.
randomize	bool	Randomize action estimates or not.
return	System

UpdateState() public method

Update Q-function's value for the previous state-action pair.

Updates Q-function's value for the previous state-action pair in the case if the next state is terminal.

public UpdateState ( int previousState, int previousAction, double reward ) : void
previousState	int	Curren state.
previousAction	int	Action, which lead from previous to the next state.
reward	double	Reward value, received by taking specified action from previous state.
return	void

UpdateState() public method

Update Q-function's value for the previous state-action pair.

Updates Q-function's value for the previous state-action pair in the case if the next state is non terminal.

public UpdateState ( int previousState, int previousAction, double reward, int nextState, int nextAction ) : void
previousState	int	Curren state.
previousAction	int	Action, which lead from previous to the next state.
reward	double	Reward value, received by taking specified action from previous state.
nextState	int	Next state.
nextAction	int	Next action.
return	void