QLearner

Description:

  A simple reinforcement learning framework that can be used to learn optimal policies for Markov decision processes using Q-learning. Q-learning is a model-free reinforcement learning algorithm that learns an optimal action-value function from experience by repeatedly updating estimates of the Q-value of state-action pairs.

Class Object: QLearner Class.

Inherits from: Object.

matrix

Type: Readonly Field.

Description:

  The matrix that stores state, action, and Q-value.

Signature:

const matrix: {{
		--[[state]] integer,
		--[[action]] integer,
		--[[Q-value]] number
	}}

update

Type: Function.

Description:

  Update Q-value for a state-action pair based on received reward.

Signature:

update: function(self: QLearner, state: integer, action: integer, reward: number)

Parameters:

Parameter	Type	Description
state	integer	Representing the state.
action	integer	Representing the action. Must be greater than 0.
reward	number	Representing the reward received for the action in the state.

getBestAction

Type: Function.

Description:

  Returns the best action for a given state based on the current Q-values.

Signature:

getBestAction: function(self: QLearner, state: integer): integer

Parameters:

Parameter	Type	Description
state	integer	The current state.

Returns:

Return Type	Description
integer	The action with the highest Q-value for the given state. Returns 0 if no action is available.

load

Type: Function.

Description:

  Load Q-values from a matrix of state-action pairs.

Signature:

load: function(self: QLearner, values: {{
			--[[state]] integer,
			--[[action]] integer,
			--[[Q-value]] number
		}})

Parameters:

Parameter	Type	Description
values	{{integer - The state, integer The action, number - The Q-value for the given state-action pair}}	The matrix of state-action pairs to load.