rl_logic.ActionSelector Class Reference

Class for selecting the action from the list of actions and their corresponding values. More...

Collaboration diagram for rl_logic.ActionSelector:

Public Member Functions
def	__init__ (self, selection_method_id="greedy")
	Constructor. More...
def	select_action (self)
	Default method for selecting the action. More...
def	select_action_greedy (self, action_values)
	Select an action using "greedy" algorithm. More...
def	select_action_e_greedy (self, action_values)
	Select an action using "e-greedy" algorithm. More...
def	select_action_softmax (self, action_values)
	Select an action using "soft-max" algorithm, based on Gibbs (Boltzmann) distribution. More...

Public Attributes
	select_action
	A reference to the selection method being used. More...
	selection_method_id
	Store a selection method ID value. More...
	eps
	Eps-value for the e-greedy selection method. More...

Detailed Description

Class for selecting the action from the list of actions and their corresponding values.

The interface is provided via select_action() method.

Constructor & Destructor Documentation

def rl_logic.ActionSelector.__init__	(		self,
			selection_method_id = "greedy"
	)

Constructor.

Parameters

self	The object pointer.
self	ID of used selection method.

Member Function Documentation

def rl_logic.ActionSelector.select_action

(

self

)

Default method for selecting the action.

It is overridden in init(). Input: {action_id: value}. Output: action_id.

Parameters

self	The object pointer.

def rl_logic.ActionSelector.select_action_e_greedy	(		self,
			action_values
	)

Select an action using "e-greedy" algorithm.

Parameters

self	The object pointer.
action_values	A dictionary containing {action_id: estimation_value}.

Returns

The selected action_id.

Here is the call graph for this function:

def rl_logic.ActionSelector.select_action_greedy	(		self,
			action_values
	)

Select an action using "greedy" algorithm.

Parameters

self	The object pointer.
action_values	A dictionary containing {action_id: estimation_value}.

Returns

The action_id with the maximum value.

Here is the caller graph for this function:

def rl_logic.ActionSelector.select_action_softmax	(		self,
			action_values
	)

Select an action using "soft-max" algorithm, based on Gibbs (Boltzmann) distribution.

See the reference in R.Sutton's book: Reinforcement Learning: An Introduction.

Parameters

self	The object pointer.
action_values	A dictionary containing {action_id: estimation_value}.

Returns

The selected action_id.

Member Data Documentation

rl_logic.ActionSelector.eps

Eps-value for the e-greedy selection method.

Default value is 0.1.

rl_logic.ActionSelector.select_action

A reference to the selection method being used.

rl_logic.ActionSelector.selection_method_id

Store a selection method ID value.

---

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

• rl_logic/__init__.py