Home > marl > agent > explorefuns > randomexplore.m

randomexplore

PURPOSE ^

Implements non-exploitive random exploration.

SYNOPSIS ^

function [a, action] = randomexplore(a, action, state, actions, rewards, params)

DESCRIPTION ^

Implements non-exploitive random exploration.
  [A, ACTION] = RANDOMEXPLORE(A, ACTION, STATE, ACTIONS, REWARDS, PARAMS)
  Implements non-exploitive random exploration with probability epsilon.
  This is the most basic (yet the most commonly used) exploration
  strategy. Can be used with any learning/policy functions combination, if 
  the required parameters listed below are fed to the learning
  initialization function.

  Supports discrete actions, with 1 action variable only.

  Required values on the agent learning parameters:
   epsilon         - the exploration probability
   epsilondecay    - the exploration probability decay ratio. Required
                   in non-episodic environments. If specified, epsilon
                   will decay with this ratio at each iteration,
                   otherwise, in episodic environments, it will decay with
                   the trials count.
  Required values on extra parameters:
   newtrial        - (in episodic environments) whether a new trial is
                   beginning


  See also act, explore

CROSS-REFERENCE INFORMATION ^

This function calls: This function is called by:
Generated on Wed 04-Aug-2010 16:55:08 by m2html © 2005