[This is preliminary documentation and is subject to change.]

This class implements a Q-learning with backpropagation neural network in the Clarion Library

The QBPNetwork..::..QBPNetworkParameters type exposes the following members.

Constructors

  Name Description
Public method QBPNetwork..::..QBPNetworkParameters(Boolean)
Initializes a parameters class for a QBPNetwork
Public method QBPNetwork..::..QBPNetworkParameters(QBPNetwork..::..QBPNetworkParameters, Boolean)
Initializes a parameters class for a QBPNetwork and sets the values of the parameters to the values specified

Methods

  Name Description
Protected method Global_OnParameterChanged
The event method used for performing global parameter changes
(Inherited from ClarionComponent..::..ClarionComponentParameters.)

Properties

  Name Description
Public property DISCOUNT
The discount factor
Public property ELIGIBILITY
A switch that can be used to specify eligibility (in the absence of a more sophisticated eligibility method)
Public property EXTRACTION_PROBABILITY
The extraction probability (i.e., the probability a rule will be extracted)
Public property INITIAL_THRESHOLDS_LOWER_BOUND
The low bound for the initial threshold
Public property INITIAL_THRESHOLDS_UPPER_BOUND
The upper bound for the initial threshold
Public property INITIAL_WEIGHTS_LOWER_BOUND
The low bound for the initial dWeights
Public property INITIAL_WEIGHTS_UPPER_BOUND
The upper bound for the initial dWeights
Protected property IsGlobal
Used to specify whether the particular instance of the parameters class is being used for global or local parameters
(Inherited from ClarionComponent..::..ClarionComponentParameters.)
Public property LEARNING_RATE
The learning rate (used for learning)
Public property MOMENTUM
The momentum (used for learning)
Public property POSITIVE_MATCH_THRESHOLD
The threshold for the positive match criterion
Public property R_ZERO
The error tolerance (used for learning)

See Also