Class ILNeuralNetwork.TILNeuralNetworkBackpropTrain
Revision as of 00:27, 22 November 2011
This is a Beta Read Only version of this page. Please review and send recommendations to mitov@mitov.com. We will enable the editing as soon as we are happy with the overall Wiki site.
Package: IntelligenceLabPkg
Unit: ILNeuralNetwork
Inherits: TILNeuralNetworkTrain
Contents |
Syntax
Delphi:
type TILNeuralNetworkBackpropTrain = class( TILNeuralNetworkTrain )
C++ Builder:
class TILNeuralNetworkBackpropTrain : public TILNeuralNetworkTrain
Visual C++ (MFC):
class CTILNeuralNetworkBackpropTrain : public CTILNeuralNetworkTrain
C# (.NET):
public ref class TILNeuralNetworkBackpropTrain : Mitov.IntelligenceLab.NeuralNetworkTrain
Summary
Trains neural networks using Backprop algorithm.
Description
This component is designed to train neural networks using Backprop algorithm.
To use the component set the Neural Network to be trained in the NeuralNetwork property.
Properties
Published
- Weights - Specifies the training weights.
From TILNeuralNetworkTrain
Properties
Published
- NeuralNetwork - Specifies the Neural Network component to be trained.
- TerminationCriteria - Criteria for terminating the training.
- NormalizeTrainingWeights - Specifies if the weights should be normalized.
Methods
Public
- function Train(ATraingFeatures : ISLRealBufferArray; AResposes : ISLRealBufferArray) : Integer - Trains the connected neural network.
- function Train(ATraingFeatures : ISLRealBufferArray; AResposes : ISLRealBufferArray; AWeights : ISLRealBuffer) : Integer - Trains the connected neural network.
Pins
- InputPin - The training data input pin of the component.
- ProgressPin - The Training Progress Pin.
Events
- OnProgress - Training progress notification event.