C# Класс NNX.Core.MultilayerPerceptron

Multilayer perceptron with the following features: * Hidden layer activation functions: tahn * Output layer activation function: softmax * Gradient calculation assumes optimization for min cross-entropy error.
Наследование: INeuralNetwork
Показать файл Открыть проект

Открытые методы

Метод Описание
CalculateGradients ( double inputs, double targets ) : double[][]

Calculates gradients of the error function (cross-entropy) with respect to weights.
FeedForward ( double inputs ) : FeedForwardResult

Activation functions: tanh for hidden nodes, softmax for output nodes.
MultilayerPerceptron ( int numInputs, int numOutputs, IList hiddenLayerSizes ) : System

Описание методов

CalculateGradients() публичный Метод

Calculates gradients of the error function (cross-entropy) with respect to weights.
public CalculateGradients ( double inputs, double targets ) : double[][]
inputs double
targets double
Результат double[][]

FeedForward() публичный Метод

Activation functions: tanh for hidden nodes, softmax for output nodes.
public FeedForward ( double inputs ) : FeedForwardResult
inputs double
Результат FeedForwardResult

MultilayerPerceptron() публичный Метод

public MultilayerPerceptron ( int numInputs, int numOutputs, IList hiddenLayerSizes ) : System
numInputs int
numOutputs int
hiddenLayerSizes IList
Результат System