c# - Using a Radial basis function with back propagation in c# using Encog library

Question

I have created a simple RBF network with a gaussian function by using,

RBFNetwork newNetwork = new RBFNetwork(28,14,1,RBFEnum.Gaussian);

I need to create a RBF Network with back propagation having 28 inputs and only one output giving 0 or 1 as a result. I could not proceed further in training it with data sets. Help needed.

Answer 1

如果您不想制定自己的停止策略，我认为您应该使用EncogUtility.TrainConsole(),EncogUtility.TrainToError()或方法。EncogUtility.TrainDialog()

请参阅Javadocs 的EncogUtility，C# 接近。唯一的问题是它内部不使用 SCG，而是使用ResilientPropagation.

要构建数据集，您必须具有相同大小和相同顺序的double[][]输入和double[][]所需输出。然后代码看起来像这样：

var trainingSet = new BasicMLDataSet(inputs, outputs);
EncogUtility.TrainConsole(network, trainingSet, 10 /*minutes training*/);

Answer 2

我认为你的答案在这里.. http://massapi.com/class/org/encog/neural/rbf/RBFNetwork.java.html 希望这对你有帮助..虽然它是在java中你可以理解它背后的逻辑