XOR Sample
This NeuronDotNet application demonstrates the use of backpropagation neural networks in non-linear classification problems.
Download LinksExclusive OR (XOR) problem is a simple non-linear classification problem. Consider a two-dimensional surface with points (0,1) and (1,0) of one type and (0,0) and (1,1) of some other type. It is not possible to separate these two types of points using a single line on the plane. (If a backpropagation network has no hidden layers, its output will be a linear function of the inputs and hence, it will not be able to classify these points. That's why, a backpropagation network must have atleast one hidden layer)
This sample application uses a simple backpropagation network to demonstrate its behaviour when subjected to XOR training set. The learning rate, number of hidden neurons and the number of training cycles is obtained from the user. The error graph shows how 'Mean Squared Error' tends to vanish with the progress in training. The trained neural network can be tested by running it against test inputs.
(Zedgraph, an open source library is used to plot the error graph)
A screenshot of the application is shown below.
