How to Create a Simple Network

A basic network task is the XOR (eXclusive-OR) problem. The requirement is that if either of two inputs are true, the output is true, but if neither or both inputs are false, the output is false.

This can be described in a 'truth table' as follows:
Input 1 Input 2 Output
false false false
false true true
true false true
true true false

It can be shown that three neurons are required to perform the processing.

As this problem uses true and false data, traditionally represented as 1 and 0, a layer of sigmoid neurons will be ideal, as this produces values in the range of 0 to 1. The Initial step is to build the actual network. Three layers will be required to emulate the network:

To place a sigmoid layer on the development pane, click on the sigmoid layer toolbar icon. The toolbar icon will appear depressed and moving the cursor over the development pane will show as a crosshair.

Click on the development pane. A sigmoid layer will be placed on the screen. Repeat this process twice more so that there are three sigmoid layers on the development pane.

Clicking on any of the layers displays a properties panel, which shows information about the layer. These parameters can be edited.

For each layer, edit the parameters as follows:
Layer Layer Name Rows
1 Input 2
2 Hidden 3
3 Output 1

Click on the layer labeled Input. A blue circle will appear. Click and drag a connection line from this layer to the Hidden layer.

Repeat this between the Hidden layer and the Output layer.

The next tutorial, 'How to Save and Open a Network' shows how to save this project to file.