Back to Tutorials.

Tutorial 2: Loading a Preexisting Neural Network

The Neural Network Applet comes with several pre-defined examples to allow you to start working with graphs without having to create one yourself. To load an example file, go to the 'File' menu and select 'Load Sample Graph and Data'. This will load a graph and its associated data. A dialog box will open with a drop-down menu allowing you to select a particular example.

• Mail Reading: This example models whether a user will read or not read an article based on whether or not the author is know, if its a new thread, the length, and if the user is at home. The inputs are known, new, short, and home and the output is reads.
  
  
• Mail Reading (simplified): Same as example above except that there is no input home.
  
  
• Assignment 12a: This example models whether a user will look for more information on a book based on whether or not they bought the book, whether they are studying the same subject as the book, whether it is their first book in that subject and whether or not they visited the website. The inputs are bought, edu, first, visited and the output is more_info.
   
• Assignment 12b: This is the same as Assignment 12a except that there are two extra hidden nodes.
   
• Boolean Example: This is a simple example modeling the AND, OR, and XOR logical operations.
  
  
• Electronics: This example models whether a user will buy a piece of electronics based on the user's age, income, if they are a student and their credit rating. The inputs are age, income, stud, cred and the outputs is buy.
  
  
• Small Car Database: This example models whether a car is acceptable or not to a user based on the price, maintenance cost, how many doors, how many people it can hold, trunk size and the safety rating. The inputs are price, maint-cost, doors, persons, trunk-size, and safety and the output is acceptable.
  
  
• Large Car Database: Same as Small Car Database except that there are more training examples. This file is large so may take some time to load.
  
  
• Matching Pennies: This is data for predicting wins by observing some properties in game of matching pennies.
  
  
• Likes TV: This predicts whether a person likes a TV program based on features of the TV program.
  
  
• Holiday: This gives data that predicts whether a person likes a holiday.

Creating a Neural Network from Data:

Neural networks can be created from raw, comma-delimited data. The only requirement is that the text file must start with a line defining the categories of the data. This line has to be of the form T:[category1],[category2], ..., [categoryN]; , in the same order as the data. For example, here is the data file for "Small Car Database":

Data can also be loaded from from normal applet save files, but it will ignore the graph information and just load the examples.

Load a sample data file by going to the 'File' menu and select 'Load Sample Data'. Once the file is loaded, the Construction Wizard dialog will pop up and query you for information on the neural network that you want to build. You should input the number of hidden layers needed, and the number of nodes for a specific hidden layer. Hidden layers can be selected using the pull-down choice menu. The number of nodes default to 2. This is an example of what the dialog box will look like for the small car database example:

You also have to choose which categories are outputs. Depress the checkbox to the left of the category name to make it an output. Input categories become input nodes, and output categories become output nodes.

Also, it may be necessary for some non-numerical categories to be declared as "ordered" by depressing the corresponding checkbox beside the category name. What this means is that this category can be represented as a continuum of numbers. The Wizard will prompt for value mappings for each element of the category. For example, the category "University" with members "SFU, UBC, UVic" cannot be represented as such, but the category "Rating" with members "Low, Medium, High" can be (one can map them as numbers 0, .5, and 1). Numerical categories are already ordered, and hence are not affected if they are declared as ordered.

Once you have loaded the small car database example with the Wizard, you may need to move the nodes around to view them better. Your network may look like this:

Once all mappings have been declared, the Wizard will create the specified neural network. Also, it will distribute the data into the training sets. You will want to move some examples to the test set. Look at Tutorial 1 on information how to do this.