Multilayer Perceptron - A Java Implementation v1.0

MULTILAYER PERCEPTRON - A JAVA IMPLEMENTATION v1.0
by Aydin Gurel

What is this?
Feed forward networks are the most widely used and the most flexible kind of neural net. This work is a java implementation of multilayer perceptron nets. Using this set of classes, you can build multilayer perceptron nets with any number of layers and any number of units very easily.

What can you exactly do with it?
You can:
- Build multilayer perceptron nets with any number of layers and units. Layers are connected to each other consecutively, each unit in a layer is connected to all of the units on the next layer (and vice versa) if there is one,
- Set units with linear and sigmoid activation functions and set them separately for each layer,
- Set parameters for sigmoid functions and set them separately for each layer,
- Use momentum, set different momentum parameters for each layer,
- Initialize the net using your own set of weights,
- Train the net using back propagation and with any training rate.

Other information:
Codes were tested using jdk 1.1.8.
Outputs of the sigmoid units are between 0 and 1

Tlu.class is class of the neural unit objects. Used by Layer.class and Mlp.class
Layer.class is the class of the layer objects. Used by Mlp.class
Mlp.class is the class of the neural net object
Programming is object oriented and Tlu.java can be used separately to represent the units of your own net
You can download Tlu.java, Layer.java and Mlp.java

Example:
This is an example of creating a net with four layers including the input layer, training it for the cosine function. You can download Example1.java. To have it work, you have to download Tlu.java, Layer.java and Mlp.java as well.

Home