A Backpropagation Neural Network - executable, Eclipse and Maven projects

      Encoder


The Encoder program visualizes the output signals of the Backpropagation Neural Network showing the input layer at the bottom, the hidden layer in the middle and the output layer at the top. The following buttons can be used to control the program:


    is used for creating the network
  is used for initializing the network with random weights
  can be used to present the given patterns one by one
  starts training
  stops training


There is also an options menu to change the network settings patternSize, codeSize, randomSeed and eta.

The core of the program consists of four classes:

  • Input.java, Hidden.java and Output.java
    model the functionalities of the input, hidden or output nodes, respectively
  • EncoderNetwork.java
    models the dynamic aspects of the Backpropagation network

The graphical user interface is established by the other classes.

  • Class EncoderFrame.java contains the program's main method

Download links: