gobrain

The version 1.0.0 of gobrain includes just basic Neural Network functions such as Feed Forward and Elman Recurrent Neural Network. Using the framework, users are able to construct a simple Feed Forward Neural Network by first creating the XOR representation pattern to train the network.

The networks structure should contain a specific number of inputs, hidden nodes and outputs. The training should run for a given number of epochs e.g. 1000. The learning rate can be set to 0.6 and the momentum factor to 0.4. Users can receive reports about the learning error by using true in the last parameter.

The BackPropagate method is used when training the Neural Network to back propagate the errors from network activation. On the other hand, the return e method is used to train the Network; it will run the training operation for 'iterations' times and return the computed errors when training. After running the relevant code the network will be trained and ready to be used. The network can then be test run using the Test method.

Users can also use the method Update to predict the output given an input with the output being a vector with values ranging from 0 to 1. gobrain features Recurrent Neural Network, a library that implements Elman's Simple Recurrent Network. To take advantage of this, one can use the SetContexts function. They can create a single context initialized with a certain value or create custom initialized contexts. However, custom contexts must have the same size of hidden nodes + 1 (bias node).