NeuroSolutions' object-oriented design breaks a neural network down into neural components. This structure provides the ability to simulate any neural network belonging to the additive model.
The best way to illustrate the flexibility provided by NeuroSolutions is to take a look at the breadboard below. This is a hybrid supervised-unsupervised topology. The unsupervised segment uses the Ojas' unsupervised rule to preprocess the data. The processed data is used to feed a modular feedforward network. Each branch of this network has a recurrent connection that provides feedback to the previous layer.
