Neuroph

Creating a neural network is not something that can be done by any amateur with a computer. Fortunately, the procedure can be learned and Neuroph can make it quite easy. The software can be used to create common neural network architectures, using a library of classes for basic neural network concepts.

The application is lightweight and it takes about a minute to install it. It is written in Java, so you can use it on Windows, Mac OS X and Linux. Neuroph is distributed under the Apache 2.0 license, which means that you can use it free of charge for as long as you want.

Although its interface looks complex and confusing at a first glance, Neuroph was created with ease of use in mind. Its main purpose is to simplify the neural network development process. The software is an excellent choice if you wish to learn how neural networks work or if you need some material for a research project. You can find numerous guides on the developer's web site, which cover mostly everything, from starting a new project to creating a complex neural network. Every function is well documented, so you should not get stuck while working on a project.

You can either create a new project from the ground up or open one of the program's samples. Once you start a project, you must create new files, for neural networks, training sets and various processes, like image recognition, text or handwritten character recognition and others. A wizard will guide you through different settings, depending on what you intend to create. For example, you may create various types of neural networks. Adaline, Perceptron, Hopfiled, supervised or unsupervised Hebbian are just a few examples.

When creating a neural network, you can specify a number of input and output nodes and select a learning rule. You can also create training set files and specify your desired input and output values as training elements. After creating a neural map and a training set, you may select various options, on the main interface, to start training your neural map, test it for the entire training set or with single input values and perform many other actions.