This toolkit is designed to be used for super simple neural network use cases. It is not designed for any advanced AI aplications but should be used as a starting point, evaluation or proof of concept for further development.

The only teaching algorithm supported is RProp because of it's independance from the settings and universality.

The features of the toolkit:

- feedforward neural network with sigmoidal unipolar activation function
- automatic linear scaling of inputs and outputs to the linear range of the neurons for optimum performance
- option to enable last layer linear activation function
- option to add BIAS inputs to every layer of the network
- one hidden or two hidden layers
- RProp teaching algorithm
- teaching in chunks from a binary file
- option to evaluate error function over validation set
- saving and loading the network from file
- documentation
- examples of interpolating, predicting and classification using the toolkit
- basic error handling

NOTE - this is a super simple neural network and it should be used only in simple cases :)