About the NN Assembly

A neural network is a system of interconnected and interactive simple processors (artificial neurons). Neural networks are not programmed, rather they are learned. Ability to learn is one of the major advantages of neural networks in comparison with traditional algorithms. Technically, learning is finding coefficients of connections between neurons. While learning, a neural network is able to reveal complex interdependencies between input and output data, and also make generalizations. This means that in case of successful learning the network can return valid result based on the data that was absent in the learning set, as well as incomplete and/or "fuzzy", partially distorted data.

The NN assembly is used to create, learn and use neural networks. This assembly allows for simple and easy integration of functionality of neural networks into applications. It allows for successful analysis, data classification and forecasting.

The main assembly interfaces are:

INeuralNetwork. The interface for working with back-propagation networks and Kohonen self-organizing maps.

Key features of the NN assembly:

High working speed.
Support of multiple-level artificial networks (back-propagation networks and Kohonen maps).
Support of direct-propagation networks and networks with synapses that are user-defined (use of "shifted input").
Learning of a back-propagation network with customizable learning parameters.
Support of Kohonen self-organizing maps (Euclidean distance and algorithm of scalar vectors multiplication).
Various functions of network activation have been implemented: threshold function, hyperbolic tangent, exponential sigmoid.
Possibility of saving a network in a text file and subsequent restoration for further use or learning.