Neural network is a group of mathematical equations. That equations can modify themself to get better results, therefore we say that neural network is an artificial intelligence.

Neural networks have significant application at matching to pattern, forecast some phenomena and when standard mathematics do not give expected results or is too complicated.

Neural network consists of neuron, inputs and outputs. Number of neurons, inputs, ouptuts, connections and how neurons are connected have influence to result of that natwork (how result is similar to expected result).

Each neuron has multiple inputs and one output. Mathematicaly it realize equation that is presented at picture.

Neural network connect neurons, inputs and outputs each other. At our web there are two kinds of connections: normal connection, and broken connection. See chapter connections for details.

There are two kinds of connections: normal connection, and broken connection. Normal connection transmit input signal without changing it, broken connection change it. Mathematical equations for that operations are shown in the picture

Notice that not every neural network can learn that what we want to learn. It depends from number of neurons and connection between them. The simplest example is, when we wont to learn that result will be 1, but all inputs are 0 and all offsets are 0. The simplest way to resolve this problem is add input that is always 1.


This simulator uses 'backpropagation' method to learn neural network. This method was detailed described by my friends.

Mariusz Bernacki, Przemysław Włodarczyk: 'Backpropagation alghoritm' (pl)