Project NeuralBasic contain implementation of multilayer neural network. This kind of neural network is the most common used neural network. Below paragraphs describe the most importants neural networks parts.
More information about neural networks can be found:


Network has some number of inputs and outputs. Network has methods Work and Train. Network has collection of layers.


Layer has some number inputs and outputs. Layer has method Work and Train. Layer has collection of neurons. First layer input it is network input. Last Layer output it is neural network output. Layer(n) output it is Layer(n+1) input


Neuron has some number of input signals and one output. Neuron has method Work Train. Below sub paragraphs describe neuron parts.


Weighs it is the most important part of neuron and whole neural network. Weights it is memory of neural network. Weights values determine neural network output


Net value it is value calculated on the base neuron input and weight coefficients. The most common is used inner dot product of input vector and weight coefficient vector.

Activation function

Output from net value is passed through activation function. Tangent hyperbolic is the most common used.

Weight initialization

Initial weight value is calculated as random value from range (-0.1,0.1;)

Last edited Aug 31, 2010 at 6:26 PM by Mariusz, version 6


No comments yet.