Neural Networks Tutorial

<< previous

answermath HOME

answermath en Español

NEXT >>

Artificial Intelligence Tutorials: Data Mining ▫ Neural Networks ▫ Fuzzy Logic ▫ Genetic Algorithms

Tutoriales de Inteligencia artificial: Redes Neuronales ▫ Lógica Fuzzy ▫ Minería Datos ▫ Algoritmos Genéticos

Mental math Tips: Addition ▫ Subtraction ▫ Multiplication ▫ Division ▫ Sine ▫ Cosine ▫ Tangent

Tips cálculo mental: Suma ▫ Resta ▫ Multiplicación ▫ División ▫ Seno ▫ Coseno ▫ Tangente

The structure of the Net: Interconnection of basic elements.

It's the way the basic units communicate their output to other unit's inputs. They are usually arranged in layers so that the outputs of a layer are fully connected to the inputs of the next layer; in this case we have a forward fully connected network. It's possible to work with partially connected units or with feedback connections from outer layers to inners ones (this has nothing to do with back propagation).

In order to achieve acceptable results the number of layers must be at least three. There is no evidence that five layers may solve problems that four-layer structure does not. So the amount of layer uses are to be three or four.


	Web http://www.answermath.com