multilayer perceptron - sagr4019/ResearchProject GitHub Wiki

multilayer perceptron

---

A multilayer perceptron is used to create a predictive model. The multilayer perceptron is build of several units or nodes, which are divided into layers.

We speak of input neurons hidden neurons to output neurons. The input neurons receive the data and the output neurons delivers the results.

Construction:

---

The MLP is also referred as a feed-forward network. This network is divided into layers. These layers consist of nodes, but each node of a previous layer is connected to each node of the subsequent layer. While calculating with such a network no layer can be skipped.

In general:

---

The MLP can solve the XOR problem by its multi-layered nature, which single-layer networks can not. The hidden neurons play the most important role. Each feature can work with non-linear networks, approximated with a hidden layer.

references

---

https://www.ibm.com/support/knowledgecenter/de/SSLVMB_23.0.0/spss/neural_network/idh_idd_mlp_variables.html https://en.wikipedia.org/wiki/Multilayer_perceptron http://campar.in.tum.de/twiki/pub/Far/MachineLearningWiSe2003/perzylo_ausarbeitung.pdf http://deeplearning.net/tutorial/mlp.html