Recurrent neural network

Or as a recurrent neural network feedback is called neural networks, which in contrast to the so-called feedforward neural networks to characterized the same or a previous layer by means of connections of the neurons of a layer. In the brain, this is the preferred Verschaltungsweise neural networks, in particular in the neocortex. In artificial neural networks, recurrent interconnection of model neurons are used to detect temporally encoded information in the data. Examples of such recurrent neural networks are the Elman net, the Jordan network, the Hopfield network, as well as the fully -connected neural network.

Recurrent networks can be subdivided as follows:

• In a direct feedback ( engl. direct feedback ) of the own output of a neuron is used as an additional input.
• The indirect feedback (English indirect feedback ) connects the output of a neuron to a neuron of the previous layers.
• The lateral feedback (English laterally feedback ) connects the output of a neuron with another neuron same layer.
• For a complete connection of each neuron output has a connection to every other neuron.

Learning of recurrent neural networks

Recurrent artificial neural networks are difficult to train through machine learning methods. A popular approach is, therefore, not to train the network, but the read-out of the network. The recurrent neural network is regarded as a so-called reservoir.