Patterns are used to create common sorts of neural networks. Information about the structure of the neural network is communicated to the pattern, and then generate is called to produce a neural network of this type.
For a list of all members of this type, see INeuralNetworkPattern Members .
| Type | Description |
|---|---|
| ElmanPattern | This class is used to generate an Elman style recurrent neural network. This network type consists of three regular layers, an input output and hidden layer. There is also a context layer which accepts output from the hidden layer and outputs back to the hidden layer. This makes it a recurrent neural network. The Elman neural network is useful for temporal input data. The specified activation function will be used on all layers. The Elman neural network is similar to the Jordan neural network. |
| FeedForwardPattern | Used to create feedforward neural networks. A feedforward network has an input and output layers separated by zero or more hidden layers. The feedforward neural network is one of the most common neural network patterns. |
| HopfieldPattern | Create a Hopfield pattern. A Hopfield neural network has a single layer that functions both as the input and output layers. There are no hidden layers. Hopfield networks are used for basic pattern recognition. When a Hopfield network recognizes a pattern, it "echos" that pattern on the output. |
| JordanPattern | This class is used to generate an Jordan style recurrent neural network. This network type consists of three regular layers, an input output and hidden layer. There is also a context layer which accepts output from the output layer and outputs back to the hidden layer. This makes it a recurrent neural network. The Jordan neural network is useful for temporal input data. The specified activation function will be used on all layers. The Jordan neural network is similar to the Elman neural network. |
| RadialBasisPattern | A radial basis function (RBF) network uses several radial basis functions to provide a more dynamic hidden layer activation function than many other types of neural network. It consists of a input, output and hidden layer. |
| RSOMPattern | A recurrent self organizing map is a self organizing map that has a recurrent context connection on the hidden layer. This type of neural network is adept at classifying temporal data. |
| SOMPattern | A self organizing map is a neural network pattern with an input and output layer. There is no hidden layer. The winning neuron, which is that neuron with the higest output is the winner, this winning neuron is often used to classify the input into a group. |
Namespace: Encog.Neural.Networks.Pattern
Assembly: encog-core-cs (in encog-core-cs.dll)
INeuralNetworkPattern Members | Encog.Neural.Networks.Pattern Namespace