A synapse is the connection between two layers of a neural network. The various synapse types define how layers will interact with each other. Some synapses contain a weight matrix, which cause them to be teachable. Others simply feed the data between layers in various ways, and are not teachable.
For a list of all members of this type, see ISynapse Members .
| Type | Description |
|---|---|
| BasicSynapse | An abstract class that implements basic functionality that may be needed by the other synapse classes. Specifically this class handles processing the from and to layer, as well as providing a name and description for the EncogPersistedObject. |
| DirectSynapse | A direct synapse will present the entire input array to each of the directly connected neurons in the next layer. This layer type is useful when building a radial basis neural network. |
| OneToOneSynapse | A one-to-one synapse requires that the from and to layers have exactly the same number of neurons. A one-to-one synapse can be useful, when used in conjunction with a ContextLayer. |
| WeightedSynapse | A fully-connected weight based synapse. Inputs will be multiplied by the weight matrix and presented to the layer. This synapse type is teachable. |
| WeightlessSynapse | A fully connected synapse that simply sums all input to each neuron, no weights are applied. This synapse type is not teachable. |
Namespace: Encog.Neural.Networks.Synapse
Assembly: encog-core-cs (in encog-core-cs.dll)