GetSynapse(layerIdx: Integer; neuronIdx: Integer; synapseIndex: Integer): Double;
GetSynapse(layerIdx: integer; neuronIdx: integer; synapseIndex: integer): double;
layerIdx. Network layer index.
neuronIdx. Neuron index in the layer.
synapseIndex. Synapse index in the neuron.
The GetSynapse method returns value of the specified synapse weight.
Indexation of the layers, neurons, synapses is continuous and starts with zero.
Maximum value of synapseIndex can be incrementally bigger per unit than the number of input synapses for the error back propagation networks. A synapse with maximum index is a special "shifted input" of a neuron. For example, there is a network with error back propagation that consists of a single layer. A layer consists of three neurons. Each neuron contains two inputs. Thus, each neuron has only two user inputs (with the indexes 0 and 1) and one unchangeable "shifted input" (with the index 2). The input is shifted only once during network initialization. The shifted input is not used in the network learning, but is rather used during its calculation.
The method use is given in the example for INeuralNetwork.CreateNetworkEx.
See also: