Implementation of stockastic gradient backpropagation learning algorithm
For a list of all members of this type, see BackPropagationLearningAlgorithm Members.
System.Object
LearningAlgorithm
BackPropagationLearningAlgorithm
[Visual Basic]
Public Class BackPropagationLearningAlgorithm
Inherits LearningAlgorithm
[C#]
public class BackPropagationLearningAlgorithm : LearningAlgorithm
Remarks
PROPAGATION WAY IN NN -------------------------> o ----- Sj = f(WSj) ----> o ----- Si = f(WSi) ----> o Neuron j Neuron i Neuron k (layer L-1) (layer L) (layer L+1) For the neuron i : ------------------- W[i,j](n+1) = W[i,j](n) + alpha * Ai * Sj + gamma * ( W[i,j](n) - W[i,j](n-1) ) T[i](n+1) = T[i](n) - alpha * Ai + gamma * ( T[i](n) - T[i](n-1) ) with : Ai = f'(WSi) * (expected_output_i - si) for output layer Ai = f'(WSi) * SUM( Ak * W[k,i] ) for othersNOTE : This is stockastic version of the algorithm because the error is back-propaged after every learning case. There is another version of this algorithm which works on global error.
Requirements
Namespace: NeuralNetwork Namespace
Assembly: NeuralNetwork.dll
See Also
BackPropagationLearningAlgorithm Members | NeuralNetwork Namespace