Next: Configuration Previous: Neural Network


Back Propagation Algorithm

In the back propagation network I'm using there is one input layer, one hidden layer and one output layer. The input's are called the 's, the hidden nodes are called 's and the output as . The weights interconnecting between the input and the hidden nodes are called and the weights interconnecting between the hidden and the output nodes are called .

The value of lambda is fixed at 5.0

is the activation of the output.

is the activation of the hidden node.

The error function is given by the following equation.

where is the target and is the output.
The weights are updated by the following way.

is a small correction factor being used whose value is fixed at 0.1.

The last two equations are used to update the weights.


gsm@ee.duke.edu
Sun Jan 14 10:17:29 1996