Neural Network Toolbox

Network Design

Because the relationship between past and future values of the signal is not changing, the network can be designed directly from examples using newlind.

The problem as defined above has five inputs (the five delayed signal values), and one output (the next signal value). Thus, the network solution must consist of a single neuron with five inputs.

Here newlind finds the weights and biases, for the neuron above, that minimize the sum-squared error for this problem.

• net = newlind(P,T);
  

The resulting network can now be tested.

Applin1: Linear Design

Network Testing

© 1994-2005 The MathWorks, Inc.