INeuralNetwork.NormalizeInputValues

Syntax

NormalizeInputValues: Boolean;

Description

The NormalizeInputValues method normalizes network input data.

Comments

The method returns the execution result. Available values:

Input data is normalized by the formula:

(Vi = Vi / ||V||)

Example

As an example, a function is given that has neural network fed into its input (the Net parameter) and the logical parameter FirstLayer that determines whether to normalize values of synapse weights only for the first layer of the network. If FirstLayer is set to True, values of synapse weights are normalized only for the first layer of the map. If FirstLayer is set to False, all the weights are normalized.

To execute the example, add a link to the NN system assembly.

Function n_Normalize(Net: NeuralNetwork; FirstLayer: Boolean): NeuralNetwork;
Begin
    If Layer Then
        // normalize synapses only of the first layer of the network
        Net.NormalizeSynapsesEx(0);
    Else
        // normalize all network synapses
        Net.NormalizeSynapses;
    End If;
    // normalize input data
    Net.NormalizeInputValues;
    Return Net;
End Function n_Normalize;

After executing the example input data and values of synapse weights are normalized.

See also:

INeuralNetwork