Specify absence of nonlinearities for specific input or output channels in Hammerstein-Wiener models
unit=unitgain instantiates an object that specifies an identity mapping F(x)=x to exclude specific input and output channels from being affected by a nonlinearity in Hammerstein-Wiener models.
Use the unitgain object as an argument in the nlhw estimator to set the corresponding channel nonlinearity to unit gain.
For example, for a two-input and one-output model, to exclude the second input from being affected by a nonlinearity, use the following syntax:
m = nlhw(data,orders,['saturation''unitgain'],'deadzone')
In this case, the first input saturates and the output has an associated deadzone nonlinearity.
For example, for a one-input and one-output model, to exclude the output from being affected by a nonlinearity, use the following syntax:
m = nlhw(Data,Orders,'saturation','unitgain')
In this case, the input has a saturation nonlinearity.
If nonlinearities are absent in input or output channels, you can replace unitgain with an empty matrix. For example, to specify a Wiener model with a sigmoid nonlinearity at the output and a unit gain at the input, use the following command:
m = nlhw(Data,Orders,,'sigmoid');