What is the sturcture of the FITNET?

These days I apply a FITNET to design my own predictive controller. Indeed, it is very convenient to use. However, there is little concrete content about the structure of this net in the MATLAB documentation. Could someone answer me these sub-questions: (1) Is it similar to a traditional BP net? And what is the difference between them, activation functions? (2) If it has the same structure as the BP net, why is it named FITNET? The 'Tips' in the documentation informs that "Function fitting is the process of training a neural network on a set of inputs in order to produce an associated set of target outputs." Does it mean the FITNET just adopts a new way to train a BP net?

Answer 1

[Greg Heath]

FITNET is the default net for curveFITting and regression.

It is the standard configuration that comes to mind whenever anyone says

            NEURAL NETWORK

for details use the commands

 help fitnet

 doc fitnet

and

 type fitnet

For classification or pattern recognition the default is PATTERNNET.

CLARIFICATION:

If you run the code in the

 help fitnet

and

 doc fitnet

you will see the circuit diagrams.

FITNET is a standard MLP with one hidden layer and a linear layer output. It is used for curve"fit"ting and regression.

PATTERNNET is a standard MLP with one hidden layer and a binary layer output. It is used for classification and "pattern" recognition.

Thank you for formally accepting my answer

 Greg

[Huang Guoming]

Thanks for your answer, but it seems that I cannot get what I want from your answer.

[Greg Heath]

FITNET is just the typical standard neural network used for curveFITting and regression.

whereas

PATTERNNET is the typical standard sigmoid neural network used for PATTERN recognition and classification.

If that isn't good enough, rephrase your question.

[Huang Guoming]

I want to know the detail structure of the FITNET, not its function.

[Greg Heath]

Comment by Huang Guoming on 26 Oct 2018 at 2:28

I have known these ways to find information about the FITNET in the Documentation of MATLAB. Indeed, that Documentation does not clearly explain the structure of the FITNET.

[Greg Heath]

If you run the code in the

 help fitnet

and

 doc fitnet

you will see the circuit diagrams.

FITNET is a standard MLP with one hidden layer and a linear layer output. It is used for curve"fit"ting and regression.

PATTERNNETNET is a standard MLP with one hidden layer and a binary layer output. It is used for classification and "pattern" recognition.

[Huang Guoming]

Thanks, standard MLP is what I want.

[Tilak Nanavati]

Hey Greg, can you please tell somthing on default activation function used in FITNET? Thanks.