Difference between FFNN and NAR

How is a normal feed forward neural network different from a NARNET? Is NARNET a RNN? What is the role of feedback delays in NARNET

Answer 1

[Greg Heath]

1. The NARXNET timeseries net is defined for

 a. Present and past (i.e., delayed) external inputs
 b. Feedback inputs from past (i.e., delayed) outputs

2. The special case TIMEDELAYNET timeseries net is defined for 1a only.

3. The special case NARNET timeseries net is defined for 1b only.

4. The FEEDFORWARDNET IS NOT A TIMESERIES NET. It is designed for time-independent input/output mapping.

a. Special case FITNET is designed for regression and 
 curvefitting.

b. Special case PATTERNNET is designed for classification and 
 pattern recognition.

The documentation for each of the above 6 nets can be obtained using the HELP and DOC commands. Fir example

 help narnet

 doc narnet

Hope this helps.

Thank you for formally accepting my answer

Greg

[shaifali gupta]

What is the difference in the architecture of feed forward net and a narnet? Please elaborate. How does NARNET architecture perform time dependent input/output mapping

[Greg Heath]

1. You did not acknowledge if you read and understood my answer.

2. You did not acknowledge if you read and understood the online documentation, examples and figures obtained from

 help fitnet  

 doc fitnet

(FITNET is the special case of FEEDFORWARDNET to be used for regression and curve fitting)

 help feedforwardnet

 doc feedforwardnet

3. You did not acknowledge if you ran any of the codes or looked at the resulting figures in the above documentation.

4. Please answer the above questions for NARNET, a special case of NARXNET.

5. As you can see from running the above examples, circuit diagrams results from the command

 view(net)

Hope this helps

Greg