How can I add one more hidden layer to ANN wizard?

I am using ANN wizard (nnstart) to run neural networks for fitting (prediction) purposes. According to the results, it appears adding one more hidden layer would produce a better predictability. However, the wizard doesn't have the capacity to add more hidden layers.
Could anyone please advise what I need to do to run a 3 layer ANN (2 hidden and 1 output) using the wizard?
Thanks

 Akzeptierte Antwort

Greg Heath
Greg Heath am 23 Nov. 2013

1 Stimme

A second hidden layer will not improve performance if your single hidden layer model is optimized.
size(input) = ? size(target) = ?
Type of network = fitnet?
No. of Hidden nodes= ?
NMSE = mse(target-output)/mean(var(target',1)) = ?
What is your NMSEgoal ? <= 1/100 ?

2 Kommentare

Bees
Bees am 29 Nov. 2013
Thank you very much for your prompt advice.
I have just started using Matlab for my project. I'm practising at the moment to get familiar with ANN on Matlab. I had seen in the literature that some people have used two-layer networks and some one-layer. So, I tried to check how I could use a two-layer if necessary.
I will soon be using my data and will provide you with full answers to get your valuable advice.
Before you do that try looking at some of my posted code
greg fitnet % etc
HTH
Greg

Melden Sie sich an, um zu kommentieren.

Weitere Antworten (1)

Bruno Pop-Stefanov
Bruno Pop-Stefanov am 22 Nov. 2013

4 Stimmen

I don't think there is a way to change the number of hidden layers using the wizard. Only the number of neurons in the hidden layer can be changed.
If you are okay not using the GUI, you can use network to create a customized neural network. This lets you specify the number of layers and customize every parameter of your neural network. See Classify Patterns with a Neural Network for details on using training and fitting command-line functions.

1 Kommentar

It is not necessary to design a custom net to use multiple hidden layers. For example
net = fitnet([ 5 5 ]);
view(net)

Melden Sie sich an, um zu kommentieren.

Kategorien

Mehr zu Deep Learning Toolbox finden Sie in Hilfe-Center und File Exchange

Gefragt:

am 22 Nov. 2013

Kommentiert:

am 5 Dez. 2013

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by