Drexel dragonThe Math ForumDonate to the Math Forum



Search All of the Math Forum:

Views expressed in these public forums are not endorsed by Drexel University or The Math Forum.


Math Forum » Discussions » Software » comp.soft-sys.matlab

Topic: neural network
Replies: 9   Last Post: May 22, 2014 4:17 AM

Advanced Search

Back to Topic List Back to Topic List Jump to Tree View Jump to Tree View   Messages: [ Previous | Next ]
Greg Heath

Posts: 5,978
Registered: 12/7/04
Re: neural network
Posted: May 22, 2014 4:17 AM
  Click to see the message monospaced in plain text Plain Text   Click to reply to this topic Reply

"isha" wrote in message <lkvij9$lmv$1@newscl01ah.mathworks.com>...
> "isha" wrote in message <lkst9a$gk5$1@newscl01ah.mathworks.com>...
> > "Greg Heath" <heath@alumni.brown.edu> wrote in message <lk091v$gcu$1@newscl01ah.mathworks.com>...
> > > "isha" wrote in message <ljvco8$60e$1@newscl01ah.mathworks.com>...
> > > > "Greg Heath" <heath@alumni.brown.edu> wrote in message <ljuuvo$5h9$1@newscl01ah.mathworks.com>...
> > > > > "isha" wrote in message <ljsm18$o04$1@newscl01ah.mathworks.com>...
> > > > > > "Greg Heath" <heath@alumni.brown.edu> wrote in message <ljrep1$ffc$1@newscl01ah.mathworks.com>...
> > > > > > > "isha" wrote in message <ljq408$k66$1@newscl01ah.mathworks.com>...
> > > > > > > > hello
> > > > > > > > I am optimising parameters like thickness of layer as output for return loss and material parameters as input.After training of the network, when i simulated it, output was crossing 2.I am need to apply constraint to the output i.e. thickness should not be greater than 2mm..
> > > > > > > > Please help me in this context.
> > > > > > > >
> > > > > > > > Regsrds
> > > > > > > > Isha

> > > > > > >
> > > > > > > Change your output transfer function to 'tansig', which is bounded, and use the smallest number of hidden nodes as you can.
> > > > > > >
> > > > > > > Hope this helps.
> > > > > > >
> > > > > > > Greg

> > > > > > thanks for the suggestion...
> > > > > > Can i use Logsig function as it is also bounded..

> > > > >
> > > > > It is not commensurate with the default maxmin [-1 1 ] normalization. If you want to
> > > > > use logsig then cancel the default and normalize to [0 1 ]before you train.
> > > > >
> > > > > Greg

> > > >
> > > > hello
> > > > Thanks for the suggestion.i need one more help how to bound the input.actually when i optimised the parameter using neural network for a given return loss and then simulated it using a software..but result doesn't match..

> > >
> > > MATLAB uses the mapmaxmin default BUT provides for a mapstd alternative. However I suggest checking for outliers BEFORE training. I use zscore (mapstd is an alternative) to do this. Once done, I have found it easier to not worry about the MATLAB default instead of turning it off.

> > hello
> > thanks for the suggestion.

>
> Sir
> i have a question that is it possible to select best material out of a database suitable for our application through neural network?


1. Make sure all of your data is valid. (e.g., use mean, std, minmax and plot).
2. Obtain multiple designs for each value of H, the number of hidden nodes, that you try.
3. For each design randomly choose the training/validation/test subsets and the initial weights.
4. Use as many default parameters as possible before trying other values.

See my examples in the NEWSWGROUP and ANSWERS

greg fitnet Ntrials
greg fitnet Hmax

Hope this helps.

Greg



Point your RSS reader here for a feed of the latest messages in this topic.

[Privacy Policy] [Terms of Use]

© Drexel University 1994-2014. All Rights Reserved.
The Math Forum is a research and educational enterprise of the Drexel University School of Education.