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: how to train dynamic neural network on multiple sequences
Replies: 5   Last Post: May 28, 2013 9:23 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: 5,965
Registered: 12/7/04
Re: how to train dynamic neural network on multiple sequences
Posted: May 26, 2013 2:19 PM
  Click to see the message monospaced in plain text Plain Text   Click to reply to this topic Reply

"Christian Vollmer" wrote in message <hjbq28$al5$1@fred.mathworks.com>...
> Hi,
>
> I'd like to train a layered recurrent network not only on a single sequence, but on multiple sequences.
> Suppose I have a set of input sequences (for example i1=[1 2 3] and i2=[1 2 2]). There is also a set of corresponding output sequences (for example t1=[2 3 4] and t2=[2 2 3]).
> One obvious way to train the network would be to concatenate all the input sequences to one big sequence i and all the target sequences to one big sequence t, then to convert them to cellarray via con2seq and then to train the network on that.


INCORRECT!

If there is no correlation between successive 3x1 sequences, then a dynamic network is inappropriate.

Instead, use a regression/curve-fitting net like FITNET or the obsolete NEWFIT.

The sizes of the input and target matrices should be

[ 3 N ] =size(input)
[ 3 N ]= size(output)

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.