[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kde-devel
Subject:    Re: Neural network window placement policy!
From:       Nicolas Brodu <brodun () aston ! ac ! uk>
Date:       1999-12-14 13:05:50
[Download RAW message or body]

Andreas Schlapbach wrote:
> 
> >From my experience with NN I know that they need a lot of training data
> (>500) in order to gain useful results. What are your expectations, when
> do you think your net will converge?

I don't know. For the initial training, I generated more than 20000 examples
by mapping a simple xwindow program and noting the input parameters and the
position returned by 'Smart' in a file each time. For the online training,
well, it can converge to something only if the user is consistent (and after
"some" time, I deliberately put a very small learning rate), and otherwise 
it'll just give not quite random answers...

> What about using a partial recurrent network (Jordan/Elman)? I think the
> previous position of windows has a _lot_ of influence on the placement
> of the next window. With the above mentioned networks this can be taken
> to account. IMHO MLP can't do so because they are feedforward only.

Because I don't know what a Jordan/Elman network is! I just implemented what 
we saw in the NN course, and we didn't learn those 2 structures (yet?).
Could you please give me information, or a pointer to it, concerning Jordan/
Elman networks (possibly off-list)?
And anyway, it would be better to take into account where the last window
ended up (possibly after the user moved it), and not only where it was
placed.
BTW, it reminds me that I should also take into account where the topmost 
window is...

Cheers,
Nicolas

-- 
Life is a sexually transmitted fatal disease. (W. Allen?)

[prev in list] [next in list] [prev in thread] [next in thread]