[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [amibroker] Training the past



PureBytes Links

Trading Reference Links

At 05.58 17/02/03 -0500, you wrote:

>I think it depends on how weights are adjusted during training.

you're right, and it depends from network topolgy. For example, time lagged 
recurrent network use non-linear recurrent adaption of weights, and so 
training can get caught in local minima.
In fact, I always train this kind of network multiple times, because it's 
not unusal to get caught in local minima: for example, the first training 
process may use N1 epochs and the error is E1, while the second train may 
use N2>N1 with a better E2<E1.


Post AmiQuote-related messages ONLY to: amiquote@xxxxxxxxxxxxxxx 
(Web page: http://groups.yahoo.com/group/amiquote/messages/)

Check group FAQ at: http://groups.yahoo.com/group/amibroker/files/groupfaq.html 

Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/