|
PureBytes Links
Trading Reference Links
|
Right, only what we are searching for, is the Neural Net to really become
predictive based on historical information. And really, isn't that what any
system must do, analyses the past to predict the future. Otherwise the whole
thing is a crap shoot. The NN does this (tries to be predictive based on
past results). What seems to fail others who have tried NNs is that they are
not breaking up the data into various configuration and point of views. The
NN needs to see data from a variety of intelligently built and carefully
designed data presentations.
Treasure
-----Original Message-----
From: Gene Pope [mailto:gene@xxxxxxxxxxxxx]
Sent: Saturday, December 14, 2002 5:53 PM
To: omega-list@xxxxxxxxxx
Subject: Re: Limited life span of mechanical systems?
>
> A Neural Net working in conjunction with the correct set of data, which
> would need to present to the NN a wide variety of input data
configurations
> (about 30 configurations of input data), and most important, providing a
> feedback loop of historical performance versus desired performance, one
> should be able to create a self optimizing system.
>
Self-optimizing in terms of resolving the past... perhaps.
With all due respect, a feed-BACK loop is no different in system trading as
in audio circuit design IMHO, it still does not solve the problem of
"prediction". An audio feedback loop does not know what "error" it will
encounter in the future, it merely corrects for this moment (and therefore
all historic moments)... and, there is lag in such a system... and, you
can't make money trading a single moment without some idea (pattern or
statistical or otherwise) what future moments might hold.
Best regards,
Gene Pope
|