|
PureBytes Links
Trading Reference Links
|
Dans un courrier daté du 24/09/98 06:05:27 , David Powell écrit :
<<
Most of us have practical experience trading off indicators/theories and
building automated trading systems. Few have hands-on experience with
neural nets.
The average person starts trading visually off indicators. He'll find
an indicator that, by appearance says, "Hey, just buy it when I do this,
sell it when I do that, you'll get rich." After spending a lifetime of
hours on the best of the best indicators, systems, theories and money
management techniques - the average (really determined) guy will start
fumbling around with programming and backtesting. Then he finds out these
indicators/theories/systems don't really stand up month-in month-out (or
day-in day-out) and he doesn't have the stomach or the pocket book to stand
the heat. Deep down this is really hard to believe because VISUALLY,
looking back, they work. He may have spent twenty years getting to this
point. Now, what does he do?
The average guy doesn't know the definition of the word neural net,
fuzzy logic, artificial intelligence. The average guy's guess might be; "uh
. . . it takes the most recent market behavior and tries to form fit, or
keep in-step-with the market's data stream real time. . . ?" But he really
doesn't know if there's a difference between fuzzy/neural/artificial or if
they are all the same thing or more importantly, if it is a tool that can
help him get where he wants to be.
I'm hoping for a dialog by people that have real, practical experience
with neural-fuzzy, money management, automated system building experience,
etc.
- an average guy
>>
Due to your message and some others from the list ("OK, Pierre, now explain
and bring us some proofs, right now"), I'm currently devising a special web
page to summary what are the pro and cons of these methods from real world
long time experience.
I have bought and used the following NN products TradeStation compatible:
* N-Train (Scientific Consulting Services), the first available NN for
TradeStation (BackProp NN), then tried "Evolve" (same NN but by Genetic Algo)
* Braincel (Backpercolation NN)
* Neuroshell 2 (Ward systems )- and originated the first DLL interface between
TS and this NN product- (several NN algorithms, including Kohonen Nets,
recursive nets, GRNN and PRNN Nets).
* MAST (DataCaster and NetCAster), NN evolved by GA
|