Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Strategies & Market Trends : Systems, Strategies and Resources for Trading Futures -- Ignore unavailable to you. Want to Upgrade?

To: Nemer who wrote (6700)	10/19/1998 11:50:00 AM
From: SE	Respond to of 44573

FWIW, I am back on sell signal. Looks like a good sell at 66 or 66.50 for a short hit. Let's see what happens..... EDIT - and just like that it was negated. If I recall correctly when my sell signals are negated this quickly without a noticable downturn, it usually is a bullish sign. Beats the heck out of me.... REEDIT- and now a buy signal. This is too tight a range for me. I just get whipped back and forth...

To: Nemer who wrote (6700)	10/19/1998 12:36:00 PM
From: Patrick Slevin	Read Replies (3) \| Respond to of 44573

<the neural net seminar> Guess I might as well, I can't seem to tag this entry just right. On the other hand, the seminar left me with mixed feelings which I have not sorted out thus far. It appears that when the markets jump into "unusual" activity, like from July through August, the models implode. This, is not something I wish to hear. Further, models should be created on a regular basis. One cannot just create a model and use it into infinity. There is a guy in S. America, for example, that has machines just creating new models that are so complex it takes a week of solid run-time just to spit out models that may or may not be suitable. This is one of the reasons I need another machine, I can't tie up the Gateway I have even overnight to spit out models. I'm not real nuts about generating them over the weekend either, as I usually wind down during that period of time. I need a break from this stuff once in a while, if for no other reason than to save my eyesight. Then, of course, the seminar was split between Ivory Tower types who desired to know every hyperbolic TAN in the programming and how the non-linear Diff-E-Q was implemented. How the Neurons were layered, the works. Fortunately, amongst some others, there was a fellow there with several advanced math and CS degrees who ALSO worked for a C.T.A.; so he was a good source. But I have a lot of work to do to bring my models up to his level. He also uses a method of automatically importing data and that avenue is not available to me. I have to physically download info and port it in to the database....when I told him that he stopped and said, "Gee, if I had to do that I probably would give it up". This, was not encouraging. Then too, there was so much ancillary stuff, like implementing Z-Score, that I really have to go back and re-read every copy of TASC for the past year or so to find specific data. Another problem....say you use variables in developing the models that were not in existence a few years back. VIX, for example, did not exist 4 or 5 years ago so implementing VIX as a variable means you cannot "train" the Net on data prior to '94 (unless you can re-create the data....in the case of VIX this is possible. Not so in all cases). Anyway, in some kind of effort to be brief, Neural Nets are not a quick and easy way to go. Using one requires a lot of work and a lot of tolerance for going down Dead Alleys. Unless a dedicated effort is made, I am certain one could try to build a successful model for years and never succeed. Further, even with a dedicated effort there are no guarantees and unusual market changes mean one has to constantly stay on top of the models. Changes in Environment also change Model effectiveness. For example, using data prior to '87 in a Net appears ineffective, as significant changes were implemented (like Circuit Breakers). So, I have mixed feelings at this point but I intend to try to move forward with it. I don't expect quick results in the near future, however.