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 : NeuroStock -- Ignore unavailable to you. Want to Upgrade?

To: Jay Hartzok who wrote (209)	10/9/1998 11:03:00 AM
From: Len Giammetta	Respond to of 805

Finally got a net on NEWZ that looks good. Been training for just over an hour but it looks like it's finally on the right track.

To: Jay Hartzok who wrote (209)	10/9/1998 1:58:00 PM
From: Len Giammetta	Read Replies (1) \| Respond to of 805

Jay and all... observation on influence settings of indices used for long term filters... I'm getting very accurate verifies when I set the influence settings for the index that more closely relates to the target to 1, and as the index becomes broader to 2 and 3. For example by setting the industry index to 1, the sector index to 2, and the general market indices to 2 or 3 has made an observable improvement in the verify. My models don't include any s/t filters, price or volume settings for the relateds but train quicker and give more accurate results than those that have. However, the net seems less profitable, as Optim has advised that it would. I continue to use the maximum number of available neurons. Since I find the whole issue of influence settings to be confusing, I welcome any opinions or observations that might lead to a "rule of thumb".

To: Jay Hartzok who wrote (209)	10/13/1998 12:16:00 AM
From: CVDave	Read Replies (2) \| Respond to of 805

Jay, After having another look at the MEH network which you sent, something is finally dawning, I think. Is it your interpretation of the inputs that there are actually a "choice" of 4 inputs--short term, long term, price, volume, where the first two are the PREPROCESSED items, understood to be functions of price, and the second two are raw price and volume data? Another way to say it would be that when you check "long term" and price, you are actually getting however many values are associated with LT and then, in addition the RAW price as well? (Or, if you check off LT, Price, Volume, it would be two RAW values?) I had been assuming (based on my reading of the help file) that LT required either Price OR Volume OR Both checked because it was a MA of either Price or Volume, depending on which one is checked. However, my interpretation of your MEH net and some of your earlier posts suggest that is WRONG...that instead, P, V are just extra inputs, in addition to the ones that get generated by ST and LT. Is this closer to what you believe to be the situation??? If so, it will really help me understand more about how you guys are constructing your nets--and why they can work. Thanks, Dave