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: Simon Thornington who wrote (765)	1/17/2000 1:10:00 PM
From: mark_andrews	Respond to of 805

Hi Simon, As the formula was given to me, it seems very clear you are correct. Add up all your influance periods for all stocks in the net. Let's call this number A. You have 2 nodes for your output and let's call this number B Neurons = SQR( A * B ) This is a rule of thumb. If you are using 2 stocks with a 20 day influance period, your Neurons would need to be 9. I have found this will keep the nets generalized. The problem I have found with NeuroStock and DeepInsight is they use Linear Regression for the input data of the Network. While this leads to perfect in sample trades, out of sample becomes inefficient. Best of luck, Mark

To: Simon Thornington who wrote (765)	1/17/2000 1:57:00 PM
From: Jay Hartzok	Read Replies (1) \| Respond to of 805

Simon, Sorry, I don't know of any books. Maybe Mark does. On neurons: I know very little about how a neural net functions. I know very little about memory or pattern detection. I only know what I've seen from training hundreds and hundreds of nets. I've tried Mark's theories on using the "proper" amount neurons based on the formula, and always ended up with less than desirable results with this program. I start all of my new nets at 50 neurons. I was going to say that I didn't think XLNX would train properly because it's out of it's historical trading range and the net doesn't have any relevant data in the training period that would give it any references to the current price range. But a few minutes ago, just out of curiosity, I started a net for XLNX with a 7 month verify period, and I just turned around and looked at the other computer, and surprisingly enough, the net doesn't look too bad. It does have a few errors in the verify period but it's only been training for 11 minutes, so I'm going to let it go and see how it turns out. Jay