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 (744)	6/6/1999 9:18:00 PM
From: CVDave	Read Replies (1) \| Respond to of 805

Jay, I've been gone so I've not been able to respond sooner. One of the obvious reasons why you might get better results with fewer neurons is that you are getting better generalization. This could be the case if you are using a relatively small amount of data. (Sorry, I don't know how much that is... it would depend on the specific case. Then, using more neurons would increase the risk of "memorization" which would be likely to result in poorer verify periods. As for the non-reproducibility, I'm surprised about that. The most likely issue there is training time. How long were you training when you were using a small number of neurons? Finally, the issue of the forget button is also surprising. One approach that I have taken when I was going to do multiple models is to "Save as" several different file names BEFORE doing any training. This allows you to use exactly the same starting point for your training experiments. Would that help your analysis? I'll try to get a little time to reproduce your experiments. Dave