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: CatLady who wrote (275)	10/21/1998 8:57:00 PM
From: Len Giammetta	Read Replies (2) \| Respond to of 805

CatLady, through trial and error, testing not as controlled as yours, I've run similar nets with the same results. Anneal just doesn't seem to add significantly to the predictive ability of Neuro. Also, nets that I've modeled with a year or less have been much less accurate than nets with those with 2 yrs or more of data. Now if we could just figure out the "perfect" influence settings... Good work CL.

To: CatLady who wrote (275)	10/22/1998 9:49:00 AM
From: Optim	Read Replies (3) \| Respond to of 805

Hi Catlady, Thanks for the response. The 3 year version trained much faster. Has anyone else tried this? What did you find? After playing with NS some more (a total of 40 nets) only four worked the way I had previously mentioned. I actually much more success extending the training time as Jay, Len and yourself have mentioned. I apologize if I lead anyone offtrack with my previous recommendation and I will be sure to scrutinze my results more thoroughly in the future. Having said that, has anyone experimented with the number of neurons? I have trained a number of nets using less than 20 and find that within 20 minutes I know if a net is good or not. Out of sample results for a 4 month period are excellent, which gives me confidence in retraining up to the current date and trading on the nets recommendations. It also has a major benefit: significantly decreased training time. I will be experimenting with increasing the number of neurons for a number of nets watching the out of sample period for any indication of memorization. Again, until I have concrete answers I will keep my results to myself! ;-) Optim