SI
SI
discoversearch

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 : Neural Nets - A tool for the 90's -- Ignore unavailable to you. Want to Upgrade?


To: Optim who wrote (82)6/8/1998 12:30:00 AM
From: Len Giammetta  Read Replies (1) | Respond to of 871
 
The initial training is long. The only way I've found around it is to maintain a manageable portfolio, 7-10 stocks. Once the training is completed I simply update, and retrain 15-30 minutes each, that's usually plenty, and I can do other things in the meantime. By the way, NS states they have an accelerator card that reduces the training time from days to minutes. I've written and asked for the price and availability but didn't get a response yet. I'll post if I get a respnse.

Although it may seem like the maintenance of the portfolios may be very time consuming, I've found I actually spend much less time. I used to spend hours a day scanning, checking fundamentals, reading threads, checking moving averages, crossovers, etc, but now I spend much less time and the results are much more profitable.

Anyway, good luck to all of you!



To: Optim who wrote (82)6/9/1998 3:21:00 PM
From: Optim  Read Replies (1) | Respond to of 871
 
Just thought I would share one of my successes with Neurostock. I recently set up a model for NN (Newbridge Networks) and included as inputs the SP500, CSCO, ASND, COMS, YHOO, IXTC, IXTO, $INX, and a few others. I trained the network for 20 hours on Backprop, then I switched to Genetic for 400 hours, and then cleaned up the net by running Simulated Annealing for 50 hours. The resulting net has pretty good predictive power and has given some good signals over the last month or so. I set it up so that new data is downloaded daily and the net is trained for about an hour. Then I make a prediction and e-mail it to myself. When I get into the office in the morning I have my buy/sell/hold recommendations waiting for me.

My goal is to do this for about 5-10 stocks. My only gripe is that it take a long time to come up with a good net. NS Trader is great for speed, but requires significantly more development time.

Optim