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 (269)	10/19/1998 2:13:00 PM
From: CatLady	Respond to of 805

I also tried using only annealing on a net, it was so slow and I saw so little training that I gave up on that line of testing. The line from the help file on annealing says, Simulated Annealing - Enables Simulated Annealing training algorithm. This very powerful training algorithm is used to fine-tune the neural network. It is very slow, sometimes taking several days to train a big network. So my further exploration has been to see what benefit annealing has after some amount of backprop has already been run. Some of those nets are still running, so more later....

To: Jay Hartzok who wrote (269)	10/19/1998 7:15:00 PM
From: Len Giammetta	Read Replies (1) \| Respond to of 805

Missed the boat… Many of the stocks I had Neuro following have done extremely well over the last few days. Unfortunately I missed the big moves. Bad signals… no way! Just the opposite. The buy signals were given practically at the bottom, 5 to 10 days ago when the market posture was bleak. As a result I failed to trust the signals, convinced that Neuro was crazy. The rest is history.

To: Jay Hartzok who wrote (269)	10/21/1998 5:41:00 PM
From: CatLady	Read Replies (1) \| Respond to of 805

To anneal or not? I've tried this little experiment on 6 different stocks. After finding related indexes and influence settings that looked OK after 2 to 3 hours of backprop training, I made two copies of each net. With the A copy, I turned on simulated annealing training in addition to backprop. I trained the A copy until all 4 passes of annealing finished. With the B copy, training was completed using backprop only. Evaluating the results based on percentage of good trades during the verify period, showed one of the six was significantly better using annealing. One was equal and four were better with only backprop. So, for me, most of the time, the answer will be NOT.