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

To: Bill Scoggin who wrote (659)	9/11/1999 9:55:00 AM
From: Optim	Read Replies (1) \| Respond to of 871

Hi Bob, Been a while since I talked with you...I have a question about the FFT and Wavelets I've been much less active in my posting these last few months, but I am still around and I am still experimenting. I have been trading the odd model here and there, but not with any regularity. That may change if these Wavelet experiments hold up though. Is what you are experimenting with concerning Wavelets basically the same as how you analyze an electrical signal using Fourier methods...ie, you assume a certain time period's waveform is a period cycle and as such is composed of some combination of harmonic sine and cosine waves with a constant component added in... I think so. Like I say, I am still new to all of this, but from what I gather Wavelets play a similar role to FFT. The allow you to isolate certain frequencies of repetitive cycles in your series. For example you could tune a Wavelet to act much like a bandpass filter, cutting out all frequencies above and below certain thresholds. By doing this to simple OHLC inputs you isolate and emphasize the underlying cycles in the data. It is very dependant of the shape or type of wavelet, as they all have different properties. I've read somewhere that the Morlet and Gabor wavelet series are the best for financial series, although I haven't actually gone so far as to verify it myself. They have performed very well so far though. You can read more about the wavelets (it's actually an add-on to NS Trader, Excel and Tradestation) here: corniceresearch.com If so, did you say that NS Trader has this feature built in? NS Trader doesn't have the Morlet or Gabor wavelets, but it does have the Daubachies and Haar wavelets. These are two of the most common wavelets and can be found in most beginners textbooks on wavelets. In addition Trader supports FFT and some PCA for dimension reduction. PCA hasn't worked well in my testing, although it does help the genetic algorithms to converge a little quicker. I imagine this is from the decorrelation aspects. There are some good books recommended on the Cornice Research site that deal with using NNs for signal processing. This is similar to the speech recognition you mentioned. There are also some good papers available on the net if you do a search. One other great place to begin is at: amara.com. Optim