SVMs are basically linear regressions. Everything else you state is nonsense. In direct contrast to NNs (or heaven forbid optimessed SMA Xovers) SVMs are not as prone to overfitting due to their inherent linearity. In addition, SVMs generally outperform NNs in prediction errors.
Contrary to what you are saying, SVMs use parameters that must be optimized and essentially do not differ from a linear regression using dummy optimized variables -- I suppose you know what that means but you may not. I can prove that an SVM is equivalent to a linear regression with a suitable number of dummy variables optimized and this will be piece of cake using elementary algebra only. Alex
Alex, if you re-read my post you will note that I state that SVMs are linear regressions. My disagreement with you is with respect to overfitting - if that is what you meant by "excessive optimisation", especially with respect to NNs or optimised SMA crossovers. Anyway, the OP was interested in pattern recognition applications. If you feel that optimised SMA crossovers are the way to go, then you are entitled to your opinion.
I never figured out what NNs really do. It appears to me they are just a fancy way of optimizing indicators but I may be wrong. It is a fact though that their use peaked in the mid 90's and they represent nothing new and I dare to say nothing promising. I have APS and I use it mainly with daily data as intraday searches take too long and I don't have a spare fast PC to dedicate to that task at the moment. I tried EURUSD 15 and 30 min intraday searches with no success but I plan to revise the parameters I used and try again. Ron
Olexandr Topchylo ' Better ' - winner of the Automated Trading Championship 2007 - $130,475.45 in 12 weeks "You wrote on your Participant's page that you were using neural networks in your EA. How complicated is such an EA? What is more difficult to create - an optimal neural network or the Expert Advisor's algorithm ?" "A neural network that forecasts the changes in exchange rates and a trading algorithm are inseparable, they belong each other. Indeed, the Expert Advisor is very complicated. This trading system was first developed in Ã++, tested and optimized. Only after that, it was re-written in MQL4 and debugged." http://championship.mql4.com/2007/news/303
Another quote from that link: "What do you input in a neural network: indicator readings or a certain set of prices on a certain interval/timeframe?" "Readings of indicators developed especially for this trading system. I won't give more details now. The only thing I can say is that they combine some moving averages." Combine moving averages? Very original!
Interesting paper that deal with important issues ... Neural Networks in Financial Engineering: A Study in Methodology Apostolos-Paul N. Refenes, A. Neil Burgess, and Yves Bentz (Invited Paper) IEEE TRANSACTIONS ON NEURAL NETWORKS, VOL. 8, NO. 6, NOVEMBER 1997, pp. 1222-1267 http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=641449
My copy is version 4.9. It was a direct download off ( with password as far as I can remember ) the web plus pdf words. I don't really see how you can get a second hand copy.