|
PureBytes Links
Trading Reference Links
|
Dear Angel,
On Mon, 17 Jan 2000 19:50:48 GMT, you wrote:
>BTW. I forget to mention another technique that I use to "stabilize" (
>as a function of time) the parameters of the optimized sistems: before
>looking for the region in which the results are better, I make a
>smooth of the "surface". With this technique I avoid those regions
>that has very goop results but only for a small range of
>paremeters....
What kind of technique do you use for smoothing, if you have to deal
with several (e.g. five) parameters?
BTW: At least for my systems, it would be very dangerous in many cases
to apply any kind of smoothing to the "surface of the optimal
parameters" over time: Because profit is a highly nonlinear function
of my parameters, the best parameter sets are represented by several
(5-dimensional) peaks, where one of them is the highest at a given
time. When the character of the market changes, the highest peak may
decrease, and during an adaptive re-optimization step at a given time
another peak can be identified to be the highest now. That is, the
best parameter set is discontinuous over time and has to "jump" to its
new values. Smoothing would give very bad results in these cases ...
mfg rudolf stricker
| Disclaimer: The views of this user are strictly his own.
|