Adaptive digital filters - page 17

New comment
 
Prival писал (а) >>

If any of the programmers are available, I'm willing to share my work on Kalman (I think the soft sign in the name is redundant). It works in Matcad.

Garfish, thank you for the offer, but I am more interested in doing it myself, some of the results are very encouraging.

For me it may be a different approach, from the other side of the world. If that, write to ddd003(dog)mail.ru

 
I will try to describe the procedure here on the Matcadet 'Random Flow Theory and FOREX' step by step and by example, a little later. Although I seem to have explained everything in that thread, but it's probably too vague. It needs to be more compact. I wanted to write an article on this subject but I cannot get to it. I will try to do it in the nearest future. I will post it here.
 

Sergey, I have a favour to ask of you. If you are going to write an article, try to make a case for Kalman compared to other mash-ups. I see that you love it for a long time, but as applied to Forex, it seems to be unrequited. Admittedly, I myself seem to have lost ground under the comparison of mashups, as each can be good for its own purposes. Let this Kalman be computed entirely in Matcad, it's not that important. What is important is your vision of its application.

 

I haven't read this thread thoroughly...

I have a simple plan.

use DF.dll

Manually generate a filter for each pair and each timeframe.

i would like to use mql4 tools to create a spectrum analyzer and a logic block that defines the cut-off frequency

 

Please use http://www.fxexpert.ru/forum/index.php?showtopic=840&st=20&start=20 for this library.

What do you want to discuss?

 

I think there is one on the Alpari forum - frantsuz or something. Big fan of spectral analysis - and probably other French perversions.

 
I can't add a message to the place where I talked to the Frenchman. On the Alpari forum go to "Sharing Expertise in Spectrum Valuation" my nickname is the same, maybe you'll find something useful.
 

Yes, yes, I think I remember that you had a discussion with him there too, Sergey. By and large the problem is standard - non-stationarity of rows. But, let's say, for a multivariant the idea might be quite workable to choose the right inputs.

 

Thanks, I'll take a look at the link.

Actually I know how to use the library

I mean, maybe there's a special library for spectrum analysis...

but now i realized that i'm dumb (probably because i never worked in the field), i remembered the analyzer device - it's a narrowband filter that scans a range of frequencies

this library can be used for filter generation and analysis

So the question becomes obsolete.)

 

I have an idea, but it's been put aside for now, to take a closed currency cluster, sum it all up, do adaptive threshold processing in the spectral domain and through periodic compensation, then reverse FFT and watch where the group rate goes. At first glance, it seems to work, but we have to check it.

At the moment I am still finishing my adaptive filter, there is still some nonsense, I can't find any way to attach it to the indicator, in which period it would be possible to specify 1 min, 5 min, 15 etc.

1...101112131415161718192021222324...41
New comment