Edge effect on the way to the GRAAL - page 13

New comment
 
faa1947:

Spit on tolerance, a totally inappropriate thing in a technical and scientific discussion.

I do get annoyed when ignorance is prided and knowledge is substituted with aggressiveness. Only proven and provable facts and ideas are of interest, at least on forums not dedicated to shamans and magic.

So show me some knowledge. Instead of aggression. You keep promising, promising... I too am waiting with interest, by the way.
 
Diamant:
So show some knowledge. Instead of aggression. You keep promising, promising... I too am waiting with interest, by the way.
I'm waiting for the article to be published. But please acknowledge that my posts are not advertising, as I attach calculations to support my point of view wherever possible. Waiting.
 
If you want to get back to the discussion, I can describe how I avoid the edge effect. In a nutshell, calculating the expansion in a loop. I'd be happy to continue the discussion
 
faa1947:

Spit on tolerance, a totally inappropriate thing in a technical and scientific discussion.

I do get annoyed when ignorance is prided and knowledge is substituted with aggressiveness.

So don't substitute knowledge with aggressiveness and you won't get annoyed.
 
serega3491:
If you want to get back to the discussion, I can describe how I avoid the edge effect. In a nutshell, calculating the expansion in a loop. I'd be happy to continue the discussion.
Yes, if you don't mind describing it.
 
I got the idea to assign myself to the camp on Spider. But I also added some of my own. I do all procedures in MATLAB. 1) Firstly, it is necessary to perform stationary! Discrete wavelet decomposition. The edge effect still remains, but it is equal to the filter length. At first I used coiflets due to their smoothness. But then I went back to Dobeshi looking almost the same, but the latter have less coefficients to calculate. 2) Calculate the decomposition in a loop! This helps to get rid of the Gibbs effect, aka edge effect. The first clarification before decomposition is to increase the number of samples. WEXTEND with parameter sp0. I used Haar wavelets, aka first level Dobeshi. Whoever builds it will see that smoothing is not too big when making one loop. And here is the main thing - in decomposition parameters I choose wavelet and number of levels. In the loop, I specify 1 level, Smoothing by 5 cycles is better than 5 levels at a time. I can lay out the code by evening. From observations I can say that the characteristics are close to JMA.
 

http://www.jurikres.com/gifs/jma_anim.gif

and my MA

http://savepic.su/1888642.htm

 
serega3491:
The idea to assign myself to the mill - found on the Spider. But added a part of myself as well. I do all procedures in MATLAB. 1) Firstly, it is necessary to perform stationary! Discrete wavelet decomposition. The edge effect still remains, but it is equal to filter length. At first I used coiflets due to their smoothness. But then I came back to Dobeshi look almost the same, but the latter have less coefficients to calculate. 2) Calculate decomposition in a loop! This helps to get rid of the Gibbs effect, aka edge effect. The first clarification before decomposition is to increase the number of samples. I used Haar wavelets, aka first level Dobeshi. Whoever builds it will see that smoothing is not too big when making one cycle. And the main thing, you choose wavelet and number of levels in decomposition parameters. I specify 1 level in the loop, Smoothing by 5 cycles is better than 5 levels at a time. I can lay out the code by evening. I can tell from observations that they are close to JMA in terms of performance.
If you are already in Matlab, why don't you use the System Identification toolbox? It seems to me that it would complement the wavelets with estimation methods.
 
I'm sorry, can you be more specific? My Matlab skills are average. At this final stage, the only task left is to create entry and exit rules. I have a script for testing. Create an independent windows will also not be difficult. I will be grateful for your help. If you need-may lay out the code of the program.
 
serega3491:
I'm sorry, can you be more specific? My Matlab skills are average. At this final stage, the only task left is to create entry and exit rules. I have a script for testing. Create an independent windows will also not be difficult. I will be grateful for your help. If you need-may lay out the code of the program.
Login-exit rules are not a leftover, but a beginning. Check out my threads. A few words won't do it justice.
1...67891011121314
New comment