Spectrum derivative (or spectrum acceleration) - page 21

 
Trololo:

FOR THE ATTENTION OF THOSE INTERESTED IN THE THREAD

due to the fact that LeoV и tara I'm asking anyone reading this thread not to take any of their posts as valid or constructive. They may be deliberately obtuse to read their posts in the thread. I cannot stop them from doing so but I warn you, blame yourself later.))))))))))


Can you do Levitan's voice?
 
LeoV:

Can you do Levitan's voice?

Grandma, go in peace before you get caned.
 
Trololo: Grandma, go in peace before you get caned.

Pity..... would have been interesting to hear....))))
 
Trololo:

FOR THE ATTENTION OF THOSE INTERESTED IN THE THREAD

due to the fact that LeoV и tara I'm asking anyone reading this thread not to take any of their posts as valid or constructive. They may be deliberately obtuse to read their posts in the thread. I cannot stop them from doing so but I warn you, blame yourself later.))))))))))

Nah. They didn't do any harm at all. Slightly added some fun to a boring and uninteresting topic. It's hard to ruin a theme. It can only be improved.
 
Zhunko:
Nah. They didn't do any harm at all. They brought a bit of fun to a boring and uninteresting topic. It's hard to ruin a theme. It can only be improved.

What are you doing up, guarding the peace lines?

What? You're shaking, Russian.

that's right, shake. (с)
 

Here is a more precise expression (I read it on the forum): this is what I meant by compound frequencies, if applied to Fourier.


Any function with a finite spectrum can be decomposed into a Fourier series. And the point of prediction is not that just decompose it, then sum it all up and go back. There are a lot of Walsh decompositions, Wavelet, etc. You need to teach the program to pick out the components of the spectrum that determine the motion (the so-called useful component), all the rest is noise, remove (filter it out), then maybe something will turn out.

Extrapolation is based on the hypothesis of probable motion(s). And you can draw a curve in the future in any way you like. You can do it with Fourier, you can do it with polynomials, you can do it with your hands.

Therefore, a person (algorithm), when selecting these or those spectral components from the spectrum and predicting them in the future, gives them (these components) preference, as it believes it will determine the further movement. But is he right? On the basis of what research he has chosen the 1, 3 and 5 garnictics, each of which has its own frequency, amplitude and phase. Or maybe he should have chosen 2, 4 and 6 and phase tweaked ? or taken 256 spectrum components etc.

The primary hypothesis (idea) that gives a statistic about the probable motion. If you can calculate the probability of further movement with Fourier, you'll be fine, and if not, you'll be out of luck.


Z.I. Fourier works, works everywhere, cops radar you light and fine, receivers all listen, cellular phones we use, etc.

 

Trololo, I'm not an expert in Fourier, but I'll give you a few comments.

Trololo: Any function with a finite spectrum can be decomposed into a Fourier series. And the point of prediction is not that you just do it, then sum it all up and go back. There are a lot of Walsh decompositions, Wavelets, etc.

We can only talk about the properties of the spectrum after selecting the basis functions of the decomposition. The types of decomposition are determined by this basis.

So you can decompose it, but what's the use? A vast majority of "researchers" immediately start decomposing on a standard sin/cos basis without even understanding what it's about.

The first and most difficult question is to choose a functional basis for the expansion.

You need to teach the program to choose those components of the spectrum that determine the motion (the so called useful component) all the rest is noise, remove (filter it out), then you might get something.

Yeah, that's right. The main thing is not to throw the baby out of the tub.

Extrapolation is based on the hypothesis of likely movement(s). And you can draw a curve in the future in all sorts of ways. You can do it with Fourier, you can do it with polynomials, you can do it with your hands.

This is where the problem lies: the hypothesis also has to come from somewhere. And will this Fourier decomposition be necessary once you have a working hypothesis of probable motion?

And I don't see any other way to draw it apart from Fourier. What the hell is a polynomial? Well, of course, if you find such a functional space, in which polynomials are orthogonal and complete basis - then yes, draw the polynomials.

Or maybe you should have chosen 2, 4 and 6 and twisted the phase? Blah, blah, blah.

What phase, twist where? What are you talking about? Once you've decomposed the function into a Fourier series, the entire phase is just in the expansion coefficients. Well, tweak the coefficients, of course, but wisely.

Z.U. Fourier works everywhere, cops radar you and fine you, all the receivers listen, we use cell phones, etc.
.

Yes, not everywhere, but only there where the spectrum on a given functional basis is effectively bounded and at least quasi-stationary. But on trig functions as applied to finruns this does not seem to be true.

Have you found such a basis yet? Or are you going to step on the rake again by fiddling with sines/cosines?

 
Mathemat:


This is a Privalow excerpt, which I think is closer to my thoughts, but described in a more scientific way.

Well the basis is not the basis, but some of the stuff is. I'm thinking about how to do the sampling process after decomposing a particular hormone, in order to assemble it back from the sampled hormones.

 

The very notion of "harmonics" only makes sense when applied to a specific basis. Well, if you have this "stuff", then go ahead!

 
what about a sine wave with dynamic parameters