Predicting the future with Fourier transforms - page 46
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
There is another nuance here. The larger the next segment after optimization, the more likely it is that the harmonics found will quickly become obsolete (no longer yield profits) on future data. Reducing this section - we get unreliability of the test.
If I get the idea right...
It would be interesting to visually look at the picture - the trajectory in the phase space {optimal harmonic, optimal initial phase}. If the trajectory is smooth enough, it can be predicted.
It would be interesting to see a visual picture - a trajectory in phase space {optimal harmonic, optimal initial phase}. If the trajectory is smooth enough, it can be predicted.
There is. But there are some patterns that can be noticed when training a network, and some training techniques that allow you to do without even a forward test. I don't know about Fourier, and I haven't heard of it.
Rather it has to do with your personal experience with neural networks. Someone who has experience with another system might have some similar observations too.
If I understand the idea correctly...
It would be interesting to see a visual picture - a trajectory in phase space {optimal harmonic, optimal initial phase}. If the trajectory is smooth enough, it could be predicted.
Rather, one wonders if there is a way to determine the most stable harmonic. We have to assume that there is.
At the very least, people who do research into the possibility of making money in the financial markets don't go for Fourier, SSA or MESA. These are outdated methods that were being spun up and down by everyone about 10 years ago. It used to work well because the calculations using these methods were not widely available. Now, because of the availability of calculations and the release of various software products based on these methods, it does not work well, or rather it has become much more difficult to find a "profitable formula" for the market ))))
At least people who do research into the possibility of making money in the financial markets do not go for Fourier, SSA or MESA. These are old-fashioned methods that were being spun up and down about 10 years ago. It used to work well because the calculations using these methods were not widely available. Now, due to availability of calculations and production of various software products based on these methods, it does not work well - or rather, it has become more difficult to find a "profitable formula" for the market.)
It's more of a religious issue))) A neural network or a digital filter is a polynomial - the sum of the products of prices and coefficients (roughly speaking).
I agree. From this point of view, any price transformation is a price transformation )))) So it's all the same ))))
You're all going crazy when I mention that...
it seems someone,
like
knows how to make good money
on some kind of
maybe a modified method
probably Fourier.
Personally, I wouldn't risk investing in a situation with this degree of uncertainty.
And what would happen if I risked revealing the heart-breaking and extremely complicated truth?
What would happen then? Cries of "it can't be!", "no!", "the world can't be this cruel to us!", "I refuse to believe it!" ?
That's why I'd rather soak.
Note, colleagues, the real DSP (read Fourieux) experts - GPWR, Prival and a couple of others - are also practically silent here. Why? Because you can get burnt in all senses of the word.
By the way Fourier:
In my youth I did research on spectral analysis and detection of broadband noise-like radio signals of a suitor in a highly noisy and noisy environment.
I am now in the midst of thinking about the extraction of trading signals in forex noise. I considered using Fourier transforms. I came to the following conclusions.
Fourier transform (forward and reverse) is an excellent method of interpolation of electromagnetic processes. And only. Acoustic (mechanical) - with a stretch. The rest is questionable.
The fact is that in electromagnetic signal, electrical and magnetic energies are converted into each other, let me put it this way, equally, symmetrically. So it became possible to use complex variable models, in which the real and imaginary components are defined in orthogonal coordinates. Hence the appearance of the sinusoid as a projection of motion of a vector of constant length along the time axis inside a "complex cylinder". And the Fourier transform operates with a set of such harmonic components. That is, the Fourier transform has a practical value - it models one of the phenomena of nature: mutual transformation of electric and magnetic energies. This is confirmed, for example, in the fact that based on the results of calculation of the power spectral density, physical filters can be made, which will confirm the results of calculations with great accuracy.
However, it does not make sense in financial quotations to talk about any energies, much less two orthogonal, inter-transformed ones, so that complex variable functions can be applied to them. Therefore the value of the Fourier transform for the analysis of such quotes is as good or better than other interpolation methods. Alas, the "physical meaning" of financial quotes is unclear. Even visually, they cannot be attributed to harmonic signals.
As for extrapolating quotes using forward and reverse Fourier transforms, with intermediate filtering. The Fourier transform is a method of interpolating a signal with a set of harmonic components. And only in its nodes (samples). The interpolation accuracy between samples is not guaranteed. The desire to extrapolate a signal by this method, even for a few readings ahead, does not make physical sense, as spectral coefficients are calculated for given time readings. This is one reason. And the second reason has to do with the unclear physical meaning of quotes. If for extrapolation of electromagnetic signal we can count on its inertia (energy conversion) and apply low-frequency decomposition coefficients, then for quotes such "low-frequency" possibility is not obvious.
I am now considering calculating the current (instantaneous) spectrum for each minute (by tick) and displaying it on the quote chart in relief. The hope remains on the brain's ability to see any patterns in these pictures...
But Fourier's scientific supervisor, Lagrange, who considered Fourier's method to be a complete nonsense, was narrow-minded and insufficiently effective: