What to feed to the input of the neural network? Your ideas... - page 82

New comment
 
Рaра Нoth #:
It is possible to memorise history without these perversions, not even close. Look at how sound is converted, by tweaking some decibels according to a formula, it is parsed into hormones, i.e. pure frequencies are obtained from which it consists. This is approximately the same thing, but neurons do everything and they parse input data in this way, selecting the correct interpretation for each group, comparing them with each other and with the history of quotations, trying to catch the correct logic between all of this

You don't seem to have understood what was written. I think the answer was about using 2 models, not about decomposition of a series. These are different things.

 
Рaра Нoth #:
I did not do it, but I saw the result with my own eyes, and caught the whole essence at once, to believe or not your problems.
Financial time series is not a sound and it cannot be decomposed into harmonics. A pure note sounding only 1 sec is 500-1000 oscillations of a periodic signal (approximately a sine wave). Only a periodic signal (which repeats hundreds or thousands of times and does not change in time) can be decomposed into harmonics.
Have you ever seen waves with the same period and amplitude 1000 times in a row in a financial series? You can get rich on a sine wave by the 10th period. And by the 1000th, all the money in the world will be in your pocket.
Everything here is random, sometimes 1 wave, sometimes 3, and for a change 5 waves can slip through, so that later you can get all those who entered on such a flat.
 
Forester #:
Financial time series is not a sound and cannot be decomposed into harmonics. A pure note sounding only 1 sec is 500-1000 oscillations of a periodic signal (approximately a sine wave). Only a periodic signal (which repeats hundreds or thousands of times and does not change in time) can be decomposed into harmonics.
Have you ever seen waves with the same period and amplitude 1000 times in a row in a financial series? You can get rich on a sine wave by the 10th period. And by the 1000th, all the money in the world will be in your pocket.
It's all random, sometimes 1 wave, sometimes 3, for a change and 5 can slip through, so that later you can take everyone who entered on such a flat.

Well, in general, you can decompose any series, but if it is not periodic, it is not necessary to use decomposition for its intended purpose.

 
Maxim Dmitrievsky #:

Well, in general, you can decompose any series, but if it is not periodic, it is not necessary to use the decomposition for its intended purpose.

What is the point of decomposing it if it is not periodic?
Each period window will have a different shape of the graph (signal for some). Accordingly in each period after decomposition of the random graph there will be a random set of frequencies, phases and amplitudes. Neither there nor there patterns can be found.
On a periodic signal the period is easily determined, but here there is no period, i.e. a window of random size (period duration).

 
Forester #:

What is the point of decomposing it if it is not periodic?
Each period window will have a different shape of the graph (signal for some). Accordingly, in each period after decomposition of the random graph there will be a random set of frequencies, phases and amplitudes. No regularities can be found there or there.

You can try :)

 
Forester #:

What is the point of decomposing it if it is not periodic?
Each period window will have a different shape of the graph (signal for some). Accordingly, in each period after decomposition of the random graph there will be a random set of frequencies, phases and amplitudes. Neither there nor there patterns can be found.
On a periodic signal the period is easily determined, but here there is no period, i.e. a window of random size (period duration).

That's right. SB often draws something similar to periodic oscillations. The main difference from real periodic oscillations is that the calculated frequency depends significantly on the width of the window.

 
Is there any other way to organise a filter with an arbitrary phase shift relative to, say, intraday volatility? Or just a time coordinate. Of course, we can take as a basis a certain point on the chart, after which we start dividing it into equal segments, but then how to choose this point correctly? :)
 
Forester #:
Financial time series is not a sound and cannot be decomposed into harmonics. A pure note sounding only 1 sec is 500-1000 oscillations of a periodic signal (approximately a sine wave). Only a periodic signal (which repeats hundreds or thousands of times and does not change in time) can be decomposed into harmonics.
Have you ever seen waves with the same period and amplitude 1000 times in a row in a financial series? You can get rich on a sine wave by the 10th period. And by the 1000th, all the money in the world will be in your pocket.
It's all random, sometimes 1 wave, sometimes 3, for a change and 5 can slip through, so that then all the people who entered on such a flat.
Legend says that there are about 5% of constant frequencies, which can be caught. At least it is not the history of recording, trying different models and hoping for results.
 
Why bother? Here in the market, the tops are making millions on neurons, ask them about their super-intelligent robots))))) Only it is desirable to pretend to be a fool, they only with such willingly communicate )))
 
Рaра Нoth #:
Why bother? Here in the market, the tops are making millions on neurons, ask them about their super-intelligent robots))))) Only it is desirable to pretend to be a fool, they only with such willingly communicate )))
Please provide the full text of the interrogation
1...75767778798081828384
New comment