Predicting the future with Fourier transforms - page 44

New comment
 
alsu:
Not really. If the sines and cosines are a finite number, then it is a series. But this number is not finite for all, but only for periodic functions. For all other functions we have some generalisation with infinite number of sines/cosines (and infinitesimal intervals between them)
For example, it is impossible to express exp(-at) by a finite number of sines
 
LeoV:

OK, where's the profit? ))))

Anywhere. Depending on the specifics of the trading system. There may be two filters with different characteristics (analogue of two MAs), may be by slope. It is possible to filter out a useful signal and apply other methods of analysis to it.
 

Fourier has no application at all to forex forecasts; it was invented for approximation.

 
alsu:
Not really. If the sines and cosines are a finite number, then it is a series. But this number is not finite for all, but only for periodic functions. For all other functions we have some generalisation with infinite number of sines/cosines (and infinitesimal intervals between them)

But no matter how you spin it... We decompose it, add it up and get the same thing. Our data is discrete, so there are a finite number of elements in the series.
 

Once again, the Fourier transform (which has an integral) is an extremely convenient tool for analytical calculations. Personally, I prefer to deal with it, and having derived the necessary final formulas (for example, for filter parameters), to process data using them, rather than to rush to quotes with a discrete transform, which, of course, does not predict anything by itself.

Incidentally, nature's Fourier transformer is the hearing organ: in our ear, an acoustic wave is converted into a set of frequencies, which we perceive as sound. And we can often guess what will happen to us in a minute. Even phase information is then discarded. Why shouldn't the analogue have a place in the forex market.

 
Integer: You decompose it, tweak the harmonics, add up - it's a filter, an advanced MA with infinite possibilities for its regulation.

Approximation is also a way of doing things. Approximated, look where it's pointing.


Wherever you want. Depending on the peculiarities of the trading system. There can be two filters with different characteristics (analog of two MAs), can be by slope. A useful signal can be filtered out and other methods of analysis can be applied to it.

It is more complicated - there are too many variables - the probability of adjustment is high. The simplest variant is to decompose it, take the harmonic that will be profitable in the future, and earn from the crossover with itself at its extremums.

But both variants (yours and mine) will generate the same question - how to identify the harmonics that will be profitable in the future. In my variant - you need to determine only one harmonic, in yours - much more, which is more likely to lead to the drain.....)))).

 
Integer:

Sorry, but this is not an explanation of Fourier, but a demonstration of its complete incomprehension.


But you seem to understand it. And of course you can explain it to me and everyone else. Or is that all you have to say?

Here's a quote from Pedivikai:

Analysis(Greek ἀνάλυσις - decomposition, dissection) is an operation of mental or real dissection of the whole (thing, property, process or relationship between objects) into its components, performed in the process of knowledge or object-practical human activity.

Suppose we have the task of decomposing the number 1 into 5 components. The variants 0, 0, 1, 0, 0 or 0.2, 0.2, 0.2, 0.2 look natural and fit for some further use. But option 1, 2, -2, 3, -3 looks unnatural and only complicates things, although it too is supposedly correct. Fourier does roughly the same thing with non-periodic functions.

 
alsu:

Once again, the Fourier transform (which has an integral) is an extremely convenient tool for analytical calculations. Personally, I prefer to deal with it, and having already derived the necessary final formulas (for example, for filter parameters), to process data using them, rather than to rush to quotes with discrete transform, which, of course, does not predict anything by itself.

Incidentally, nature's Fourier transformer is the hearing organ: in our ear, an acoustic wave is converted into a set of frequencies, which we perceive as sound. And we can often guess what will happen to us in a minute. Even phase information is then discarded. Why shouldn't the analogue have a place in the forex market.


I.e. you can do these calculations, like decompose the data into harmonics, adjust amplitudes, phases, add up, only instead you can calculate coefficients to calculate results like in FATL, SATL indicators - just multiplying prices by coefficients and adding up.
 
AlexeyFX:


But you seem to understand everything.


You got it right))
 
LeoV:

It's more complicated this way - there are so many variables - there's a high probability of fitting. The simplest option - decompose it, take the harmonic that will be profitable in the future, and earn on the intersection with itself in the extremums.

But both variants (yours and mine) will generate the same question - how to identify the harmonics that will be profitable in the future. In my variant, you need to identify only one harmonic, in yours - many more, which will most likely lead to a drain.....)))).


So with neural networks you get even more parameters. Using only one harmonic is a special case of using multiple - same multiple, only all have 0 amplitude except one. If we use only one, then we come to Herzl, to MESA.
1...373839404142434445464748495051...55
New comment