Random Flow Theory and FOREX - page 34
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
Yes, I wrote and wrote and tried to explain. But the result is astonishing.
One thinks that I am talking about fractals and looking for a black cat (I do not see where I said about fractals, I must be blind).
And this is just great.
And after such statements to recommend a book in English about wavelets.
I'll open a great secret for you then, if Kotelnikov's theorem is not fulfilled, then wavelets don't work. I recommend you to take another close look at this one.
If you want to use formula Fd>2*fmax and calculate on a piece of paper, what is the high-frequency component during the gap and compare it with sampling frequency (if it is difficult, just compare the gap period and data arrival period). Maybe these values are equal? If so, then sorry, the device is not working correctly
Maybe you missed the paragraph "How is the price measured?" and so you think that the nature of this phenomenon is unknown to me. Or you mistakenly think that DSP has a nature, I thought all my life that it is a tool invented by man, which helps him to analyze many phenomena of nature.As for zoologists, biologists, ..... or even gynecologists. If the techniques developed by them will help create a model and allow you to predict the "behaviour" of the price in the future. I will gladly use it.
P.S. Maybe it's easier for me to create a topic. Help! Looking for a script (indicator, procedure) that collects ticks, removes false ticks from them, approximates the price by 5 ticks using the least squares method (the number of ticks can be changed) and reflects the price as a point with a confidence interval of evaluation. Not writing 555 pages here for what it's all about.
P.P.S. For some reason I thought that by writing this big text (the previous one). I draw your attention to the problem of accuracy of indirect measurements, what Laplace, Legendre, Gauss thought about. About the fundamental principle of all natural sciences: measurement of magnitude without accompanying accuracy is meaningless.
And it turns out that's not what it says :-(, too bad. I tried very hard to write lucidly, but I guess I never learned how to state my thoughts well and correctly. Sorry.
to Prival
It says it right, I recommend thinking about it. I understand Kotelnikov's theorem helps to shoot down enemy AF, but I'm afraid to create TC will not help, the nature of this theorem is different. Mistake in the beginning: forex is not a flight of enemy aircraft and there is no way to make air defence.
Don't tell me - he has got a nerve.
They work even for series that are obtained in violation of Kotelnikov's theorem. This happens for the simple reason (now it's my turn to open secrets) that it's just mathematics and it doesn't give a damn how exactly a signal is digitized and whether it can be reconstructed from the original one.
I think I have clearly explained the cause of the gap and you will NEVER be able to eliminate it, remember your plane on the ground and my elementary example. And no ridiculous recommendation of DC will help.
You all know :o).
If you listen to ancient Chinese - everything has a nature, even stone. Every tool has its own nature and its own limitations. By the way, I was referring to the nature of forex, not DSP. You have too much faith in DSP, that's the mistake. Do you believe that a gap indicates a violation of the theorem? And you ask the business tycoons why they need the theorem?
I am glad that the row of "Laplace, Legendre," has been expanded by another thinker, a mathematician... Prival - what you wrote about is clear, "measure more often - it will be more accurate" or "give the fulfillment of the theorem". What I'm trying to tell you is work with what you have and don't make it up. By the way, why do you need such an accurate series with such an exorbitant sampling rate? Do you seriously believe that this is a key to a successful TS? Let's listen to a gynaecologist together on this.
One thinks that I am talking about fractals and looking for a black cat (I cannot see where I said about fractals, I must be blind).
It's probably about me again. About fractals I corresponded with Candid, not with you (if carefully read), we have old discussion, sorry, that in your branch, and absolutely not on a theme. I have enclosed a book to you from the bottom of my heart, very kind, by the way.
You succeeded, I remind you of your own words: "Wishes if you are told some seemingly indisputable things. Always question everything. This is the only true way of a researcher. I try to live up to the wish, dear Prival. Have I lived up to my wish?
PS: Prival, if you think I don't belong here - give me a hint, I won't play your opposition. And most importantly, don't be angry. DSP so DSP - it's up to you to implement it, not me. I've said practically everything what I think about it. In a couple of days the next test will finish work, and I will not interfere in the search for your black cat in the room, sincerely hope that it is there at least.
The point is that the problem is very complex. And it takes time for others to figure it out. And you need patience to gradually explain the problem and its importance. Try to explain what is a variable sampling rate, what to eat it with and whether such a view will do any good.
By the way, and I like the book on wavelets already. Thanks, grasn. Such a solid, unifying approach to the overall DSP problem.
P.S. Prival, how do you think neural networks were treated here a year ago?
to grasn
Most often they talk about continuous wavelet transform. http://www.basegroup.ru/filtration/wavelet_for_bussines.htm Everything is fine there. But when you go from continuous to discrete, it's the sampling rate that really matters. Imagine you have continuous process (Mexican hat) and imagine that wavelet transform should work fine, but if ADC is bad (sample rate = 1 Hz) and time during which this Mexican hat exists (1 sec), then instead ofseries of numbers you get one number - say 5.
The key word here is row.
Therefore this phrase of yours IHMO is wrong
Они работают даже на рядах, которые получены с нарушением теоремы Котельникова. А происходит это по той простой причине, (теперь моя очередь открывать тайны) что это просто математика и ей по барабану, на сколько точно оцифрован сигнал и можно ли по нему восстановить исходный.
The question of sampling frequency is even more serious for wavelet transform application; it should ideally be equal to infinity. Unfortunately it is not enough in FOREX, that's why the gaps are there. The only way to reduce the gap (to become not 40 points, but say 38) is to get a big heap of quotes providers (I wrote about it). But anyway the gaps will exist, maybe their value will be a bit less.
Maybe that's more clear what I was talking about.
P.S. Don ' t leave a forum, as a result of disputes sometimes truth is born. Only don't concern military as Budyonny who sent cavalry to tanks in 1941. You can meet intelligent and competent people among them sometimes.
The point is that the task you've set is a very difficult one. It takes time for others to get to grips with it. And you need patience to gradually explain the problem and its importance. Try to explain what is a variable sampling rate, what to eat it with and whether such a view will do any good.
By the way, and I like the book on wavelets already. Thanks, grasn. Such a solid, unifying approach to the overall DSP problem.
P.S. Prival, how do you think neural networks were treated here a year ago?
Wavelets are good and even very good, they allow to do many things. But everything is not so simple there, nothing is given for free, you always have to give something in exchange. And you should apply them with good understanding of what you are doing. This is a tool of analysis (like a surgeon's scalpel) and one must be able to use it, as well as the NS, by the way. I always regarded NS as a dumb algorithm, which is nowhere near the intelligence attributed to it.
Yes, I would like to draw your attention once again to the market model that I propose, it has the properties that Peters writes about "Chaos and Order" p. 21-25. It also has the properties of Brownian motion (the model of independent increments). It is just that you and I are often confused by different designations and different terms from different books. I don't claim that this model is something unknown in science, it seems to me that it (the model) hasn't been applied to the market yet or I haven't found it in open sources.
But we will talk about that later, when we will check the adequacy of the model. The main thing now is to get a qualitative range of data.
The algorithm is dumb, Prival, and 90% of intelligence is the authors', ie one who has directed this NS to properly prepared data, and even choosing the correct architecture. Nevertheless - look what a brouhaha Better has raised... And the algorithm of fitting of scales can hardly be called a dumb search too, agree.
That's the thing: 90%, in my opinion 99% at all. NS is a tool of recognition theory. In terms of NS construction, I want to build a network architecture and it will consist of models of price "behaviour" with different parameters. I will input only quotes and you know about the output. But this algorithm cannot be called an NS.
And Better is good, and he has those necessary (correct) elements from my point of view.
Hello all! You've written a lot in three days...
Sergey, why doesn't this link http://grasn.narod.ru/002.djvu work?
I'm interested in that book too, but I couldn't find it on the rides.