Machine learning in trading: theory, models, practice and algo-trading - page 938
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
Maybe it is because the 10 o'clock is first, after a long pause from 24 o'clock? That's why it predicted well? Is there something specific about 24 hours, other than low volatility?
So the trading session starts at 10 o'clock, after a pause from 23-50 - so there are always strong movements and disorder, so it is logical that this moment is exceptional in relation to all the other days.
Now I will try to combine days of the week with time, the news days and time of their release should appear - let's check the theory.
I keep getting into it.
I've been learning this stuff for 5 years and couldn't even imagine that it could be applied to a forum.
In short, NS is an ordinary adaptive recursive filter.
There are many variations.
With a teacher, without a teacher, etc..
Predictors are digital filtering coefficients. It's hilarious because it's elementary...
Tell me, did you already make a blind one or use input/output with training?
// Blind is when you don't know what's coming out and there's a foregone conclusion, i.e. there's nothing to compare it to...First of all, unusual. Actually, NS is strictly non-linear shit. Second, nonrecursive, although there are also recursive NS.
It only looks elementary, but when you dig into it... "Many of the generals found the hunters and took, but come up, it happens - no, wise. It seems to be easy to look at, but when you look at it, it's a hell of a thing!" (с)
I only use the standard BP with the teacher. + my own modifications of learning (this is manual).
I don't do prognosis, or don't do it yet - I don't know. Only classification on MLP.
First of all, unusual. Actually, NS is strictly nonlinear shit. Second, it is nonrecursive, although there are recursive NS.
It only looks elementary, but when you dig into it... "Many of the generals found the hunters and took, but come up, it happens - no, wise. It seems to be easy to look at, but when you look at it, it's a hell of a thing!" (с)
I only use the standard BP with a teacher. + my own modifications of learning (this is manual).
this is the easiest, fastest way to do it
i.e. is your own teacher
With your eyes you look
I don't do prognosis, or don't do it yet - I don't know. Only classification on MLP.
you have a signal on the input, there is a signal on the output // one previous tick watch at least
Yuri, you make and trade forecasts. Whatever you want to call it, but it's a fact.
Okay, I'm off to program.
// And yes, it's not recursive, my mistake. Or maybe I just don't want to use it. Time will tell.
So the trading session starts at 10 o'clock, after a pause from 23-50 - so there are always strong movements and disorder, so it is logical that this moment is exceptional in relation to all the other days.
Now I will try to combine days of the week with time, the news days and their time of release should be displayed - let's check the theory.
I can't determine the news by day+hour nexus, maybe because of the time factor - winter/summer time change...
Maybe some predictors are interfering...this is the easiest, or rather the fastest, thing to do
that is, you are your own teacher
You look at it through your eyes.
I look at intermediate results after H epochs and modify parameters of further learning.
you have a signal on the input, you have a signal on the output //you watch the previous tick, at least
Yura, you make and trade a forecast. Whatever you want to call it, but it's a fact.
There is only a time series applied to the NS and nothing else. There are no indicators-predictors, there is nothing of that. Indicators before that work and ordinary logic decides whether to feed anything to the NS or screw it.
The weather is supposed to be good, or the weather will be bad. That is the entire forecast. And how good or bad it is not defined and such task is not set at all. For NS it is a task of classification.
I watch the intermediate results after H epochs, and modify the parameters of further training based on the results.
Only the time series is fed to the NS, and nothing else. There are no indicators-predictors-none of it. Indicators before that work and ordinary logic decides whether to feed anything to the NS or not.
The weather is supposed to be good, or the weather will be bad. That is the forecast. And how good or bad it is not defined and such task is not set at all. For the NS, it is a classification task.
You didn't understand what NS does with it....
Why use made up stuff, write it yourself.
I found a 5-row 64th order VFD on the forum and converted it to a 4-row, added opener, high and low.
Try to read the code, and you'll understand, that it's easy to do, and so is the teacher....
Anyway, you didn't understand what NS.... does with it.
You said yourself, the NS is an analogue of the filter + also non-linear. What does it do? - It selects the filter coefficients, and at the same time it builds the necessary indicators-predictors inside itself.
Why should I use an invented one?
I've found a 5-row 64th order LPF on the forum, and converted it to 4-row.
Try to get into the code and you'll realize that it's all done in a snap and the teacher is the same....
I believe that there are people who have done this (unlike me) professionally, and we should use their results, not invent bicycles. By the way, this is one of the basic principles of OOP. And not inheritance classes etc.
And we can make our own filters, our qualification allows us).
You said yourself, the NS is an analogue of the filter + also non-linear. What does it do? - It selects filter coefficients, and at the same time builds the necessary indicators-predictors inside itself.
That's what I mean, too.
To find the coefficients you have to see the result first.
And I'm telling you it adapts to forecasts.
Well, right?
You said yourself, the NS is an analogue of the filter + also non-linear. What does it do? - It selects filter coefficients, and at the same time it builds the necessary indicators-predictors inside itself.
I believe that there are people who have done all this (unlike me) professionally, and one should use their results, rather than inventing bicycles. By the way, this is one of the basic principles of OOP. And not inheritance classes etc.
(We can also make filters ourselves, our qualifications allow).
They will not lay out what we need.