You're a strange passenger -- always whining -- regularly someone doesn't show you something, doesn't give you something.
Back in 1987, there was a film with Stallone called "To the Best of Your Ability" -- the key phrase there: "Nobody in the world will go out of their way to help you. If you want something, take it yourself."
Thanks, Cap. Now I know who to turn to for helpful advice.
Maxim Dmitrievsky #: Training on shorter intervals requires an analytical f-ya, which does not correspond to training on unknown data of unknown nature. For training on a large number of examples it is no longer needed, the algorithm starts to catch patterns (if there are any in the data).
The ph is stationary and periodic, so it is easy to predict.
It's hard to understand whether you don't understand the essence of the problem or you are deliberately trolling.
Above I have shown three sections of the function, which has an analytical formula and, therefore, it is possible to test training methods on it. First, take the first section as shown in the picture (the range t is specified), split the function into any number of points (even a million) and try to train the network on these points (simulate discrete data). And then see how the network behaves on the next section as in the picture.
If these points are not clear, things are very bad.
This is a very simple example to check, in fact real market data contains not one hidden process function, but many, where the CVR is the sum of these processes.
It's hard to tell whether you're missing the point or deliberately trolling.
Above I have shown three sections of the function, which has an analytical formula, and, therefore, you can test training methods on it. First, take the first section as shown in the picture (the range t is specified), split the function into any number of points (even a million) and try to train the network on these points (simulate discrete data). And then see how the network behaves on the next section as in the picture.
If these points are not clear, then everything is very bad.
This is a very simple example to check, in fact real market data contains not one hidden process function, but many, where the CVR is the sum of these processes.
The problem is set incorrectly, as usual. Use three points to plot three other points. I took a large number of points and predicted them. Questions?
It seems to be a non-periodic function. But at small X it can be like this. Theorising, haven't built it.
100% of the research was on which analytic functions are bad at MO.
The man just took a section where several periods of the function fit and pretends he can easily approximate the series. In reality, this is not the case on the CVR, the length of periods and how many processes are contained there. Therefore, as always, he showed fake results and passed them off as truth.
Participation in this thread is rapidly losing its meaning.
Maxim Dmitrievsky #: The task is incorrectly set, as usual. Use three points to construct three other points. I took a large number of points and predicted. Any questions?
I have no questions for you. Read my post above, take a million points on the specified area and try to predict on the next area as in the pictures. This exercise is very useful to understand the topic, I recommend it.
took a section where several periods of the function are stacked and pretends that it can easily approximate the series
Unfortunately, it is not obvious to me from the analytical form of the function that it is periodic. But if several periods fall into the training interval, even a human can predict its behaviour. That is, taking such a learning interval is not interesting at all.
It is much more indicative to take an interval, for example, two times smaller than the period, but without restrictions on the number of training points.
Andrey Dik #: I don't and never have had any questions for you. Read my post above, take a million points on the specified area and try to forecast on the next area as in the pictures. This exercise is very useful to understand the topic, I recommend it.
Allow the use of cookies to log in to the MQL5.com website.
Please enable the necessary setting in your browser, otherwise you will not be able to log in.
You are signed in. Two-factor authentication has been disabled. Set it up again and scan a new QR code, or continue with two-factor authentication disabled.
You're a strange passenger -- always whining -- regularly someone doesn't show you something, doesn't give you something.
Back in 1987, there was a film with Stallone called "To the Best of Your Ability" -- the key phrase there: "Nobody in the world will go out of their way to help you. If you want something, take it yourself."
It seems to be a non-periodic function. But at small X, it may look like one. I'm theorising, haven't built it.
100% research was which analytic functions are bad for MO.Very difficult to predict
Training on shorter intervals requires an analytical f-ya, which does not correspond to training on unknown data of unknown nature. For training on a large number of examples it is no longer needed, the algorithm starts to catch patterns (if there are any in the data).
It's hard to understand whether you don't understand the essence of the problem or you are deliberately trolling.
Above I have shown three sections of the function, which has an analytical formula and, therefore, it is possible to test training methods on it. First, take the first section as shown in the picture (the range t is specified), split the function into any number of points (even a million) and try to train the network on these points (simulate discrete data). And then see how the network behaves on the next section as in the picture.
If these points are not clear, things are very bad.
This is a very simple example to check, in fact real market data contains not one hidden process function, but many, where the CVR is the sum of these processes.
It seems to be a non-periodic function. But at small X it can be similar to this one. Theorising, haven't built it.
100% of the research has been on which analytic functions are bad at MO.It's hard to tell whether you're missing the point or deliberately trolling.
Above I have shown three sections of the function, which has an analytical formula, and, therefore, you can test training methods on it. First, take the first section as shown in the picture (the range t is specified), split the function into any number of points (even a million) and try to train the network on these points (simulate discrete data). And then see how the network behaves on the next section as in the picture.
If these points are not clear, then everything is very bad.
This is a very simple example to check, in fact real market data contains not one hidden process function, but many, where the CVR is the sum of these processes.
It seems to be a non-periodic function. But at small X it can be like this. Theorising, haven't built it.
100% of the research was on which analytic functions are bad at MO.The man just took a section where several periods of the function fit and pretends he can easily approximate the series. In reality, this is not the case on the CVR, the length of periods and how many processes are contained there. Therefore, as always, he showed fake results and passed them off as truth.
Participation in this thread is rapidly losing its meaning.
The task is incorrectly set, as usual. Use three points to construct three other points. I took a large number of points and predicted. Any questions?
took a section where several periods of the function are stacked and pretends that it can easily approximate the series
Unfortunately, it is not obvious to me from the analytical form of the function that it is periodic. But if several periods fall into the training interval, even a human can predict its behaviour. That is, taking such a learning interval is not interesting at all.
It is much more indicative to take an interval, for example, two times smaller than the period, but without restrictions on the number of training points.
I don't and never have had any questions for you. Read my post above, take a million points on the specified area and try to forecast on the next area as in the pictures. This exercise is very useful to understand the topic, I recommend it.