Filters again. And who will do the filters? And what are these particular phases? How are they? -Algorithmically? This is not a tsar's job.
I guess so: they went to the DM - let him identify everything himself.
Well, filters, not filters... I don't know what to call it.
If they have started to change, but they are not yet sufficient for the NS to properly learn, then it, like all the other counters, will go berserk.
If I take my first bot, that is being trained in the optimizer - it trades cool on similar patterns, for example, if the market has been growing for 3 months at a certain pace, but with slight differences, it still works. As soon as the fundamental dependencies have changed henceforth goodbye.
This all relates to TSs that make a lot of deals... I'll say nothing about position-taking, I'm not interested in it.
Well, Yuri, judge for yourself. On the input - the purest increments, the weighting coefficients will be the values of probability density function of these increments. We are not dealing with a price, but with a wave packet of prices. I.e. with that very t2-distribution of increments (though multiplied by the exponent, but more about it in due time). And that's it, I guess. On the output, as soon as this wave packet takes a certain form (pattern) - the signals to open/close a trade. The amplitude of probabilities is calculated as a proportion of the root of t.
By the way, it is probably possible to feed the incremental velocities to the input, as Feynman did (in his mind!!!).
In general, it's all about increments, not prices. Without filters, logarithms and other stuff.
Well, Yuri, think about it. On the input - the purest increments, your weighting coefficients will be the values of the probability density function of these increments. We are not dealing with a price, but with a wave packet of prices. I.e. with that very t2-distribution of increments (though multiplied by the exponent, but more about it in due time). And that's it, I guess. On the output, as soon as this wave packet takes a certain form (pattern) - the signals to open/close a trade. The amplitude of probabilities is calculated as a proportion of the root of t.
By the way, it is probably possible to feed the rates of increments to the input as well, as Feynman did (in his mind!!!).
In general, it's all about increments, not prices. No filters, logarithms, or any other nonsense.
Yuriy Asaulenko: How many inputs will the NS have?
I think there are two. One - for increments, the second - for their speeds.
PS Please don't judge too harshly - I've just started working on NS. Looking for NeuralNet for VisSim. If you have any, please send me the link in your PM.
I think there are two. One - for increments, the second - for their speeds.
PS Please don't judge too harshly - I've just started working on NS. Looking for NeuralNet for VisSim. If you have one, please send me a link in your PM.
Where is the distribution? Classic MLP has no memory.
Yuriy Asaulenko: Then where is the distribution? Classic MLP has no memory.
I told you, I need NeuralNet for VisSim. I will see - what is possible and what is not.
PS I'm a theorist - I can generate ideas in high-frequency mode, but in practice - there's only 1 deal on my account, and it's a curve... Although I've started 20 pairs at once today. Let's see :)))
PS I'm a theorist - I can generate ideas in high-frequency mode
It feels. As Feynman said, a thought experiment costs nothing. You can start up a thermonucleus like two fingers - what energy is needed there - only 4-5 times more than in a TV kinescope.
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.
Filters again. And who will do the filters? And what are these particular phases? How are they? -Algorithmically? This is not a tsar's job.
I guess so: they went to the DM - let him identify everything himself.
Well, filters, not filters... I don't know what to call it.
If they have started to change, but they are not yet sufficient for the NS to properly learn, then it, like all the other counters, will go berserk.
If I take my first bot, that is being trained in the optimizer - it trades cool on similar patterns, for example, if the market has been growing for 3 months at a certain pace, but with slight differences, it still works. As soon as the fundamental dependencies have changed henceforth goodbye.
This all relates to TSs that make a lot of deals... I'll say nothing about position-taking, I'm not interested in it.Filters again. And who will make the filters? And what are these particular phases? How are they? -Algorithmically? This is not a tsar's job.
So I guess: left to the DM - let him himself to identify everything.
You don't need any filters! You need to feed the purest price increments to the input of NS and it's done.
Hi!
More details please, how, where?)
Alexander will explain it to us all.)
Hi!
More details please, how, where?)
Hi!
Well, Yuri, judge for yourself. On the input - the purest increments, the weighting coefficients will be the values of probability density function of these increments. We are not dealing with a price, but with a wave packet of prices. I.e. with that very t2-distribution of increments (though multiplied by the exponent, but more about it in due time). And that's it, I guess. On the output, as soon as this wave packet takes a certain form (pattern) - the signals to open/close a trade. The amplitude of probabilities is calculated as a proportion of the root of t.
By the way, it is probably possible to feed the incremental velocities to the input, as Feynman did (in his mind!!!).
In general, it's all about increments, not prices. Without filters, logarithms and other stuff.
Hi!
Well, Yuri, think about it. On the input - the purest increments, your weighting coefficients will be the values of the probability density function of these increments. We are not dealing with a price, but with a wave packet of prices. I.e. with that very t2-distribution of increments (though multiplied by the exponent, but more about it in due time). And that's it, I guess. On the output, as soon as this wave packet takes a certain form (pattern) - the signals to open/close a trade. The amplitude of probabilities is calculated as a proportion of the root of t.
By the way, it is probably possible to feed the rates of increments to the input as well, as Feynman did (in his mind!!!).
In general, it's all about increments, not prices. No filters, logarithms, or any other nonsense.
How many inputs will the NS have?
I think there are two. One - for increments, the second - for their speeds.
PS Please don't judge too harshly - I've just started working on NS. Looking for NeuralNet for VisSim. If you have any, please send me the link in your PM.
I think there are two. One - for increments, the second - for their speeds.
PS Please don't judge too harshly - I've just started working on NS. Looking for NeuralNet for VisSim. If you have one, please send me a link in your PM.
Where is the distribution? Classic MLP has no memory.
If you know English, you can also read Bishop.
Then where is the distribution? Classic MLP has no memory.
I told you, I need NeuralNet for VisSim. I will see - what is possible and what is not.
PS I'm a theorist - I can generate ideas in high-frequency mode, but in practice - there's only 1 deal on my account, and it's a curve... Although I've started 20 pairs at once today. Let's see :)))
PPS Aha. I'll read it at my leisure. Thank you.
PS I'm a theorist - I can generate ideas in high-frequency mode