Machine learning in trading: theory, models, practice and algo-trading - page 2624

[JeeyCi]

Maxim Dmitrievsky # :
not from layering 😀

Is "profane speech" from the word profane...?

so you're part of the same herd... - "the laughing and shitting firm" ... in such a one, one really only learns and works with low probability, thinking that Bayes will do everything for them...

well, if the pack has decided to make a revolution in statistics and mo - I won't get in the way... don't forget to try your own things!

[Maxim Dmitrievsky]

JeeyCi #:

Is "profane speech" from the word "profane"?

So you're part of the same pack... - "the laughing and shitting firm"... in one of those, one really only learns and works with a low probability, thinking that Bayes will do everything for them...

well, if the pack has decided to make a revolution in statistics and mo - I won't get in the way... don't forget to try your own things!

I'm not even in denial.

But having recently learned about context and forget gate, teaching Masha with a clever look, it's not a good idea to give advice.

[mytarmailS]

Maxim Dmitrievsky #:

Have you seen anything about symbolic regression?

[Maxim Dmitrievsky]

mytarmailS #:
Have you seen anything about symbolic regression?

No, I haven't, describe your idea of what to do, maybe with an example, I can do some digging. Maybe there are other ways through the MT5 optimizer. The idea is not clear yet

[Renat Akhtyamov]

arrived ;)

https://www.mql5.com/ru/forum/86386/page2615#comment_28692929

[mytarmailS]

Maxim Dmitrievsky #:
Not yet, you describe the idea of what to do, maybe with an example, you could poke around. And maybe there are other ways through the MT5 optimizer. The idea is not clear yet

The idea is to generate patterns of any complexity, without limitations

[Maxim Dmitrievsky]

mytarmailS #:

The idea is to generate patterns of any complexity, without limit

We take fetches and targets and look for a f-u to which the values belong? I got it, you can use an optimizer or a ready made package.

https://gplearn.readthedocs.io/en/stable/

https://medium.com/analytics-vidhya/python-symbolic-regression-with-gplearn-cbc24dbbc271

I don't see how this is better than NS or boosting, except for lower error rate sometimes

[mytarmailS]

Maxim Dmitrievsky #:
We take the features and targets and look for a f-function, to which the values belong? Well, I got it, you can do it with the optimizer or a ready-made package.

https://gplearn.readthedocs.io/en/stable/

https://medium.com/analytics-vidhya/python-symbolic-regression-with-gplearn-cbc24dbbc271

I don't see how this is better than NS or boosting, except for the lower error rate sometimes.

This article is no better, probably even worse, the usual approximation...

Imagine that you need to find a graphical pattern. It looks like this: weekly low, the price breaks it down, then up, then down again, then up again, and then the price touches the moving average from the top and then you have to buy.

That's the rule... Just the sequence of events, without time reference, without indices, the pattern can be formed as much as 2 hours or as much as 12 hours...

Even if we know this pattern, it's not so easy to program it, and we don't know it... The question is how can such ideas be generated automatically?

[Rorschach]

Has anyone checked the MO for lice? Take 100 bars for example. 50% feed in like this, 50% randomly mix up the order. Will the net/forest be able to tell where random is? Forest is even more interesting, you can look at the signs.

[Maxim Dmitrievsky]

mytarmailS #:
This article is no better, probably even worse, the usual approximation...

Imagine you need to find a graphical pattern. It looks like this: weekly low, price breaks it down, then up, then down again, then up again, and then the price touches the moving average down from above and then you have to buy.

That's the rule... Just the sequence of events, without time reference, without indices, the pattern can be formed as much as 2 hours or as much as 12 hours...

Even if we know this pattern, it's not so easy to program it, and we don't know it... The question is how can such ideas be generated automatically?

Nothing comes to mind so far, in the context of symbolic regression 😀 we have chunks of building blocks like MAs, nadal minima and the like. The question is what comes out