Machine learning in trading: theory, models, practice and algo-trading - page 3382

New comment
 
Catbust, for example, has about 15 built-in commonly used "eval metrics". If the dataset is balanced, accuracy is commonly used, otherwise balanced accuracy. The others do not make a significant difference. If the data is rubbish, the metrics do not help.
 
fxsaber #:

There's a certain slyness to it. The links are just to make sure they open. No one who is "interested" will delve into them. No one will read Andrei's chewed-up articles, let alone works of an academic nature.


Has anyone seen this easy-to-understand TOP with the ability to calculate the ranking of their own optimisation algorithm?

https://habr.com/ru/users/belyalova/publications/articles/

No, it's just that the topic is really complex and there are few obvious works on the topic. And the productive works are even less))))))

 

What's been going on here for the last few pages? The topic is interesting, but it keeps getting lost in "who's the Sensei" squabbles.

How about some kind of applied problem ? Whoever solves it is the Jedi.

 
fxsaber #:
In terms of assessing the robustness of the algorithm. Because in your case it is a part of this algorithm. The topic is about MO, so I am writing in terms of MO, not a special case - MT5 optimiser.
 
Aleksey Nikolayev #:
I would like a link to specific algorithms on multi-criteria optimisation in function space. But if not willing to provide one, it's best to gloss over it for clarity) Not willing to waste time searching.

Here, unfortunately, it's true. Few people will want to dig through thousands of books and sift through them, separating grain from chaff. I did it 14 years ago, downloaded, collected, sifted, sifted, put in folders. And in the end, what? Accusations and slander.

That doesn't apply to you. I hope that this list of references will be useful to you in your search for the information you need. And, specifically I recommend Karpenko and Simon, there are more solid sources in terms of the volume of information, but more difficult to digest the material.


ZЫ. If someone needs it, can put links to interesting literature in this thread. But nobody needs it, nobody will do it, everybody needs ready-made.

 
A perfect example of how the choice of FF has little effect on the quality of the result if the data is rubbish. Try the catbust metrics, run with each. If the data isn't rubbish, then all work.

Just theory theorists cancer howling and trying to prove something to someone, and reality.
 
Maxim Dmitrievsky #:
The FF is the same, right?

I think he's confusing the FF with the optimisation surface

 
fxsaber #:

It's not about translation.

It's clear that a hundred sets depend on FF.

If you have the possibility to vary the FF, then in order to use the average values of parameters it probably makes sense to take the FF that gives the most heap distribution for the resulting hundred variants.

I'm not sure if this looks meaningful within the framework of trading tasks.

 
fxsaber #:

Has anyone seen this easy to understand top with the ability to calculate the ranking of their optimisation algorithm?

https://habr.com/ru/users/belyalova/publications/articles/

Yes, I have. The lady has several interesting articles on the topic of algorithm comparison. Unfortunately, no codes and bench conditions are given to reproduce the results.

 
mytarmailS #:

I think he's confusing FF with optimisation surface.

I don't know what anyone thinks anymore, it's like some kind of banter.
And it's all pointless in terms of finding patterns or evaluating robustness
1...337533763377337833793380338133823383338433853386338733883389...3526
New comment
Reason: