Machine learning in trading: theory, models, practice and algo-trading - page 2983

[Vladimir Perervenko]

Renat Fatkhullin #:

It's just that you are not aware of the new functionality of MQL5 standard matrix methods:

• https://www.mql5.com/ru/docs/matrix

A big step has been made in basic matrix and vector maths. The writing can be considerably shorter.

That's true.

But there is a lack of simple but very necessary vector and table functions. Such as lag(), lead(), diff(), roll(), apply() and others constantly used in calculations. Besides, since you have started to implement MO, you should add a dataframe structure with accompanying manipulation functions, which is actually standard in MO. As a wish.

Good luck

[Aleksey Nikolayev]

Vladimir Perervenko #:

That's right.

But it lacks simple but very necessary vector and table functions. Such as lag(), lead(), diff(), roll(), apply() and other functions that are constantly used in calculations. Besides, since you have started to implement MO, you should add a dataframe structure with accompanying manipulation functions, which is actually standard in MO. As a wish.

Good luck

That's for sure. And some people have already tried to adapt matrices in their place)

[Forester]

Aleksey Nikolayev #:

That's right. And some people have already tried to put matrices in their place).

What is the advantage of dataframes? I looked it up on the internet:

You use data frames if columns (variables) can be expected to be of different types (numeric/character/logical etc.). Matrices are for data of the same type.

Even when I was experimenting with R, I did everything on matrices. I had only numbers on the inputs. And even now only numbers. Sometimes (for experimentation) I assign a column to a categorical one, but categories are also defined by numbers.

[Aleksey Nikolayev]

Forester #:
What's the advantage of dataframes? I searched the internet: Even when I was experimenting with R, I did everything on matrices. I only had numbers on the inputs. And even now only numbers. Sometimes (for experimentation) I assign a categorical column, but categories are also defined by numbers.

Even numbers are different - int, long, float, double. And if you suddenly want to work with text data - news or data from the Internet?

The point is that dataframes make it possible to work with heterogeneous data in a string. Whether a particular person needs it or not at a particular moment of time is a deeply secondary issue. Any opportunity for choice is always more useful to have than not to have.

[Maxim Kuznetsov]

Forester #:
What's the advantage of dataframes? I searched the internet: Even when I was experimenting with R, I did everything on matrices. I only had numbers on the inputs. And even now only numbers. Sometimes (for experiment) I assign a categorical column, but categories are also defined by numbers.

In short: tabular organisation of data, by columns. It's pretty damn convenient, you can use the set you need and correlate it. Plus, it's got a well-established terminology and methods.

just don't pull it into the terminal...or they'll do it again like they always do :-) the SQLite example is illustrative.

[mytarmailS]

Interesting video

https://www.youtube.com/watch?v=vOIotMC_PQY

[Maxim Dmitrievsky]

mytarmailS #:
What is it?

Causal inference.

Anomalies are okay, I haven't seen the rest of the vid yet

[Roman]

Renat Fatkhullin #:

It's just that you are not aware of the new functionality of MQL5 standard matrix methods:

• https://www.mql5.com/ru/docs/matrix

A big step has been made in basic matrix and vector maths. The writing can be considerably shorter.

Finish Sort() method
Some functions are not relevant without it

[mytarmailS]

Maxim Dmitrievsky #:
Causal Inference

Anomalies are a good topic, I haven't watched the rest of the video yet.

at the end of the video they talk about an interesting package, you can try to move it.

[Aleksey Vyazmikin]

mytarmailS #:

Interesting video

https://www.youtube.com/watch?v=vOIotMC_PQY

FidoNet 2.0 means - interesting :)

The question of automatic generation of signs is very interesting of course. Has anyone tried this approach for trading? I have seen programmes that do this, but it is very complicated to implement, so I didn't bother to look into it. Therefore, the issue of integrating these converters into MQL5 for use in the same ONNX models is important.