Machine learning in trading: theory, models, practice and algo-trading - page 567

New comment
 
Vladimir Perervenko:

The solution is as follows:

Load the R library reticulate (R interfface to Python) and follow the documentation and examples.

By the way, R has the same package xgboost. Or is Python's better? You can see an example here

Good luck

I don't understand something, why do we need R here? Comrade needs MQl. So, write a DLL, add #include <python.h> and so on, and call Python functions. Correct me if I'm wrong.
 
Yuriy Asaulenko:
I don't understand, why do we need R here? Comrade needs MQl. So, write the DLL, add #include <python.h> and call the Python functions. Correct me if I'm wrong.

Correction.

1. R has been working steadily in MT/MQL for a long time. Checked.

2. The reticulate package was written by the RStudio team to provide a robust interface between R -> Python/ It was conceived mainly to use TensorFlow, CNTK, Keras and other numerous Python developments in R. And now all of these packages run as native in R.

So I don't think a library written by a knee-jerk enthusiast (without disparaging his competence in any way) can take precedence over reticulate. You should read the description of the package. It's very clear there.

Good luck

 
Yuriy Asaulenko:
I don't understand, why do we need R here? My friend here needs MQl. So, write a DLL, add #include <python.h> and so on, and call the Python functions. Correct me if I'm wrong.

It's kinda cooler this way - call python through R and then R through dll from MT5... or rather the reverse sequence... it's too easy to live in Russia

 
Maxim Dmitrievsky:




 
Maxim Dmitrievsky:

It's kinda cooler this way - call python through R and then R through dll from MT5... or rather the reverse sequence... it's too easy to live in Russia


 
SanSanych Fomenko:



I know all this :)

"
  • The algorithm is prone to overtraining, especially on noisy tasks. This problem can be partially overcome by adjusting the parameterr(see below). A similar problem, only more pronounced, is observed in the original Random Forest algorithm (seeMachine Learning Benchmarks and Random Forest Regression). It should be noted that this shortcoming was not noticed by its authors, who assumed that the algorithm was not prone to overtraining, a misconception shared by some machine learning practitioners and theorists.

"


http://alglib.sources.ru/dataanalysis/decisionforest.php

 
Maxim Dmitrievsky:

I know it all :)


Let it be for others, maybe it will be useful.

 
Vladimir Perervenko:

Correction.

1. R has been working steadily in MT/MQL for a long time. Checked.

2. The reticulate package was written by the RStudio team to provide a robust interface between R -> Python/ It was conceived mainly to use TensorFlow, CNTK, Keras and other numerous Python developments in R. And now all of these packages run as native in R.

So I don't think a library written by a knee-jerk enthusiast (without disparaging his competence in any way) can take precedence over reticulate. You should read the description of the package. It's very clear there.

Good luck

No, I don't doubt the competence of the developers of PCtudio.

But why make a mess of things like Python -> R -> Dll -> MQl? If you can go straight to Python -> Dll -> MQL? And we're not talking about creating libraries, but calling specific Python program functions, which already provide for interaction with C/C++ - from C-API to Boost.Python, etc.

I would like to understand the logic.

 
Yuriy Asaulenko:

No, I don't doubt the competence of PCtudio developers at all.

But why make a mess like Python -> R -> Dll -> MQl? If you can go straight to Python -> Dll -> MQL? And we're not talking about creating libraries, but calling specific Python program functions, which already provide for interaction with C/C++ - from C-API to Boost.Python, etc.

I'd like to understand the logic.


Is there a dll for python?

 
SanSanych Fomenko:

Is there a dll for python?


Well, above is the link

1...560561562563564565566567568569570571572573574...3495
New comment
Reason: