Machine learning in trading: theory, models, practice and algo-trading - page 1973

New comment
 
Maxim Dmitrievsky:

So, what's up there? Have you figured out a little bit about the logic of the script?

 
mytarmailS:

So, what is it? Have you figured out a little bit of the logic of the script?

No, I haven't understood the logic of the NS yet, I still can't understand it.

I'll send you my example of learning from quotes. But first, read a book about the syntax, at least in passing, otherwise you'll get stuck )

The main thing is working with lists (a la arrays), how to separate them, select elements, etc.

and loops

 
Maxim Dmitrievsky:

or you'll be stupid.)

There's no more ))

Maxim Dmitrievsky:
 Well, the main thing is to work with lists (a la arrays), how to separate them, select elements, etc.

I think it's confusing. In P, everything is separated, a matrix is a matrix, a vector is a vector, a sheet is a sheet, but in Python it's all in one pile ((( but I will try to do it )

 
mytarmailS:

There's nowhere else to go.)

it's simple, it's the easiest language

 
mytarmailS:

There's nowhere else to go.)

That's right, it's confusing... In R-ka everything is separated, matrix is a matrix, vector is a vector, list is a list, while in python it's all in one pile ((( but I'll try to

In python, an array is a list. There can be both numeric and other elements. There's also tuple, dictionary, set... but those are rarely used. Mostly lists.

If you need dataframes, use the Pandas package, it has a lot more features. Including for working with time series.

A one-dimensional list is a vector, a two-dimensional list is a matrix, etc.

 
Aleksey Vyazmikin:

The longer the path, the longer the delay.

Well, rewrite it in mql - a deep convolutional network, it will take you only two or three years to write the code .... But at the same time, you will win Tenth of a second to transfer the data, because it is so important when you trade minutes!

So do it! :)

 
mytarmailS:

Well, rewrite it in mql - a deep convolutional network, it will take you only two or three years to write the code .... But at the same time, for that you will win Tenth of a second on the data transfer, because it is so important when you trade minutes!

So do it! :)

I wanted to tell you about Pipe Channels as an alternative to transferring data, but it turned out that this issue is not covered in R, alas.

But I found an article on getting macroeconomic statistics through R, it may be useful to include fundamental data in datasets.

I also found some interesting informationhere:

"

14. fast R builds

Revolution Analytics, later bought by Microsoft, has done a tremendous job optimizing R. Their Revolution R product is now known as Microsoft R Open.

One ofthe benefits of this build is the use of theIntel Math Kernel Library, which greatly speeds up matrix operations (noticeable when working with large matrices).

Where to download:https://mran.microsoft.com/download
Installation instructions:https: //mran.microsoft.com/documents/rro/installation

"

"

15. JIT compiler

In the days of version 2.13 of the R interpreter, one of its leading developers created thecompiler package, which greatly improved the performance of R when executing code with loops. The effect of its use is sometimes so noticeable that the package is now included in the standard R package, and for all standard functions the bytecode is generated at the build stage of the corresponding packages.

"

Have you tried any of these - is there really a performance gain?

 
mytarmailS:

Now I can't even run Metatrader )))) ahah, I feel like a helpless nerd))


Hear me, I've got Eclipses figured out and so can you. Wipe your drooling weakling and keep working :-)

 
Maxim Dmitrievsky:

can you give me a link to the source code on githab? maybe there is some theory there

Hello Maxim Dmitrievsky. I just read your email. There is nothing on the githab. I took some simple network code and modified it for me. It was decisive for me to weed out the garbage from the tick history. Python

df = df['bid'].drop_duplicates(keep='last').values
df = df[-240000::1]
 
mytarmailS:

It started the terminal, but I don't think it did anything else, 4 errors again


By the way, it's a packet analyzer error. I fixed it this way:

The standard there is jedi in the settings. I installed Pylance and changed 'jedi' at the bottom to Pylance in the studio settings via search. Now it sees all the fields of the package and doesn't give out any errors.


1...196619671968196919701971197219731974197519761977197819791980...3589
New comment