Machine learning in trading: theory, models, practice and algo-trading - page 3088

 
Andrey Dik #:

Why are they so loud, talking like adults, smoking, swearing?

Because you can do anything on the forum and you won't get anything for it.

 
Maxim Dmitrievsky #:

because you can do anything on the forum and you'll get nothing for it.

apparently so. Amen.
 
Andrey Dik #:

I really wonder what is the basis for blind faith in packages? maybe somewhere there are comparative tests of packages on AO? i'm lost in guesswork....


Blind faith NOT in packages, but in a professional development environment.

The very first sign of a professional environment is the possibility to find something in this environment. If we are talking about R, it is in the field of statistics.

Optimisation is not actually a statistic, but for obvious reasons R contains packages related to optimisation. In TWO clicks I found a link to a prohibitively large list of packages related to optimisation see above.

A package in R is a set of software tools that meet moderation requirements in terms of composition, design, testing and maintenance.

I take the first package in the list - optimx.

It is referenced at https://cran.r-project.org/web/packages/optimx/index.html with the following information:

Version: 2022-4.30
Imports: numDeriv
Suggests: knitr,rmarkdown,setRNG,BB,ucminf,minqa,dfoptim,lbfgsb3c,lbfgs,subplex
Published: 2022-05-10
Author: John C Nash [aut, cre], Ravi Varadhan [aut], Gabor Grothendieck [ctb]
Maintainer: John C Nash <nashjc at uottawa.ca>
Licence: GPL-2
NeedsCompilation: no
Citation: optimx citation info
Materials: NEWS
In views: Optimisation
CRAN checks: optimx results are the results of the package check.

Documentation:

Reference manual: optimx.pdf
Vignettes: Using and extending the optimx package
Rvmmin15
SNewton

Downloads:

Package source: optimx_2022-4.30.tar.gz


I won't comment on all positions, let's just look at the manual https://cran.r-project.org/web/packages/optimx/optimx.pdf.

It turns out that the package contains a couple of dozens of functions.

I would like to point out an extremely important point: there is a link to the description of the package algorithms - this is a common practice in R - I haven't met any packages without algorithm description. All R packages are NOT black boxes, there is always a description of algorithms, which usually has a list of literature on discussion and approbation.

References Nash, John C. and Varadhan, Ravi (2011) Unifying Optimisation Algorithms to Aid Software System Users: optimx for R, Journal of Statistical Software, publication pending

All of the above defines R as a professional development environment and an environment for statistical professionals. To add to this, there is a version of R that Microsoft bought and supports. In the field of statistics today everything else is a "kolkhoz", which does not stand next to R. In another 5-10 years there were competitors, for example, SPSS, and today there are none.

Dick! What can you oppose to this professional approach in programming? I quite admit that you have written something so brilliant. What's in it for us? Don't you understand that no sane programmer would NOT entrust money to a home-made programme? If you have an ingenious optimisation algorithm, then package it and put it on CRAN. But the distance from what you have to CRAN is huge. It takes a huge effort to turn your homemade algorithms into a professional and generally available tool. By the way, optimisation algorithms in R are only a shell in R, and the algorithm itself is either C++ or Fortran.

optimx: Expanded Replacement and Extension of the 'optim' Function
optimx: Expanded Replacement and Extension of the 'optim' Function
  • cran.r-project.org
Provides a replacement and extension of the optim() function to call to several function minimization codes in R in a single statement. These methods handle smooth, possibly box constrained functions of several or many parameters. Note that function 'optimr()' was prepared to simplify the incorporation of minimization codes going forward. Also implements some utility codes and some extra solvers, including safeguarded Newton methods. Many methods previously separate are now included here. This is the version for CRAN.
 
СанСаныч Фоменко #:

Blind faith NOT in packages, but in a professional development environment.

The very first sign of a professional environment is the possibility to find something in this environment. If we are talking about R, then find it in the field of statistics.

Optimisation is not actually a statistic, but for obvious reasons R contains packages related to optimisation. In TWO clicks I found a link to a prohibitively large list of packages related to optimisation see above.

A package in R is a set of software tools that satisfy moderation requirements in terms of composition, design, testing and maintenance.

I take the first package in the list - optimx.

There is a link to it https://cran.r-project.org/web/packages/optimx/index.html with the following information:

Version: 2022-4.30
Imports: numDeriv
Suggests: knitr,rmarkdown,setRNG,BB,ucminf,minqa,dfoptim,lbfgsb3c,lbfgs,subplex
Published: 2022-05-10
Author: John C Nash [aut, cre], Ravi Varadhan [aut], Gabor Grothendieck [ctb]
Maintainer: John C Nash <nashjc at uottawa.ca>
Licence: GPL-2
NeedsCompilation: no
Citation: optimx citation info
Materials: NEWS
In views: Optimisation
CRAN checks: optimx results are the results of the package check.

Documentation:

Reference manual: optimx.pdf
Vignettes: Using and extending the optimx package
Rvmmin15
SNewton

Downloads:

Package source: optimx_2022-4.30.tar.gz


I will not comment on all positions, let's see only the manual https://cran.r-project.org/web/packages/optimx/optimx.pdf.

It turns out that the package contains a couple of dozens of functions.

One more extremely important point: there is a link to the description of the package algorithms - this is a common practice in R - I have not met any packages without algorithm description.

References Nash, John C. and Varadhan, Ravi (2011) Unifying Optimisation Algorithms to Aid Software System Users: optimx for R, Journal of Statistical Software, publication pending

All of the above defines R as a professional development environment and an environment for statistical professionals. To add to this, there is a version of R that Microsoft bought and supports. In the field of statistics today everything else is a "kolkhoz", which does not stand next to R. In 5-10 years there were competitors, for example, SPSS, but today there are none.

Dick! What can you oppose to this professional approach in programming? I quite admit that you have written something so brilliant. What's in it for us? Don't you understand that no sane programmer would NOT entrust money to a home-made programme? If you have an ingenious optimisation algorithm, you can package it and put it on CRAN, but the distance from what you have to CRAN is huge. It takes a huge effort to turn your homemade algorithms into a professional and generally available tool. And the most important obstacle is to publish and get recognition of professional community for your ingenious algorithm. By the way, optimisation algorithms - there is only a shell in R, and the rest is either C++ or Fortran.

Wow! 87 pages of description! Cool, it must be a good thing!

I thought so, blind faith.

You, Fomenko, don't seem to understand that there is no sorcery in packages, they were written by ordinary mortal people.


"It's not the packages, but the local users of these packages :) Like bums digging through them, for no particular purpose." (C)

 
СанСаныч Фоменко #:

and there's a lot of interesting stuff here.

https://cran.r-project.org/web/views/Finance.html



I asked you once how to find out if a vehicle has been retrained.

here you go https://cran.r-project.org/web/packages/pbo/index.html

https://github.com/mrbcuda/pbo

CRAN Task View: Empirical Finance
CRAN Task View: Empirical Finance
  • cran.r-project.org
This CRAN Task View contains a list of packages useful for empirical work in Finance, grouped by topic.
 
mytarmailS #:

What was the problem, Renate? CRAN didn't miss it?

They have religiosity even in the registration process.

You see, they don't work with companies. They only want copyright/personal registrations.

Been stalling for weeks. They.
 
Andrey Dik #:

Is it possible to call the .ex5 library from the MT5 integrated R programme?

This is an external package that can request data from Metatrader.

It is not planned to run inside Metatrader, as it was done for Python scripts.
 
СанСаныч Фоменко #:

Blind faith NOT in packages, but in a professional development environment.

....

All of the above is what defines R as a professional development environment and an environment for statistical professionals.

.....

NO sane programmer would NOT entrust money to a homemade programme? If you have an ingenious optimisation algorithm, that formalise the package and put it on CRAN. But the distance from what you have to CRAN is huge. It takes a huge effort to turn your homemade algorithms into a professional and generally available tool. By the way, optimisation algorithms in R are only a shell in R, and the algorithm itself is either C++ or Fortran.

I would not say that packages for R are written by super programmers and they get perfect absolutely accurate code. The code approaches the ideal one as you devote enough time to it, work with it and test it, find and fix bugs.

When Vladimir's articles with the Darch package appeared, I experimented with it a lot. I devoted enough time to it.
With these experiments I made some suggestions to improve the package and even found 2-3 bugs.

The author corrected a lot of things, but then suddenly rolled everything back to the version before all the corrections. apparently the new edits changed something else somewhere and did not want to deal with it and waste time. As I understood at that moment he had already abandoned the project and was doing other work. Judging by the fact that the last edits were 5-6 years ago - nothing has changed. The project is abandoned and buggy. Luckily it has already been removed https://cran.r-project.org/web/packages/darch/index.html

So any of us can create decent code better than abandoned Darcha with bugs if we work hard on it.

So out of hundreds of packages I would trust only those that are given time and fixes. Like katbusta etc. with funding (or without funding, but with enthusiasm and not abandonment).

Issues · maddin79/darch
Issues · maddin79/darch
  • maddin79
  • github.com
Create deep architectures in the R programming language - Issues · maddin79/darch
 
Renat Fatkhullin #:

We've been at it for weeks. They.

What happened next? )

 
Renat Fatkhullin #:
They have religiosity even in the registration process.

You see, they don't work with companies. They only want author/personal registrations.

We've been at it for weeks. They.

Wo - they don't work with companies either....

And the privateers drop their crafts when they go on other jobs/projects.