[R] R for linear algebra

Avi Gross @v|gro@@ @end|ng |rom ver|zon@net
Sat Jun 25 20:38:42 CEST 2022

Just adding to my previous post, I looked back and this forum has at times had discussions in which the use of packages related to R STUDIO and/or the tidyverse was considered somehow not aligned with the mailing list. 

Many of the above are, in MY opinion, way more relevant to people using R in their daily lives as some are very popular packages.

I am not saying a package focused on teaching Linear Algebra is not useful or that the audience is limited. I was saying that R was partially designed specifically for many things done in linear Algebra to be doable in relatively few steps and thus a hospitable place all on it's own and a great place to host such an add-on package.

In my experience, though, most packages and modules out there do not stand alone. For serious uses, you must glue together various parts of what the package supplies with fairly routine code such as to read in data and convert it to the form needed, or to use results in other packages and so on.

So my earlier replies were focused far wider on R indeed being a hospitable environment regardless of any specific packages supplied that zoomed in on parts of a specialized field of endeavor. I am glad someone supplied one such package for consideration but I also think a course could be taught where most of the functionality needed was from base R and some other packages already widely used except perhaps for some esoteric parts.

Just as a BTW, I have seen packages that use a GUI to automate loading in data, applying various transformations and analyses on it and producing some kinds of output. Much of it is checkboxes and the like. But the part of it that I appreciated the most was that it was able to show you the underlying R code it generated which you could copy and modify for all kinds of additional needs you may have that the GUI did not cover or that ...

Some tools need not replace but augment and for professionals, it can be better to be able to see what is going on under the hood, especially if problems arise.

So if the class being taught is to people who will continue using it for years, I might teach it differently as compared to those just needing three credits. It is sort of like the multiple tracks I have seen for undergraduate classes in a subject like Physics. My school had classes with or without derivations and Calculus and other math so some just had equations thrown at them and others got to understand why those equations worked the way they did and how they were "invented" and how to make others with some thought and so on. But it seems some schools now have a third track called something like "Physics for Poets" which cuts back on even showing too many equations!

Linear Algebra and R can also be taught at multiple levels and being a purist, my preference is to UNDERSTAND things not just on the surface. But most people who are just passing by do not really need to learn much of it, let alone all of it. Yet it becomes very sad when some such cruisers later on decide to stick around and do additional sloppy work using what little they know or making R do things as similar to the methods they used in some other programming language.

That is part of why I made comments about the package as to whether it really is something people should continue to use heavily if they want to do lots of work in the field. 

-----Original Message-----
From: Avi Gross via R-help <r-help using r-project.org>
Cc: r-help using r-project.org <r-help using r-project.org>
Sent: Sat, Jun 25, 2022 2:09 pm
Subject: Re: [R] R for linear algebra


I am not in any way disparaging the matlib package and it seems well-built for the limited purpose of teaching Linear Algebra rather than R. It is probably a better answer to a question about how to teach linear algebra while making some more complex tasks doable. I recall the frustration of multiplying matrices by hand as well as other operations, necessarily on smaller matrices.

My comments were more along the lines of the charter of this group which seems far narrower. Yes, people can ask for suggestions for a package that does something that may interest them but getting help on any one of thousands of such packages here would get overwhelming.
>From what you said though, others looking for a package to use for real projects might well beware as it may indeed not be particularly fast or in some cases perhaps not as flexible.

-----Original Message-----
From: John Fox <jfox using mcmaster.ca>
To: Avi Gross <avigross using verizon.net>
Cc: r-help using r-project.org <r-help using r-project.org>
Sent: Sat, Jun 25, 2022 1:34 pm
Subject: Re: [R] R for linear algebra

Dear Avi,

The purpose of the matlib package is to *teach* linear algebra and 
related topics, not to replace or even compete with similar 
functionality in base R.

Consider, e.g., the following example for matlib::Inverse(), which 
computes matrix inverses by Gaussian elimination (I've elided most of 
the steps):

 > example("Inverse")

Invers>  A <- matrix(c(2, 1, -1,
Invers+                -3, -1, 2,
Invers+                -2,  1, 2), 3, 3, byrow=TRUE)

Invers>  Inverse(A)
      [,1] [,2] [,3]
[1,]    4    3  -1
[2,]  -2  -2    1
[3,]    5    4  -1

Invers>  Inverse(A, verbose=TRUE, fractions=TRUE)

Initial matrix:
      [,1] [,2] [,3] [,4] [,5] [,6]
[1,]  2    1  -1    1    0    0
[2,] -3  -1    2    0    1    0
[3,] -2    1    2    0    0    1

row: 1

  exchange rows 1 and 2
      [,1] [,2] [,3] [,4] [,5] [,6]
[1,] -3  -1    2    0    1    0
[2,]  2    1  -1    1    0    0
[3,] -2    1    2    0    0    1

  multiply row 1 by -1/3
      [,1] [,2] [,3] [,4] [,5] [,6]
[1,]    1  1/3 -2/3    0 -1/3    0
[2,]    2    1  -1    1    0    0
[3,]  -2    1    2    0    0    1

  multiply row 1 by 2 and subtract from row 2
      [,1] [,2] [,3] [,4] [,5] [,6]
[1,]    1  1/3 -2/3    0 -1/3    0
[2,]    0  1/3  1/3    1  2/3    0
[3,]  -2    1    2    0    0    1

. . .

  multiply row 3 by 2/5 and subtract from row 2
      [,1] [,2] [,3] [,4] [,5] [,6]
[1,]  1    0    0    4    3  -1
[2,]  0    1    0  -2  -2    1
[3,]  0    0    1    5    4  -1
      [,1] [,2] [,3]
[1,]    4    3  -1
[2,]  -2  -2    1
[3,]    5    4  -1

And similarly for the other functions in the package. Moreover, the 
functions in the package are transparently programmed in R rather than 
calling (usually more efficient but relatively inaccessible) compiled 
code, e.g., in a BLAS.


On 2022-06-24 9:57 p.m., Avi Gross via R-help wrote:
> Yes, Michael, packages like matlib will extend the basic support within base R and I was amused at looking at what the package supported that I had not thought about in years!
> https://www.rdocumentation.org/packages/matlib/versions/0.9.5
> However, once you through in packages/modules/libraries and other add-ons, I suggest many languages share such gifts that then allow a serious amount of what we call linear Algebra to be done, albeit with some work.
> In some ways my review of Linear Algebra in recent years showed me that some things have changed from when I took it as part of a math degree in college but a big change has been in finding so many uses of it now that computers can do complicated things fast and even faster when using vectors and matrices as a way to organize and consolidate operations, sometimes even in parallel.
> Packages can be nice and especially if they gather together lots of related functions to teach a subject but anyone doing serious work should first make sure they know what is in the base.
> If your matrix is A, you could load a package like psych to get a trace of A:
> psych::tr(A)
> Or install  matlib with oodles of dependencies
> matlib::tr(A)
> Or without worrying if the user had installed and made available a library, use built-ins like diag() and sum():
> sum(diag(A))
> And what does  matlib::Det(A) gain you most of the time that det(A) does not tell you?
> A policy of this forum is to point out mostly what is part of standard R and obviously there can  be specialized functionality in packages, albeit some functions  in that package are unlikely to be taught in a first undergraduate Linear Algebra course. I note matlib is billed as also being for learning Multivariate Statistics, and that is fine, just perhaps a tad of overkill.
> If the users have a question about the base R functionality, they can ask about it here. But if they want to know why the cholesky function in matlib is not working for them, ...
> -----Original Message-----
> From: Michael L Friendly <friendly using yorku.ca>
> To: org (R-help using r-project.org) <r-help using r-project.org>
> Cc: Avi Gross <avigross using verizon.net>; abouelmakarim1962 using gmail.com>, <abouelmakarim1962 using gmail.com>
> Sent: Fri, Jun 24, 2022 9:12 pm
> Subject: Re: R for linear algebra
>  <!--#yiv2977149432 _filtered {} _filtered {}#yiv2977149432 #yiv2977149432 p.yiv2977149432MsoNormal, #yiv2977149432 li.yiv2977149432MsoNormal, #yiv2977149432 div.yiv2977149432MsoNormal {margin:0in;font-size:11.0pt;font-family:"Calibri", sans-serif;}#yiv2977149432 a:link, #yiv2977149432 span.yiv2977149432MsoHyperlink {color:#0563C1;text-decoration:underline;}#yiv2977149432 p.yiv2977149432MsoPlainText, #yiv2977149432 li.yiv2977149432MsoPlainText, #yiv2977149432 div.yiv2977149432MsoPlainText {margin:0in;font-size:11.0pt;font-family:"Calibri", sans-serif;}#yiv2977149432 span.yiv2977149432EmailStyle17 {font-family:"Calibri", sans-serif;color:windowtext;}#yiv2977149432 span.yiv2977149432PlainTextChar {font-family:"Calibri", sans-serif;}#yiv2977149432 .yiv2977149432MsoChpDefault {font-family:"Calibri", sans-serif;} _filtered {}#yiv2977149432 div.yiv2977149432WordSection1 {}-->Hello    The {matlib} package was developed for just this purpose – teaching and illustrating linear algebra concepts and techniques. In my old 6140 Multivariate Data Analysis course, I spent ~5-6 weeks on linear algebra.    CRAN: https://cran.r-project.org/web/packages/matlib/index.html GitHub: https://github.com/friendly/matlib 6140: http://euclid.psych.yorku.ca/www/psy6140/    There are quite a few vignettes    --- Michael Friendly     Email: friendly AT yorku DOT ca Professor, Psychology Dept York University      Voice: 416 736-2100 x66249 4700 Keele Street    Web:http://www.datavis.ca | @datavisFriendly Toronto, ONT  M3J 1P3 CANADA          Message: 11 Date: Fri, 24 Jun 2022 02:31:24 +0000 (UTC) From: Avi Gross <avigross using verizon.net> To: "abouelmakarim1962 using gmail.com" <abouelmakarim1962 using gmail.com>,                "r-help using r-project.org" <r-help using r-project.org> Subject: Re: [R] R for Linear Algebra Message-ID: <1658548037.4476854.1656037884865 using mail.yahoo.com> Content-Type: text/plain; charset="utf-8"    R is a language specially made to do many things with vectors and matrices and arrays with any reasonable number of dimensions and also has a host of functions that do things like make diagonal matrices, calculate determinants and multiply them in several ways,  invert them, get eigenvalues and much more if you look for packages. So the question becomes, what else do you feel you need for a Linear Algebra course that a simple web search would not have found like this: https://towardsdatascience.com/intro-to-r-linear-algebra-2a4de2506c93    https://towardsdatascience.com/intro-to-r-linear-algebra-2a4de2506c93#:~:text=R%20is%20especially%20handy%20with,solvers%20and%20dynamic%20indexing%20capabilities. Unless I missed something, yes, R is more easily able to be used to do such things than many other languages albeit just about any language tends to have access to a variety of functions that may do these things but are not built in.       -----Original Message----- From: AbouEl-Makarim Aboueissa <abouelmakarim1962 using gmail.com> To: R mailing list <r-help using r-project.org> Sent: Thu, Jun 23, 2022 10:39 am Subject: [R] R for Linear Algebra    Re: R for Linear Algebra          Dear All:                In my STA 591 topics in Biostatistics Fall 2022, I am planning to spend about 5 weeks on Linear Algebra.          I am wondering if someone has or knows the source for some R functions/programs written for linear algebra calculations.          Also if you know a book in Linear Algebra with R, will be a good help too.       Any help in this matter is highly appreciated in advance.
>     [[alternative HTML version deleted]]
> ______________________________________________
> R-help using r-project.org mailing list -- To UNSUBSCRIBE and more, see
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
John Fox, Professor Emeritus
McMaster University
Hamilton, Ontario, Canada
web: https://socialsciences.mcmaster.ca/jfox/

    [[alternative HTML version deleted]]

R-help using r-project.org mailing list -- To UNSUBSCRIBE and more, see
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.

	[[alternative HTML version deleted]]

More information about the R-help mailing list