Hi,
I'm a college student who discovered operations research last year and
have been planning on majoring in it. I find it highly interesting. But
I'm not sure the discipline fits well with my strengths, and am thus
considering majoring in pure math instead.
I'm very good at discrete math and geometry, but not great with
algebra. I think this is because I'm good at solving problems through
visualization (as is often done in graph theory and combinatorics);
this kind of problem solving seems to be less frequently used in
algebra.
I understand that linear algebra is the bread and butter of R, so I am
wondering if someone who's not very good at it is asking for trouble by
entering the field. Maybe there is a subfield of R that would allow me
to employ the kinds of problem-solving techniques I'm good at?
Thank you,
Rex
p.s.: I could elaborate on what I mean by this "visual" vs. "abstract"
dichotomy I'm setting up.

Rex Eastbourne wrote:
Hi,

I'm a college student who discovered operations research last year and
have been planning on majoring in it. I find it highly interesting. But
I'm not sure the discipline fits well with my strengths, and am thus
considering majoring in pure math instead.

I'm very good at discrete math and geometry, but not great with
algebra. I think this is because I'm good at solving problems through
visualization (as is often done in graph theory and combinatorics);
this kind of problem solving seems to be less frequently used in
algebra.

I understand that linear algebra is the bread and butter of R, so I am
wondering if someone who's not very good at it is asking for trouble by
entering the field. Maybe there is a subfield of R that would allow me
to employ the kinds of problem-solving techniques I'm good at?

Thank you,

Rex

p.s.: I could elaborate on what I mean by this "visual" vs. "abstract"
dichotomy I'm setting up.

I'll let the practitioners weigh in how far you can get *doing* R
without a solid basis in linear algebra. Having taught R courses, I
feel fairly confident in saying that someone who is uncomfortable with
the either the fundamentals of linear algebra and matrix theory
(including things like linear dependence/independence, basis of a vector
space, inverse of a matrix, ) will struggle in pretty much any R
college course.

All versions of optimization (linear, nonlinear and integer programming,
and optimal control, to name the common ones) rely heavily on linear
algebra. In the real world, you may not be inverting matrices yourself,
and if you do you will surely use software, but learning optimization
theory (and algorithms) requires linear algebra. The same is true of
stochastic processes (including queueing theory). There's a pretty good
chance you could go through a simulation course or two without seeing
linear algebra, but any regression analysis or multivariate statistics
class worth a darn is going to use it heavily.

All that said, a solid first course in linear algebra should be enough
to carry you through the R course work. Also, while I'll buy into your
visual-abstract dichotomy for the sake of argument (although personally
I find a Klein bottle visually abstract), I'm not convinced that linear
algebra lies in the abstract camp. Some of the algorithmic elements
(such as Cramer's rule for inverting a matrix) might be abstract, but to
me vectors and linear operator on vector spaces are quite "visual".

/Paul

Sun, 11 Jun 2006 18:51:32 -0400, "Paul A. Rubin" <rubin@msu.edu>
wrote:

>
>I'll let the practitioners weigh in how far you can get *doing* R
>without a solid basis in linear algebra. Having taught R courses, I
>feel fairly confident in saying that someone who is uncomfortable with
>the either the fundamentals of linear algebra and matrix theory
>(including things like linear dependence/independence, basis of a vector
>space, inverse of a matrix, ) will struggle in pretty much any R
>college course.

"R courses" or "optimization algorithms" courses?

A.L.

A.L. wrote:
Sun, 11 Jun 2006 18:51:32 -0400, "Paul A. Rubin" <rubin@msu.edu>
wrote:
>
>I'll let the practitioners weigh in how far you can get *doing* R
>without a solid basis in linear algebra. Having taught R courses, I
>feel fairly confident in saying that someone who is uncomfortable with
>the either the fundamentals of linear algebra and matrix theory
>(including things like linear dependence/independence, basis of a vector
>space, inverse of a matrix, ) will struggle in pretty much any R
>college course.
>
"R courses" or "optimization algorithms" courses?

A.L.

Mix of optimization courses (theory and algorithms) and stochastic
processes courses (mainly theory), plus simulation. I think I know
where you're headed, though, and I agree. We no longer teach R here,
but when we did we had a "traditional" R curriculum, which contained
effectively no courses on R practice. I suppose we were also somewhat
complicit in the trend away from the early days of R (recognize the
nature of the problem and develop an appropriate tool) toward what I
think has been the latter day approach (teach a set of tools, then make
every problem fit one of the tools).

That said, I'm pretty sure our curriculum with consistent with those at
most other schools in the U.S., meaning that someone majoring in R will
take courses in optimization, stochastic processes, statistics and
simulation, and all but simulation will rely on linear algebra.

/Paul

Mon, 12 Jun 2006 11:16:15 -0400, "Paul A. Rubin" <rubin@msu.edu>
wrote:

>
>Mix of optimization courses (theory and algorithms) and stochastic
>processes courses (mainly theory), plus simulation. I think I know
>where you're headed, though, and I agree. We no longer teach R here,
>but when we did we had a "traditional" R curriculum, which contained
>effectively no courses on R practice. I suppose we were also somewhat
>complicit in the trend away from the early days of R (recognize the
>nature of the problem and develop an appropriate tool) toward what I
>think has been the latter day approach (teach a set of tools, then make
>every problem fit one of the tools).

From the place where I am now (company providing software for
logistics management and associated consulting; I am a sort of "chief
scientist") current R curriculum looks like the attempt to teach car
design engineers and give them advanced theory of materials as the
only professional skill. However, designing a car requires a bit more
than just this

Current R graduates have very narrow skills that make them
practically useless in industry. They know how to program in AMPL, but
the request to provide APIs in C/C++ or Java to make R solution
useable for other groups results in statement: "We are R specialists,
THEY will do the API". The problem is that there are no "THEY".
Similarly, knowledge of AMPL, pivoting techniques and branch-and-cut
is totally useless when it is necessary to perform analysis of
practical problem, formalize the problem, build a model and design the
architecture for solution.

INFRMS is lamenting that R discipline has no sufficient recognition
and hired PR agency to promote "Science of Better" campaign. This is
pure nonsense. The problem is not in lack of recognition. The problem
is curriculum, designed by academics, implemented by academics and
oriented towards educating new academics (I don't criticize you
personally, Paul. I was committing the same crime when I was in the
academia :)

You may ask what is the solution. I have some ideas, but I don't have
the answer. I think that discussion could be facilitated by INFRMS as
a replacement for nonsense "Science of Better" campaign.

A.L.

You might look into CS as well, particularly the theory side. Lots of
graph theory, combinatorics, discrete math and discrete algorithms, and
usually light on algebra. Seems to fit more into the visual side of things.

Rex Eastbourne wrote:
Hi,

I'm a college student who discovered operations research last year and
have been planning on majoring in it. I find it highly interesting. But
I'm not sure the discipline fits well with my strengths, and am thus
considering majoring in pure math instead.

I'm very good at discrete math and geometry, but not great with
algebra. I think this is because I'm good at solving problems through
visualization (as is often done in graph theory and combinatorics);
this kind of problem solving seems to be less frequently used in
algebra.

I understand that linear algebra is the bread and butter of R, so I am
wondering if someone who's not very good at it is asking for trouble by
entering the field. Maybe there is a subfield of R that would allow me
to employ the kinds of problem-solving techniques I'm good at?

Thank you,

Rex

p.s.: I could elaborate on what I mean by this "visual" vs. "abstract"
dichotomy I'm setting up.

Shaddin Doghmi wrote:
You might look into CS as well, particularly the theory side. Lots of
graph theory, combinatorics, discrete math and discrete algorithms,
and usually light on algebra. Seems to fit more into the visual side
of things.

But I suspect that linear algebra, or at least matrix algebra, is a big
deal if you get into graphics.

/Paul

11 Jun 2006 00:02:45 -0700, "Rex Eastbourne"
<rex.eastbourne@gmail.comwrote:

>Hi,
>
>I'm a college student who discovered operations research last year and
>have been planning on majoring in it. I find it highly interesting. But
>I'm not sure the discipline fits well with my strengths, and am thus
>considering majoring in pure math instead.
>
>I'm very good at discrete math and geometry, but not great with
>algebra. I think this is because I'm good at solving problems through
>visualization (as is often done in graph theory and combinatorics);
>this kind of problem solving seems to be less frequently used in
>algebra.

What means "you are no good"? Study more and you will be good.
Even elephant can dance. Moreover, linear algebra has nice
geometrical interpretations of almost all constructs. Try to
"visualize" algebra.

In the future you will confront a lot of things that "you are not
good in". Study hard and you will be good. This is American
educational system that lets students believe that studying is fun
and only fun, and that this is teacher's duty to "teach me".
Studying is work, work, work. Fun is when you finally become good.

A.L.

"AL." <alewando@ziuta66.comwrote in message
@4ax.com
Mon, 12 Jun 2006 11:16:15 -0400, "Paul A. Rubin" <rubin@msu.edu>
wrote:
>
>
>Mix of optimization courses (theory and algorithms) and stochastic
>processes courses (mainly theory), plus simulation. I think I know
>where you're headed, though, and I agree. We no longer teach R here,
>but when we did we had a "traditional" R curriculum, which contained
>effectively no courses on R practice. I suppose we were also somewhat
>complicit in the trend away from the early days of R (recognize the
>nature of the problem and develop an appropriate tool) toward what I
>think has been the latter day approach (teach a set of tools, then make
>every problem fit one of the tools).
>
>
From the place where I am now (company providing software for
logistics management and associated consulting; I am a sort of "chief
scientist") current R curriculum looks like the attempt to teach car
design engineers and give them advanced theory of materials as the
only professional skill. However, designing a car requires a bit more
than just this

Current R graduates have very narrow skills that make them
practically useless in industry. They know how to program in AMPL, but
the request to provide APIs in C/C++ or Java to make R solution
useable for other groups results in statement: "We are R specialists,
THEY will do the API". The problem is that there are no "THEY".
Similarly, knowledge of AMPL, pivoting techniques and branch-and-cut
is totally useless when it is necessary to perform analysis of
practical problem, formalize the problem, build a model and design the
architecture for solution.

INFRMS is lamenting that R discipline has no sufficient recognition
and hired PR agency to promote "Science of Better" campaign. This is
pure nonsense. The problem is not in lack of recognition. The problem
is curriculum, designed by academics, implemented by academics and
oriented towards educating new academics (I don't criticize you
personally, Paul. I was committing the same crime when I was in the
academia :)

You may ask what is the solution. I have some ideas, but I don't have
the answer. I think that discussion could be facilitated by INFRMS as
a replacement for nonsense "Science of Better" campaign.

A.L.
>
>
>

From my narrow perspective (I'm an R. practitioner since 1985, but I
started without any frmal R. education) in a large corporation, being the
responsible for R. consulting and development in my country, I can not do
anything but fully agree with A.L.

While I take for granted that an R. expert should now the math it is
required to use, and should be able to use the tools the community makes
available, there more important skills are
1) in the 'modeling' part: how to understand a real world problem (its
business drivers) and translate it into model (and process) formulations
accurate enough and implementableand computable models.
2) in the implementation part: how to make models really part of an industry
grade application (API's, model exception handling, sound Data Modeling, GUI
to let users understand what the model actually suggests )

Usually both these skills are very missing in the R. practitioners I
interview.

Stefano Gliozzi

Wed, 28 Jun 2006 17:41:11 +0200, "Stefano Gliozzi"
<stefano_gliozzi@it.ibm.comwrote:

>
>"AL." <alewando@ziuta66.comwrote in message
[]
>INFRMS is lamenting that R discipline has no sufficient recognition
>and hired PR agency to promote "Science of Better" campaign. This is
>pure nonsense. The problem is not in lack of recognition. The problem
>is curriculum, designed by academics, implemented by academics and
>oriented towards educating new academics (I don't criticize you
>personally, Paul. I was committing the same crime when I was in the
>academia :)
>>
>You may ask what is the solution. I have some ideas, but I don't have
>the answer. I think that discussion could be facilitated by INFRMS as
>a replacement for nonsense "Science of Better" campaign.
>>
>A.L.
>>
>>
>>
>
>From my narrow perspective (I'm an R. practitioner since 1985, but I
>started without any frmal R. education) in a large corporation, being the
>responsible for R. consulting and development in my country, I can not do
>anything but fully agree with A.L.
>
>While I take for granted that an R. expert should now the math it is
>required to use, and should be able to use the tools the community makes
>available, there more important skills are
>1) in the 'modeling' part: how to understand a real world problem (its
>business drivers) and translate it into model (and process) formulations
>accurate enough and implementableand computable models.
>2) in the implementation part: how to make models really part of an industry
>grade application (API's, model exception handling, sound Data Modeling, GUI
>to let users understand what the model actually suggests )
>
>Usually both these skills are very missing in the R. practitioners I
>interview.

Unfortunately, the academia is not interested in changing the status
quo. Mostly because the majority of professors also don't possess
these skills.

A.L.

A.L. wrote:
Wed, 28 Jun 2006 17:41:11 +0200, "Stefano Gliozzi"
<stefano_gliozzi@it.ibm.comwrote:
>
>"AL." <alewando@ziuta66.comwrote in message
[]
INFRMS is lamenting that R discipline has no sufficient recognition
and hired PR agency to promote "Science of Better" campaign. This is
pure nonsense. The problem is not in lack of recognition. The problem
is curriculum, designed by academics, implemented by academics and
oriented towards educating new academics (I don't criticize you
personally, Paul. I was committing the same crime when I was in the
academia :)

You may ask what is the solution. I have some ideas, but I don't have
the answer. I think that discussion could be facilitated by INFRMS as
a replacement for nonsense "Science of Better" campaign.

A.L.


>>From my narrow perspective (I'm an R. practitioner since 1985, but I
>started without any frmal R. education) in a large corporation, being the
>responsible for R. consulting and development in my country, I can not do
>anything but fully agree with A.L.
>>
>While I take for granted that an R. expert should now the math it is
>required to use, and should be able to use the tools the community makes
>available, there more important skills are
>1) in the 'modeling' part: how to understand a real world problem (its
>business drivers) and translate it into model (and process) formulations
>accurate enough and implementableand computable models.
>2) in the implementation part: how to make models really part of an industry
>grade application (API's, model exception handling, sound Data Modeling, GUI
>to let users understand what the model actually suggests )
>>
>Usually both these skills are very missing in the R. practitioners I
>interview.
>
Unfortunately, the academia is not interested in changing the status
quo. Mostly because the majority of professors also don't possess
these skills.

A.L.

Hey! I (rep)resent that remark!

There's probably some truth to the fact that some of us lack coding
skills (and even more of us lack database skills), some of us lack
experience with the more hands-on elements of getting models to run
correctly, and some of us may lack all of the above. But the "status
quo" of R in academe is declining enrollments and declining
institutional support, to the extent that a number of schools (including
mine) have essentially phased out R programs. If we crank up the
requirements for the students, that's not likely to stimulate
enrollments. What's needed is for industry to come to us and say we
need (and will hire) people with the following skills <insert laundry
list here>.

/Paul (who is not personally responsible for the demise of R here,
rumors to the contrary)