|
|
 | | From: | Sascha Herpers | | Subject: | general question about notations | | Date: | Wed, 29 Dec 2004 14:01:05 +0100 |
|
|
 | Hi,
I have a rather philosophical question.
In OR (as in other research areas) there are a lot of notations, e.g.
the Kendall/Lee notation. Whenever somebody introduces a notation, he
usually states how many position there are, and what values the position
may have. Then a few examples are given and maybe some other remarks.
Even though these descriptions are complete, it is mostly lengthy. On
top of that, each auther has his/her own idea about a good descriptive text.
Why does seemingly nobody use a formal language to describe a notation?
Why not use for example the (extended) Backus/Naur [(E)BNF] form? If you
don't like that particular language, pick your own one.
I must admit, that it takes some time to read a rather large notation
description stated in EBNF, but if the description is complete, there
are no questions left open, whether, e.g., value /a/ of property /1/ may
be combined with value /b/ of property /2/, or which one of the
properties is mandatory or optional.
Am I devoting to much of my mental energy to a problem that only I see
or is the generic scientist happy with whatever notation description he
comes up with, as long as it is basicly understandable.
What is you experiance?
Thanks
Sascha
|
|
| | From: | Paul A. Rubin | | Subject: | Re: general question about notations | | Date: | Wed, 29 Dec 2004 13:55:08 -0500 |
|
|
| Sascha Herpers wrote:
> Hi,
>
> I have a rather philosophical question.
>
> In OR (as in other research areas) there are a lot of notations, e.g.
> the Kendall/Lee notation. Whenever somebody introduces a notation, he
> usually states how many position there are, and what values the position
> may have. Then a few examples are given and maybe some other remarks.
>
> Even though these descriptions are complete, it is mostly lengthy. On
> top of that, each auther has his/her own idea about a good descriptive
> text.
>
> Why does seemingly nobody use a formal language to describe a notation?
> Why not use for example the (extended) Backus/Naur [(E)BNF] form? If you
> don't like that particular language, pick your own one.
>
> I must admit, that it takes some time to read a rather large notation
> description stated in EBNF, but if the description is complete, there
> are no questions left open, whether, e.g., value /a/ of property /1/ may
> be combined with value /b/ of property /2/, or which one of the
> properties is mandatory or optional.
>
> Am I devoting to much of my mental energy to a problem that only I see
> or is the generic scientist happy with whatever notation description he
> comes up with, as long as it is basicly understandable.
>
> What is you experiance?
>
> Thanks
> Sascha
EBNF (or something equivalent) would not replace the examples and other
remarks necessary to help the reader understand the new notation. It
would only replace the list of positions and possible entries in each
position.
Are you thinking that EBNF would be a clearer way to present that
information, or just more compact? Do you have an example in mind where
EBNF would be better (on either dimension)? For Kendall/Lee notation,
I'm inclined to think that EBNF would be overkill: not any clearer than
a tabular presentation, and somewhat but not significantly more compact (?).
-- Paul
******************************************************************************************
Paul A. Rubin Phone:
(517) 432-3509
Department of Management Fax: (517)
432-1111
The Eli Broad Graduate School of Management E-mail: rubin@msu.edu
Michigan State University
http://www.msu.edu/~rubin/
East Lansing, MI 48824-1122 (USA)
******************************************************************************************
Mathematicians are like Frenchmen: whenever you say something to them,
they translate it into their own language, and at once it is something
entirely different. J. W. v. GOETHE
|
|
| | From: | Sascha Herpers | | Subject: | Re: general question about notations | | Date: | Wed, 29 Dec 2004 22:21:18 +0100 |
|
|
| Hi,
> EBNF (or something equivalent) would not replace the examples and other
> remarks necessary to help the reader understand the new notation.
certainly not. What I have in mind is, that people sometimes don't mark
optional parts of a notation very clearly and if they do, they do it
sometimes in different way (some with square brackets, some put an
asteriks next to a symbol and mentation somewhere in the text "oh, and
by the way, symbols with * are optional").
As I mentioned in my post, some descriptions simply don't tell you,
which values must not appear a the same time or which symbols have to be
accompanied by a subset of other values. OK, if you have some knowledge
in the area, where the notation is used, you might know these
releations, by if you are new in that field, you most likely don't know.
And that is where a complete and sound definition of a notation come in
very handy. What else is complete and sound if not a formal language as
EBNF. To say it once more: I am not somebody who wants to make everybody
use EBNF. If you have a better formal language to write down notations, ok.
> It
> would only replace the list of positions and possible entries in each
> position.
Yes, but EBNF as a placeholder for a formal language of you choice is more
than a way to write down lists of values and positions. It give you tool
for writing down grammer and that's what notations are.
> Are you thinking that EBNF would be a clearer way to present that
> information, or just more compact?
That is a point, which I don't like myself. Even though I would like the
idea of having a way to specify a notation, I see your point at the same
time: if you want it precise, it'll get very lengthy. But that's
probably a trade-off. Someone who knows how to read EBNF is able to use
any notation in whatever area of research it is used. Small notations
probably don't have enough grammer to them to make it worth the time and
paper to write them down that way.
> Do you have an example in mind where
> EBNF would be better (on either dimension)? For Kendall/Lee notation,
> I'm inclined to think that EBNF would be overkill: not any clearer than
> a tabular presentation, and somewhat but not significantly more compact
> (?).
Kendall/Lee is probably on the border of to small to have a big
advantage with EBNF and to large not to try it. ...hmmm...no good
example...I don't know. Maybe I can come up with something in the next
days, but I don't promise anything.
You probably (or better most certainly) have more experiance with
notatios. Have you never thought, that the introduction of a notation in
a paper or book lacked something or could have been more precise?
Shouldn't there be at least some (more or less) formal guidelines when
writing down notation?
Good night (at least here in Germany)
Sascha
|
|
| | From: | Paul A. Rubin | | Subject: | Re: general question about notations | | Date: | Wed, 29 Dec 2004 16:47:06 -0500 |
|
|
| Sascha Herpers wrote:
> Hi,
>
>> EBNF (or something equivalent) would not replace the examples and
>> other remarks necessary to help the reader understand the new notation.
>
> certainly not. What I have in mind is, that people sometimes don't mark
> optional parts of a notation very clearly and if they do, they do it
> sometimes in different way (some with square brackets, some put an
> asteriks next to a symbol and mentation somewhere in the text "oh, and
> by the way, symbols with * are optional").
I think this tends to reflect their background in computing. People
with formal training in computer languages tend to be at least somewhat
familiar with BNF, but I suspect a significant number of people
(including many who write code, at least for their own use) are not. On
the other hand, people who program in certain languages (Perl and Java
come to mind) may be familiar with "regular expressions", which use
annotations such as ?, + and * to indicate optional repetitions (0 or 1,
1 or more, 0 or more respectively, if I recall correctly).
> As I mentioned in my post, some descriptions simply don't tell you,
> which values must not appear a the same time or which symbols have to be
> accompanied by a subset of other values. OK, if you have some knowledge
> in the area, where the notation is used, you might know these
> releations, by if you are new in that field, you most likely don't know.
Quite true. Even more distressing, there are areas that have generally
accepted notational conventions, and yet every so often some author
feels obligated to create his or her own notation. In a way, I find
that even more confusing, since my brain instinctively reacts to a new
symbol by assuming that it must not represent something for which a
known symbol already exists.
> And that is where a complete and sound definition of a notation come in
> very handy. What else is complete and sound if not a formal language as
> EBNF. To say it once more: I am not somebody who wants to make everybody
> use EBNF. If you have a better formal language to write down notations, ok.
I see your point, but it would require a general (if not universal)
adoption of the new notational system. Whether or not EBNF turns out to
be that system, I think it is instructive to consider EBNF. It seems to
be the definitive method for describing a computer language's
grammar/syntax, and yet I'm not sure that it has achieved anywhere near
universal adoption even among people working with languages.
[snip]
>> Do you have an example in mind where EBNF would be better (on either
>> dimension)? For Kendall/Lee notation, I'm inclined to think that EBNF
>> would be overkill: not any clearer than a tabular presentation, and
>> somewhat but not significantly more compact (?).
>
> Kendall/Lee is probably on the border of to small to have a big
> advantage with EBNF and to large not to try it. ...hmmm...no good
> example...I don't know. Maybe I can come up with something in the next
> days, but I don't promise anything.
>
> You probably (or better most certainly) have more experiance with
> notatios. Have you never thought, that the introduction of a notation in
> a paper or book lacked something or could have been more precise?
> Shouldn't there be at least some (more or less) formal guidelines when
> writing down notation?
Good question. I have definitely enocuntered this with journal
articles, but typically not so much as a result of an incomplete
specification as the result of the author using symbols before defining
them at all (or assuming, incorrectly in my case, that I would already
be familiar with the notation).
> Good night (at least here in Germany)
> Sascha
Guten Abend
-- Paul
|
|
| | From: | Sascha Herpers | | Subject: | Re: general question about notations | | Date: | Thu, 30 Dec 2004 16:47:12 +0100 |
|
|
| Hi,
> People with formal training in computer languages tend to be at least somewhat
> familiar with BNF, but I suspect a significant number of people
> (including many who write code, at least for their own use) are not.
yes, I admit, this is another downside of my wishful thinking.
> Quite true. Even more distressing, there are areas that have generally
> accepted notational conventions, and yet every so often some author
> feels obligated to create his or her own notation.
Yes, but that is a problem of young researchers (and I am one of them).
We always want to find or create something new. And we have to.
Uhm...let's see, what do we have here? Ah...well, let's call it 'a',
then the thing next to it must be 'b'...heureca, let's write a paper
about the symbolism of a and b. (I guess I am a little silly today. Sorry.)
> In a way, I find
> that even more confusing, since my brain instinctively reacts to a new
> symbol by assuming that it must not represent something for which a
> known symbol already exists.
Maybe that is another pro for my wish to formaly approach notations.
Take for example a capacitated lot sizing problem. There are several
auther who instinctivly mean the same thing, when writing about a CLSP,
but there a certain attributes they omit (because, these attributes are
to obvious to mention) and some which they name differently than the
other. If a CLSP would be formaly written down as an aggregation of
other symbols, which maybe in turn are aggregations (but eventualy boil
down to something atomic), it would certainly make it easier for newbies.
> [...EBNF...] and yet I'm not sure that it has achieved anywhere near
> universal adoption even among people working with languages.
Pity. It's probably because, it is not very readable if you are not
familiar with it.
Hm. Well. I gues the bottom line is, that there is no generally best or
at least good approach for writing down notations. Relaxing formality
would probably leave me with using the same way of defining a notation
as everybody else is using.
> Guten Abend
Uuh. Just out of curiosity, did you just pick up a few German words or
did you learn the language?
Sascha
|
|
| | From: | Paul A. Rubin | | Subject: | Re: general question about notations | | Date: | Thu, 30 Dec 2004 16:55:32 -0500 |
|
|
| Sascha Herpers wrote:
> Hi,
>
>> People with formal training in computer languages tend to be at least
>> somewhat familiar with BNF, but I suspect a significant number of
>> people (including many who write code, at least for their own use) are
>> not.
>
> yes, I admit, this is another downside of my wishful thinking.
>
>> Quite true. Even more distressing, there are areas that have
>> generally accepted notational conventions, and yet every so often some
>> author feels obligated to create his or her own notation.
>
> Yes, but that is a problem of young researchers (and I am one of them).
> We always want to find or create something new. And we have to.
> Uhm...let's see, what do we have here? Ah...well, let's call it 'a',
> then the thing next to it must be 'b'...heureca, let's write a paper
> about the symbolism of a and b. (I guess I am a little silly today. Sorry.)
Let's just say that I've reviewed equally silly papers whose authors
were apparently being serious when they submitted them. :-)
>
>> In a way, I find that even more confusing, since my brain
>> instinctively reacts to a new symbol by assuming that it must not
>> represent something for which a known symbol already exists.
>
> Maybe that is another pro for my wish to formaly approach notations.
> Take for example a capacitated lot sizing problem. There are several
> auther who instinctivly mean the same thing, when writing about a CLSP,
> but there a certain attributes they omit (because, these attributes are
> to obvious to mention) and some which they name differently than the
> other. If a CLSP would be formaly written down as an aggregation of
> other symbols, which maybe in turn are aggregations (but eventualy boil
> down to something atomic), it would certainly make it easier for newbies.
I agree in part and disagree in part. One concern that comes to mind is
whether the sort of notation you are suggesting would make underlying
assumptions (for instance, zero lead times, or free disposal of excess
inventory) more apparent or less apparent. One problem I've had with
some papers, typically those in which a mathematical programming model
is used in a business context, is that authors fail to state relevant
assumptions. In some cases, I'm not even sure the authors realize they
(implicitly) made those assumptions, and in other cases I suspect the
authors do not realize that their conditions would not be universally
applicable and hence automatically assumed by the reader. I'm afraid
that a mathematical formalism building models from symbols representing
other models (or submodels) might hide rather than expose the assumptions.
>
>> [...EBNF...] and yet I'm not sure that it has achieved anywhere near
>> universal adoption even among people working with languages.
>
> Pity. It's probably because, it is not very readable if you are not
> familiar with it.
>
> Hm. Well. I gues the bottom line is, that there is no generally best or
> at least good approach for writing down notations. Relaxing formality
> would probably leave me with using the same way of defining a notation
> as everybody else is using.
>
>> Guten Abend
>
> Uuh. Just out of curiosity, did you just pick up a few German words or
> did you learn the language?
I took classes in public school and university, to the point of
acquiring a reading knowledge (but not fluency -- I could order a beer,
but only because "bier" is a cognate :-). Then, after I got to graduate
school, I promptly forgot almost all of it (except "bier", of course).
-- Paul
>
> Sascha
>
>
|
|
| | From: | Sascha Herpers | | Subject: | Re: general question about notations | | Date: | Fri, 31 Dec 2004 08:03:16 +0100 |
|
|
| Hi,
> I'm afraid
> that a mathematical formalism building models from symbols representing
> other models (or submodels) might hide rather than expose the assumptions.
You are probably right.
> I took classes in public school and university, to the point of
> acquiring a reading knowledge (but not fluency -- I could order a beer,
> but only because "bier" is a cognate :-). Then, after I got to graduate
> school, I promptly forgot almost all of it (except "bier", of course).
At least you know the basics ;o)
Thanks for taking the time to write all this down.
Regards and best wishes for 2005...
Sascha
|
|
|
