On Aug 18, 6:13 pm, António Marques <
ento...@gmail.com> wrote:
> On Aug 18, 5:48 pm, Jerry Friedman <
jerry_fried...@yahoo.com> wrote:
> > On Aug 18, 9:30 am, "Peter T. Daniels" <
gramma...@verizon.net> wrote:
...
> > > Hmm, is there more than one empty set? Is the set of all unicorns the
> > > same as the set of all living people 150 years old?
>
> > In mathematics there's only one empty set. (I don't know whether
> > there might be some non-standard approach or non-mathematical approach
> > with different empty sets.)
>
> In computer programming, there sure are different empty sets; an array
> of type X is not generally replaceable with an array of type Y, even
> if both are empty.
I'd call those empty arrays. If programmers call them empty sets,
then I'd say there's only one empty set in the sense of set theory.
> And you can add elements to an empty set or values to a numeric object
> of value 0, whereas you can't do anything with a null set or a null
> number. The point here is that even if you dismiss these as artefacts
> of notation or memory management, they do serve purposes. For
> instance, you can have a function the purpose of which is to determine
> the position of the first occurrence of a character in a string. If
> the character isn't part of the string, it's not correct to reply '0'.
> The usual solution to problems like these has been to reply '-1',
> which will sort of worjk as long as all the valid replies are
> positive; however, the elegant and meaningful reply would be 'null'.
> And boolean values? True is 'yes', false is 'no'. What is null? The
> correct thing to consider is that it means 'undetermined'.
Fine, but that's not the empty set in the sense of set theory.
> For me, if you say a 'line segment' has an area of 0, I'll consider
> that you're in fact talking about a closed polygon with at least one
> side of length 0.
I'm talking about a set of points in two or more dimensions. It
doesn't have to be bounded--for instance, the area of the points (x,
y) with x > 0 and 0 < y < e^(-x), which is the area under the curve of
y = e^(-x) from zero to infinity, is well defined (and equal to 1).
> Otherwise, take a triangle and remove one of the
> sides. What does 'area' mean when applied to the remaining shape?
We've got three points in a plane, A, B, and C. The area of the set
of points bounded by the line segments AB, BC, and CA--a triangle--is
given by our choice of well-known formulas [•]. By "remove one side",
I take it you mean "the set of points bounded by the line segments AB
and BC". This expression is meaningless; there is no such set of
points. So indeed, "area" doesn't mean anything when applied to it.
(On the other hand, the area of the three line segments, that is, the
sides of the triangle without the interior, is 0, and if you remove
one of them, it's still 0.)
[From another post]
> I think what you're saying is that there is a way to define 'area' so
> that it applies to things which's
Heh.
> definition can be used for line segments.
As long as it's clear that "you" includes all mathematicians, except
possibly for a set of measure 0 :-) For instance, see page 5 at
http://terrytao.files.wordpress.com/2011/01/measure-book1.pdf
for the idea that points have 0 length and "boxes" can have 0 volume.
(That includes line segments with 0 area, volume, etc.)
This is also stated at Wikipedia, which I generally trust on math, and
at Mathworld, where it's not quite so explicit as it is in Terence
Tao's book.
> That doesn't mean that our concept of line segments admits
> area.
Mine does. Intuitively, as you make a rectangle thinner, it
approaches a line segment, and its area approaches 0.
Here's another appeal to intuition. It's part of our intuitive idea
of measure that if A and B and disjoint (non-intersecting) sets, the
measure of the union of A and B equals the measure of A + the measure
of B, right?
Consider two subsets of the real-number line: A = [0, 1) (the numbers
from 0 to 1 including 0 but not 1), and B = {1}. The union of those
sets is C= [0, 1] (the numbers from 0 to to 1 including both 0 and
1). Both A and C have length 1. It works perfectly to say that the
length of C = the length of A plus the length of B, as 1 = 1 + 0. To
say that the length of B is undefined and we can't use the equation is
to make a simple and perfectly workable thing complicated and
unworkable.
(Apologies to those who didn't need the definitions of the intervals,
etc. Some here may have wanted them.)
> Something's been lost in the abstraction. Like, but not
> necessarily homomorphic, with 'no is maybe and maybe is yes'.
I have no idea what part of measure theory your "no is maybe" phrase
is supposed to connect to. Of all the things to accuse of lacking
rigor and self-consistency! Or am I misunderstanding you.
[*] "Well-known" doesn't necessarily mean I remember them, other than
A = bh/2.
--
Jerry Friedman