home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: sci.math
- Path: sparky!uunet!psinntp!scylla!daryl
- From: daryl@oracorp.com (Daryl McCullough)
- Subject: Re: Those Naughty Category Theorists
- Message-ID: <1992Dec14.170338.4977@oracorp.com>
- Organization: ORA Corporation
- Date: Mon, 14 Dec 1992 17:03:38 GMT
- Lines: 67
-
- pratt@Sunburn.Stanford.EDU (Vaughan R. Pratt) writes:
-
- >Category theory: The ordered pair <a,b> is an entity uniquely
- >determined by entities a and b and uniquely determining those same a
- >and b via a rule which for every ordered pair determines some a and b.
-
- It took me forever to get to the point where the category theory
- definitions of cross products, disjoint unions, etc. made sense
- to me. (Now, they are all intuitively obvious, of course.) Let
- me try my hand at it:
-
- First of all, category theory doesn't really define what it means to
- be *an* ordered pair, it only defines what it means for one object to
- be the cross product of two other objects.
-
- I'm going to use type theory, rather than category theory, but the
- concepts can all be translated into category theory. Let's write down
- some properties we want a cross product to have:
-
- If an object C is to act like A x B, then we need to have left and
- right projection functions: l: C -> A, and r: C -> B. Given *any*
- C and any l and r, we can represent some ordered pairs as elements of
- C. Element c of C represents the ordered pair <l(c), r(c)>
-
- Now, although any C,l, and r can represent some ordered pairs of A x B,
- there can be two kinds of flaws in our representation: (1) Some pairs
- may not be represented, and (2) Some pairs may be represented twice--that
- is, there may be c1 and c2 such that c1 is not equal to c2, but
- the corresponding pairs, <l(c1), r(c1)>, <l(c2),r(c2)> are equal.
-
- To take care of the first problem, we demand that C, l, and r can
- represent as many pairs as any other representation can. That is, if
- we have some other candidate for representing pairs D, f, g, then
- there is for every pair represented in D a corresponding pair
- represented in C. Mathematically, if D is some object, f: D -> A,
- g: D -> B, then D also represents pairs. To say that C represents
- as many pairs as D does means that there is a function q: D -> C
- such that for any d, q(d) represents the same pair as d. That is,
- l(q(d)) = f(d) and r(q(d)) = g(d). This can be written without
- mentioning elements by using function composition:
-
- C, l, r represents as many pairs as D, f, g if there is some
- q: D -> C such that l o q = f and r o q = g, (where l o q is
- the function defined by l o q (d) = l(q(d))).
-
- (Note, I may have the order wrong for composition; I'm not sure
- if the convention is l o q or q o l.)
-
- We can then say that C, l, r represents all pairs in A x B if
- for all D,f,g it represents as many pairs as D, f, g.
-
- To take care of the second problem, we demand that there not be two
- different ways of representing the same pair in C, l, r. We can insure
- this by asking that for all D, f, and g, there must be a *unique*
- function q such that l o q = f and r o q = g.
-
- Thus we get the category theory definition of cross-product:
- C,l,r represents the cross product A x B if l: C -> A, r: C -> B,
- and for all D,f,g such that f: D -> A, g: D -> B, there exists a
- unique q: D -> A such that l o q = f and r o q = g.
-
- (For category theory, read f: D -> A as "f is a morphism from
- D into A")
-
- Daryl McCullough
- ORA Corp.
- Ithaca, NY
-