Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I know that there are several posts on the same question. They all ask for examples for morphisms that are not functions. So, morphisms are more general than functions; they are the arrows connecting the objects of a category. However, I still cannot avoid the idea that they are functions.

By function, I mean exactly that the relation is well-defined. If $(a, b_1), (a, b_2)$ are inside the graph of $f: A \to B$, then $b_1 = b_2$. Therefore, if a morphism, say $g: A \to B$, is not a function, then I am allowed to have $(a, b_1), (a, b_2)$ inside the graph of $g$.

But, without the condition that they are well-defined, I do not see how associativity holds. For example, let:

  1. The graph of $f: A \to B$ contains $(a, b)$;
  2. The graph of $g: B \to C$ contains $(b, c_1), (b, c_2)$;
  3. The graph of $h: C \to D$ contains $(c_1, d_1), (c_2, d_2)$.
  4. The graph of $f': A \to D$ contains $(a, d_1), (a, d_2)$

So, $h((g \circ f)(a)) = d_1$ or $d_2$? Similarly, $(h \circ g)(f(a)) = d_1$ or $d_2$? This is what I mean; if I am forced to choose either $(a, d_1)$ or $(a, d_2)$, then indeed I am dealing with functions ....

share|cite|improve this question
8  
Morphisms are not necessarily relations at all! – Anthony Carapetis Dec 31 '15 at 3:31
1  
But there is in any case a perfectly good associative composition of relations. $xR\circ S z$ just means there's there's a $y$ with $xRy,ySz$. – Kevin Carlson Dec 31 '15 at 5:27
1  
Perhaps the following two "extreme" cases might help: (a) a monoid (or group) can be considered as a category with only one object, say $*$, and with morphisms the elements of the monoid which are considered as morphisms $*\to *$; (b) a preoder can be considered as category with objects the elements of the preorder and where there is a morphism $a \to b$ whenever $a\leq b$. – Nex Dec 31 '15 at 5:33
    
I never thought of associativity as Kevin Carlson described. But it works without morphisms being functions, as case (b) demonstrates. – Andy Tam Jan 2 at 3:10
    
Actually, case (a) also demonstrates that morphisms need not be functions. In case (a) the category has only 1 object (which could be anything, its arbitrary) and the morphisms are the elements of a monoid (which may not be functions). – kaiten Jan 2 at 4:14
up vote 11 down vote accepted

First, you're making too many assumptions. Why should morphisms be relations? Why should they have a graph? Why should A and B have elements?

Second, even with these assumptions, morphisms still don't need to be functions. Consider the category of Sets with arrows reversed. Then each morphism is like the reverse of a function. In particular, there are 2 distinct morphisms from {x,y} to {x} even though there is only one such function.

share|cite|improve this answer
1  
Does the term relation signify something specific? All I mean is that you can put $a \in A, b \in B$ (yes, by assumption, $A, B$ have elements) into a pair. This seems reasonable because a morphism is based on two objects. – Andy Tam Jan 2 at 2:50
1  
Thats basically what relation means, but in general there is no reason to assume the members of your category have elements. And even when they do have elements there is no reason to assume your morphism has anything to do with individual elements (which means it may not be a relation). – kaiten Jan 2 at 4:06

Morphisms, as already pointed out, could not be functions, or relations at all. In my category algebra class we saw morphisms as "arrows" between objects of the category. If we are in a concrete cathegory, as in the cathegory of sets or groups, a morphism is reasonably a function, a homomorphism, or something like this. However in an abstract category, such as in a category with two natural numbers $n, m$ as objects and morphisms are given by $n\times m$ matrices, relations don't play any role. You still want to assure that the morphisms are componible.

share|cite|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.