2

The following question is taken from Arrows, Structures and Functors the categorical imperative by Arbib and Manes

In chapter 7 of Arbib and Manes about Functors. The authors introduce the category of generalised monoid as motivation and precursor to introducing functors.

$\color{Green}{Background:}$

$\textbf{(1) Definition:}$

  1. We have seen that a one-object category is a monoid, let us now see a sense in which any category is a $\textit{generalized}$ monoid. Given $\textbf{K},$ let $$\textbf{M}_{\textbf{k}}=\prod_{A,B\in \mathrm{Obj }\textbf{K}}\textbf{K}(A,B)$$ be the collection of all $\textbf{K}-$morphisms. Composition definees a $\textit{partial}$ function $$\textbf{M}_{\textbf{k}}\times\textbf{M}_{\textbf{k}}\to\textbf{M}_{\textbf{k}}:(f,g)\mapsto g\circ f.$$

Let us see how we may recapture the identities of $\textbf{K}$ from this function: We say that $u$ is an identity if $g\circ u=g$ whenever $g\circ u$ is defined, and if $u\circ f=f$ whenever $u\circ f$ is defined. As before, each $f$ has exactly one identity $u$ such that $u\circ f$ is defined, call it $C(f)$ for the identity of the $\textit{codomain}$ of $f$; and exactly one $v$ such that $f\circ v$ is defined, call it $D(f)$ for the $\textit{domain}$ of $f.$ Thus our partial function $\textbf{M}_{\textbf{k}}\times \textbf{M}_{\textbf{k}} \to \textbf{M}_{\textbf{k}}$ satisfies the conditions:

$2a)$ There are total functions $$C,D: \textbf{M}_{\textbf{k}}\to \text{identities of } \textbf{M}_{\textbf{k}}$$

for which

$$2a-i)\quad \quad D(D(f))=D(f)=C(D(f));$$ $$2a-ii) \quad D(C(f))=C(f)=C(C(f)); \text{ and,}$$ $$2a-iii)\quad \quad f\circ D(f)=f=C(f)\circ f.$$

$2b)$ $g\circ f$ is defined iff $C(f)=D(g).$ If $g\circ f$ is defined then $D(g\circ f)=D(f)$ and $C(g\circ f)=C(g).$ Moreover, if either $(h\circ g)\circ f$ or $h\circ (g\circ f)$ is defined, then both are defined and $(h\circ g)\circ f=h\circ (g\circ f).$ [In fact, given a partial function satisfying $1$ and $2$ we can reconstitute a category, with objects being the identities.]

$\textbf{(2) Exercise:}$

Let $\textbf{M}\times \textbf{M}\to \textbf{M}$ be a partial function satisfying $2a.$ and $2b.$ Introduce the notation $f:k'\to k$ to indicate that $D(f)=k'$ and $C(f)=k.$ Deduce that this yields a category, with objects the identities, whose 'generalized monoid' is $\textbf{M}\times \textbf{M}\to \textbf{M}.$

[Solution:]

Let $\textbf{M}$ be a category and if $k, k', k''' \in \textbf{M},$ are objects and given any map $f:k'\to k,$ in $\textbf{M}$ and from the notation in the question for $D(f)=k',$ we can interpret it to mean $D(f)=k':=i:k'\to k',$ similarly, for $C(f)=k:=i:k\to k.$ So let $i \text{ (identity map) }, f:k'\to k, g:k\to k'', h:k'\to k''',$ be maps in $\textbf{M}$ and let $v,u$ be identity maps where they are respectively defined as $v:D(-)\to D(-), u:C(-)\to C(-),$ and $D(-)=(-)\circ v, C(-)=u\circ (-).$ Then according to $(2a-i)-(2a-iii)$ we have the following table for what $D(-).C(-).D(D(-)).C(D(-)).D(C(-)).C(C(-)).(-)\circ D(-).C(-)\circ (-). C(-)=D(-). D((-)\circ (*)) . C((*)\circ (-)) . D((-)\circ (*))=D(*).C((*)\circ (-))=C(*)$ equal to:

$\tiny\begin{array}{c|c|c|} & & 1 & 2 & 3 & 4 & 5 & 6 & 7 & 8 & 9\\ \hline &(-)/(-)\circ (-)/(-)\circ ((-)\circ (-))/((-)\circ (-))\circ (-) & D(-)&C(-)&D(D(-))&C(D(-))&D(C(-))&C(C(-))&(-)\circ D(-)&C(-)\circ (-)& C(-)=D(-)\\ \hline 1&i & D(i) & C(i) \\ \hline 2& f & k' & k &k'& k'& k & k & f & f \\ \hline 3& g & k & k'' & k & k &k''& k''& g & g \\ \hline 4& h & k'' & k''' & k'' & k'' & k''' & k''' & h & h &\\ \hline 5& g\circ f & k' & k'' & k' & k' & k'' & k'' & g\circ f & g\circ f & C(f)=D(g)=k \\ \hline 6& h\circ g & k & k''' & k & k & k''' & k''' & h\circ g & h\circ g & C(g)=D(h)=k'' \\ \hline 7&(h\circ g)\circ f & k' & k''' & k' & k' & k''' & k''' &(h\circ g)\circ f &(h\circ g)\circ f & C( f)=D(h\circ g)=k \\ \hline 8&h\circ (g\circ f) & k' & k''' & k' & k' & k''' & k''' &h\circ (g\circ f) &h\circ (g\circ f) & C(g\circ f)=D(h)=k''\\ \hline \end{array}$

$\tiny\begin{array}{c|c|c|} & & 10 & 11 & 12 & 13\\ \hline &(-)/(-)\circ (-)/(-)\circ ((-)\circ (-))/((-)\circ (-))\circ (-) & D((-)\circ (*)) & C((*)\circ (-)) & D((-)\circ (*))=D(*)&C((*)\circ (-))=C(*)\\ \hline 1&i & & \\ \hline 2& f & & & & \\ \hline 3& g & & & & \\ \hline 4& h & & & &\\ \hline 5& g\circ f & k' & k'' & k' & k'' \\ \hline 6& h\circ g & k & k''' & k & k'''\\ \hline 7&(h\circ g)\circ f & k' & k''' & k' & k'''\\ \hline 8&h\circ (g\circ f) & k' & k''' & k' & k'''\\ \hline \end{array}$

$\color{Red}{Questions:}$

To show that $\textbf{M}\times \textbf{M}\to \textbf{M},$ the partial function satisfying (2a) and (2b) is a category for the exercises above, I have to show that it has the identity, and composition morphisms, as well as they satisfy the rule for associativity according to the rules (2a-i) to (2a-iii) along with the values of $D(-).C(-).D(D(-)).C(D(-)).D(C(-)).C(C(-)).(-)\circ D(-).C(-)\circ (-). C(-)=D(-). D((-)\circ (*)) . C((*)\circ (-)) . D((-)\circ (*))=D(*).C((*)\circ (-))=C(*)$ for the morphisms $i,f, g, h, h\circ (g\circ f), (h\circ g)\circ f.$ I constructed the table showing their values. When the exercise asks: $\textbf{"Deduce that this yields a category, with objects the identities, whose 'generalized monoid' is}$ $\textbf{M}\times \textbf{M}\to \textbf{M}.$" Do they mean the following:

To show that the map $\textbf{M}\times \textbf{M}\to \textbf{M}.$ defined by partial functions satisfies the definitions of the monoid axioms, we first note that as in the definition from (1) above, let $\textbf{M}$ be a category, and objects in $\textbf{M}$ are set $K,$ and the set of morphisms $\textbf{M}(K)$ consist composition maps ${\textbf{M}}(K)\times {\textbf{M}}(K)\to {\textbf{M}}(K)$ defined by $(f,g)\mapsto g\circ f,$ $f, g\in \textbf{M}(K),$ and $f,g:X\subset K\to K$ Hence we can let $k',k, k'',k'''$ all be subsets of $K.$

The identity morphism $i\in \textbf{M}(K)$ composing with itself $(i,i)\mapsto i\circ i$ is certainly in ${\textbf{M}}(K)\times \textbf{M}(K)\to {\textbf{M}}(K),$ since $i\circ i=D(i)$ and $i\circ i=C(i).$

Now any morphism $f\in \textbf{M}(K)$ can be written as a composition of $f$ and the identity morphism $i$. Meaning $(f,i)\mapsto i\circ f$ and $(i,f)\mapsto f\circ i$ are both in ${\textbf{M}}(K)\times \textbf{M}(K)\to {\textbf{M}}(K).$ If we let $f$ to be the map $f:k'\subset K\to k\subset K$ and $i$ the identity map, then $i\circ f=C(f)$ and $f\circ i=D(f).$

For composition of morphism, again, let $i$ be the identity map, the map $f$ as just defined and $g\in \textbf{M}(K),$ defined by $g:k\subset K\to k'\subset K,$ then $(i,g\circ f)\mapsto (g\circ f)\circ i$ and $(g\circ f, i)\mapsto i\circ (g\circ f)$ are both in ${\textbf{M}}(K)\times \textbf{M}(K)\to {\textbf{M}}(K).$ From the above table, we can see that $i\circ (g\circ f)=C(g\circ f),$ and $(g\circ f)\circ i=D(g\circ f).$

For associativity of morphisms, let $h\in \textbf{M}(K)$ defined by $h:k''\subset K\to K'''\subset K,$ $i$ the identity morphism and morphisms $f,g$ as before. Then from the table above, $i\circ (h\circ (g\circ f))=C(h\circ (g\circ f))$ and $(h\circ (g\circ f))\circ i=D(h\circ (g\circ f)).$ Hence $(i, h\circ (g\circ f))\mapsto (h\circ (g\circ f))\circ i,$ and $(h\circ (g\circ f),i)\mapsto i\circ (h\circ (g\circ f)),$ are both in ${\textbf{M}}(K)\times \textbf{M}(K)\to {\textbf{M}}(K).$ Meaning composition of morphisms holds for ${\textbf{M}}(K)\times \textbf{M}(K)\to {\textbf{M}}(K).$ So ${\textbf{M}}(K)\times \textbf{M}(K)\to {\textbf{M}}(K).$ satisfies the definition for a monoid and hence it is a generalized monoid.

$\textbf{Also,}$ is the way I constructed the table above clear enough in terms of details?

One other minor questions: Are there tags for question clarifications over wordings, and not just for notations.

Thank you in advance

Seth
  • 4,043
  • 1
    Super long formulas seem to just go off the margins on this site, so please refrain from using them. Also, the edit history is public, so no need to say "edited". We will be able to see each revision of your edit regardless. – Trebor Sep 29 '23 at 12:55
  • @Treber I don't know how to keep tables within margins, especially when there is a large table that need to be created. – Seth Sep 29 '23 at 12:58
  • Since the function $M\times M\to M$ isn't really a function, just a partial function, I fail to see why we can expect the full monoid axioms to hold. Do your authors ever define "generalised monoid"? If they don't, then I just wouldn't worry about it. Just focus on creating a category, with objects the identities, and leave it at that – FShrike Sep 29 '23 at 13:21
  • @FShrike ah where I quoted in (1) Definition above is the first time I seen the term "generalised monoid". I did manage to show how $f$ can be written in terms of $(f,i)$ and $(i,f)$. also the same for $g\circ f$ and $h\circ(g\circ f), (h\circ g)\circ f.$ I think that is all I need to show that the monoid axioms are satisfied as well as the axioms for categories by using partial maps $(f,g)\mapsto g\circ f.$ – Seth Sep 29 '23 at 14:30

0 Answers0