-
-
Notifications
You must be signed in to change notification settings - Fork 0
/
chap-filt-power.tex
144 lines (115 loc) · 5.87 KB
/
chap-filt-power.tex
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
\chapter{Power of filters}
\section{Germs of functions}
\begin{defn}
Functions $f, g \in \mathbf{Rel} (\Ob \mathcal{X} , B)$
\emph{are of the same $\mathcal{X}$-germ} for a filter object
$\mathcal{X}$ iff there exists $X \in \up \mathcal{X}$ such that $f|_X
= g|_X$.
\end{defn}
\begin{prop}
Being of the same germ is an equivalence relation.
\end{prop}
\begin{proof}
~
\begin{description}
\item[Reflexivity] Take arbitrary $X \in \up \mathcal{X}$.
\item[Symmetry] Obvious.
\item[Transitivity] Let $f|_X = g|_X$ and $g|_Y = h|_Y$. Then $f|_{X \cap
Y} = h|_{X \cap Y}$.
\end{description}
\end{proof}
\begin{defn}
A \emph{germ} is an equivalence class of being the same germ.
\end{defn}
\begin{obvious}
Every germ is a filter on $\mathbf{Set}$.
\end{obvious}
\begin{thm}
Let $A$, $B$ be sets.
The following are mutually inverse bijections between monovalued reloids $f
: A \rightarrow B$ with $\dom f = \mathcal{X}$ and $\mathcal{X}$-germs
$S$ of functions $A \rightarrow B$ for $\mathcal{X} \in \mathscr{F} A$:
\begin{enumerate}
\item $f \mapsto \up^{\mathbf{Set}} f$;
\item $S \mapsto s|_{\mathcal{X}}$ if $s\in S$.
\end{enumerate}
The second bijection can also be written as $S \mapsto \left(\bigsqcap^{\mathsf{RLD}} S\right)|_{\mathcal{X}}$ or
if $\card B\ne 1$ as $S \mapsto \bigsqcap^{\mathsf{RLD}} S$.
\end{thm}
\begin{rem}
$s|_{\mathcal{X}}$ is always defined because $S$ is nonempty (it is an equivalence class).
\end{rem}
\begin{proof}
First prove that $\up^{\mathbf{Set}} f$ is an
$\mathcal{X}$-germ. Really, $F \in \up^{\mathbf{Set}} f
\Leftrightarrow F \sqsupseteq f \Leftrightarrow F|_{\mathcal{X}} = f
\Leftrightarrow \exists X \in \up \mathcal{X} : F|_X \sqsupseteq f$;
thus $F, G \in \up^{\mathbf{Set}} f \Rightarrow \exists X \in
\up \mathcal{X} : F|_X \sqsupseteq f \wedge \exists Y \in \up
\mathcal{X} : G|_Y \sqsupseteq f \Rightarrow \exists X \in \up
\mathcal{X} : F|_{X \cap Y} \sqsupseteq f \wedge \exists Y \in \up
\mathcal{X} : G|_{X \cap Y} \sqsupseteq f \Rightarrow \exists Z \in
\up \mathcal{X} : (F|_Z \sqsupseteq f \wedge G|_Z \sqsupseteq f)
\Rightarrow \exists Z \in \up \mathcal{X} : (F \sqcap G) |_Z
\sqsupseteq f$ and $F \in \up^{\mathbf{Set}} f \wedge \exists
X \in \up \mathcal{X} : F|_X = G|_X \Rightarrow F \sqsupseteq f \wedge
F|_{\mathcal{X}} = G|_{\mathcal{X}} \Rightarrow G|_{\mathcal{X}} \sqsupseteq
f \Rightarrow G \in \up^{\mathbf{Set}} f$. We have proved
that $\up^{\mathbf{Set}} f$ is an equivalence class of the
suitable equivalence relation, that is $\up^{\mathbf{Set}} f$
is an $\mathcal{X}$-germ.
That $\bigsqcap^{\mathsf{RLD}} S$ is a monovalued reloid is obvious.
Also $\im \bigsqcap^{\mathsf{RLD}} S = \mathcal{X}$ is obvious.
Now prove that our correspondences are mutually inverse.
Let $f_0 : A \rightarrow B$ be a monovalued reloid and $\dom f =
\mathcal{X}$. Let $S = \up^{\mathbf{Set}} f_0$ and $f_1 =
s|_{\mathcal{X}}$ for an~$s\in S$. We need to prove $f_1 = f_0$. Really,
$f_1 = F|_{\mathcal{X}}$ for an $F\in\up^{\mathbf{Set}}f_0$; thus $f_1=f_0$.
Let $S_0$ be an $\mathcal{X}$-germ of functions $A \rightarrow B$. Let $f =
s|_{\mathcal{X}}$ for an~$s\in S_0$ and $S_1 =
\up^{\mathbf{Set}} f$. We need to prove $S_1 = S_0$. Really,
\[
S_1 = \up^{\mathbf{Set}}(s|_{\mathcal{X}}) =
\setcond{F\in\mathbf{Set}}{F\sqsupseteq s|_{\mathcal{X}}} =
\setcond{F\in\mathbf{Set}}{\exists X\in\up\mathcal{X}:F|_X\sqsupseteq s|_X} =
\setcond{F\in\mathbf{Set}}{\exists X\in\up\mathcal{X}:F|_X=s|_X} = S_0.
\]
$\left(\bigsqcap^{\mathsf{RLD}} S\right)|_{\mathcal{X}} =
\bigsqcap^{\mathsf{RLD}}_{s\in S} s|_{\mathcal{X}} = s|_{\mathcal{X}}$
for every choice of~$s\in S$.
We can assume that $B \neq \emptyset$ because otherwise the theorem is
obvious. Thus we can assume $\card B>1$.
If $\mathcal{X} = X$ then obviously $S$ has just one element $F$ and
$\im \bigsqcap^{\mathsf{RLD}} S = \im F = X =
\mathcal{X}$. Otherwise for every $X \in \up \mathcal{X}$ there are
elements $F$, $G$ of $S$ such that $\dom (F \sqcap G) \sqsubseteq X$
(using $\card B > 1$).
By properties of generalized filter bases $X \times \top \sqsupseteq
\bigsqcap^{\mathsf{RLD}} S \Leftrightarrow \exists F, G \in S : X
\times \top \sqsupseteq F \sqcap G \Leftrightarrow X \sqsupseteq
\mathcal{X}$. Thus $\im \bigsqcap^{\mathsf{RLD}} S =
\mathcal{X}$.
\end{proof}
\section{Power of filters}
Let's define $\mathcal{Y}^{\mathcal{X}}$ for filters~$\mathcal{X}$,~$\mathcal{Y}$:
First define $Y^{\mathcal{X}}$ for a set~$Y$:
\[ Y^{\mathcal{X}} = \setcond{ f \in \mathsf{RLD} (\Ob \mathcal{X}
, Y) }{ \dom f = \mathcal{X} \wedge f\text{ is monovalued} } . \]
Now $\mathcal{Y}^{\mathcal{X}} = \bigsqcap^{\mathsf{RLD}}_{Y \in
\up \mathcal{Y}} Y^{\mathcal{X}}$.
\cite{filt-cat}~defines an isomorphic to this way to define ``exponentiation'' of filters.
TODO: Check $\mathcal{Y}^1 \cong \mathcal{Y}$; $\mathcal{Z}^{\mathcal{X}
\times^{\mathsf{RLD}} \mathcal{Y}} \cong
(\mathcal{Z}^{\mathcal{X}})^{\mathcal{Y}}$; $\mathcal{Z}^{\mathcal{X} \amalg
\mathcal{Y}} \cong \mathcal{Z}^{\mathcal{X}} \times^{\mathsf{RLD}}
\mathcal{Z}^{\mathcal{Y}}$; $\mathcal{Y}^2 \cong \mathcal{Y}\times^{\mathsf{RLD}}\mathcal{Y}$;
$\mathcal{Y}^0 \cong 1$; $\mathcal{Y}^N \cong \prod^{\mathsf{RLD}}_{n\in N}\mathcal{Y}$.
More formulas at \url{https://en.wikipedia.org/wiki/Cartesian_closed_category}.
Andreas Blass says in a private email that it is not cartesian closed: ``Unfortunately, the two categories of filters in my paper are
not cartesian closed. This is mentioned in a parenthetical comment
near the bottom of page 141. The operation of cartesian product with
the cofinite filter on the natural numbers has no right adjoint,
because it does not preserve infinite coproducts.''
about~\cite{filt-cat}.
But it is probably a braided closed monoidal category?
See \cite{filt-cat} for more categorical properties of filters.