Coends basically have the following shape: you take an algebraic structure (covariant), and a geometric structure (contravariant) and you glue things together to form a global structure.
Let Fβ:βπopβ Γβ πβββπ± be an enriched functor, where π± is a symmetric monoidal category and π is π±-enriched.
Consider the product over all objects in π, βAMor(A,βA).
A wedge wβββF is an object w and a collection of maps wAβ:βwβββF(A,βA) such that the following commutes
w β F(A,A) β β F(B,B) $ \xrightarrow{(f,1)}$ F(A,B)
for all fβ:βAβββB.
An end is a universal wedge
Ends are ways of taking two kinds of structure and restricting to those pieces which behave the same in both structures.
Natural transformations: These are the
main example. The two types of structure are precomposition and
postcomposition
Think of the maps wAβ:βwβββF(A,βA) as something like the maps that make up a natural transformation
Because your end is a limit, you're basically looking at a collection {wA} such that they are invariant under altering the components of F(A, B) independently
F(A, A) $\xrightarrow{(1,f)}$ F(A, B) β β F(B, A) β w
We denote the coend by β«A F(A, A)
Unlike an end, which restricts to compatible structure; the coend glues two different kinds of structure together
Geometric realization: Glues together
the simplicial set blueprint with actual physical simplices
Tensor modules: Glues together two
different modules
The maps wAβ:βF(A,βA)βββw define your gluing
You essentially glue together the maps that change the components of F(A, B) independently.
This is meant to be something like the tensor product of modules
Similarly, here you want to think of C (its morphisms) as your ring, and the functors as your modules.
If we have F: C β D and K: C β E
C β D β E
The left Kan extension of F along K is
(LK F)[Y] = β«X Mor(K(X), Y) Δ F(X)
where Δ is the copwoer
β«A Mor(A,B) β F(A) β F(B)
β«A Hom(Mor(B, A), F(A)) β F(B)
Algebra: The simplex functor ΞβββTop that assigns a physical
simplex to each combinatorial simplex.
Geometry: We look at a fixed simplicial
set Xβ:βΞopβββSet.
Take a ring viewed as an abelian-enriched one object category.
Algebra: A left R-module Nβ:βRβββAb
Geometry: A right R-module Mβ:βRopβββAb
MβRNβ=ββ«RMβ Γβ N.
Algebra: Some functor that carries
algebraic data, which you'd like to extend to a larger space Gβ:βπββββ°
Geometry: A hom-functor Morβ(F(β),βX)β:βπopβββSet
The left Kan extension is β«cβββπMor(F(c),βX)β Γβ G(c).
If G acts on a topological space X, you often want to look at the orbit space X/G.
Algebra: We take a trivial map from the
delooping of G to the
singleton. BGβββTop.
Geometry: We take the map BGopβββTop
that sends the unique object to the space X, and the maps to a right
action.
The orbit space is the coend X/Gβ=ββ«BGXβ Γβ *.
Consider the category of open sets of a manifold
Algebra: A functor assigning smooth
functions to each open set Open(M)βββAlg
Geometry: The open sets Openβ(M)opβββTop
The end takes all of the open sets and produces the global functions that can be appropriately restricted to each local open set.
Algebra:
Geometry:
We again consider the delooping of some group acting on a space.
Algebra: We take the representation of
G acting on the left on our
space BGβββTop.
Geometry: We take the map BGopβββTop
that sends the unique object to the singleton.
The end is the space of G-invariants
β«BGMor(*,βX)
Algebra:
Geometry: