EPR?

Resources on Lenses

Doctrines

Abstract Finite categories Kind of like a graph data structure. Kind of like a finite group. A list of objects, a list of morphisms with domain and codomain, and a multiplication table. Any particular question can be answered via enumeration

Finitely Generated Categories

Axioms

Category is objects and morphisms. Morphisms have a partial operation called composition and there is an identity morphism for every object

Examples

Groups and Monoids

The algebraic laws of monoid basically work for a category with one object, where multiplication is composition

A group can be modeled as a category with one object, where ever morphism has an inverse. Then the axioms of cateogory corresponds to the axioms of a group

PoSet

partially ordered sets. There is at most one morphism between each object

Lattices are an interesting further subcase.

https://en.wikipedia.org/wiki/Partially_ordered_set

FinSet

Finite sets as objects, finite maps (dicitonaries) as morphisms

Can also consider partial finite maps.

FinVect

vector spaces are objects (dimensionality or labelled). Linear maps are morphisms (~ matrices)

Categories and Polymorphism

Part of the appeal of category theory comes from it’s relationship to polymorphism. Polymorphism is counterintuitive really in regards to the simple model of types being sets.

https://bartoszmilewski.com/2014/10/28/category-theory-for-programmers-the-preface/

https://arxiv.org/pdf/2209.01259.pdf Category Theory for Programming

Combinators

https://en.wikipedia.org/wiki/Combinatory_logic

Combinators are little functional pieces you can combine like lego blocks. When you have a language of combinators, variable binding issues are gone. They are hard to program with

Combinators are not synonymous with categories, but categories can inform a particular style of combinators

Encodings

Encoding category theory to first order, higher order logic, and dependent type theory. Generalized algebraic theories.

Relational Composition comp(F,G,H)

Comp as Partial function comp(F,G) = Some(H)

Objects as Sort

Blog posts

• Egglog 2: Automatically Proving the Pullback of a Monic is Monic https://www.philipzucker.com/egglog2-monic/
• Egglog Examples: Pullbacks, SKI, Lists, and Arithmetic https://www.philipzucker.com/egglog-3/
• Rewriting Monoidal Categories in the Browser with Egg

Functors

Functors are mappings between categories. This means they are a tuple of map from objects to objects and morphism to morphism such that composition plays nice (commutes sort of) Functors are mappings between categories (a object to object map and morphism to morphism map) that plays nice with composition and identity. F(id) = id. F(f.g) = F(f) . F(g)

F = (Fo, Fm) Fm(f . g) = Fm(f) . Fm(g)

Representations of groups. Functor from group as a category to

Free/Forgetful

Galois Connections. Abstraction and concretization

Closure/Rounding and lift

Abstract interpretation Intervals <-> sets polytopes convexsets <-> sets

Vect

Higher Category

https://en.wikipedia.org/wiki/Higher_category_theory

Morphisms go between objects 2-morphisms go between morphisms

An example I like is Rel. Morphisms are binary relations. binary relations can be compared via inclusion (they form a partial order themselves).

Topos

Use functions instead of a notion of element being member of set. functions can essentially select elements. functions into “truth value” set (indicator functions) are analogs of subsets.

https://jonsterling.notion.site/Topoi-inside-and-out-7b0b86e39eeb43aeaee3c3af1dd91f2a jon sterling blog post

To say something is a topos is to say it is

1. a category
2. a special category with some special constructs

Weird notions of truth value are fun.

Time

Categorical Databases

schema is a finite category Data lives over it Mappings between schema describe

Logic

Rising Sea

A proof of A |- B is the basic “morphism”. A proof is a tree. We can perhaps annotated this sequent with free variables and unification variables in play (signature). This makes this morphism floating over some kind of variable set, something sheafy? Or a set floats of variables floats over the proof

• Objects are propositions.
• Cut is compose.
• Axiom is id.
• Existentials and universals as adjunctions Different proofs are different morphisms between the same objects.

Computational Category Theory

hey I got on hacker news

catlab.jl

homotopy.io globular dicopy cartographer

From a programming perspective I think there are a couple contributions:

• Category theory has a number of very intuitive looking graphical notations which nevertheless translate to very exact algebraic expressions. This is pretty dang nice. Category theory is a road to a very principled formulation of things that are already done in dataflow languages, tensor diagrams, and UML and things like that. These graphical languages are a reasonable user facing interface.

• Empirically in mathematics, category theory has been found to be a flexible and “simple” language to describe and interconnect very disparate things. In programming terms that means that it is a plausible source of inspiration for consistent interfaces that can do the same.

• Point free DSLs are really nice to work with as library/compiler writer. It makes optimizations way easier to recognize and implement correctly. Named references are phantom links attaching different points in a syntax tree and make it hard to manipulate them correctly. In slightly different (and vague) terms, category theory gives a methodology to turn abstract syntax graphs into trees by making all links very explicit.

I can’t comment much on what category theory contributes to mathematics, but I think there are similar points. I think category theory can make precise analogies between extremely disparate fields like logic, set theory, and linear algebra. I think categorical intuitions lead to clean definitions when specialized to some particular field. Combinatory categorical logic does make some aspects of logic less spooky.