Dozens of diagrams pointing here and there don't help. It is somehow as if the abstractness of category theory is abstract in the wrong way for me.
For example, just recently I came across this text: "Foundations of Algebraic Topology", by Eilenberg and Steenrod. Its preamble is highly readable and engaging, see below. We have a topology and compute some algebraic structure from it. Sounds easy!
------------------------------
The principal contribution of this book is an axiomatic approach to the part of algebraic topology called homology theory. It is the oldest and most extensively developed portion of algebraic topology, and may be regarded as the main body of the subject. The present axiomatization is the first which has been given. The dual theory of cohomology is likewise axiomatized. It is assumed that the reader is familiar with the basic concepts of algebra and of point set topology. No attempt is made to axiomatize these subjects. This has been done extensively in the literature. Our achievement is different in kind. Homology theory is a transition (or function) from topology to algebra. It is this transition which is axiomatized. Speaking roughly, a homology theory assigns groups to topological spaces and homomorphisms to continuous maps of one space into another. To each array of spaces and maps is assigned an array of groups and homomorphisms. In this way, a homology theory is an algebraic image of topology. The domain of a homology theory is the topologist's field of study. Its range is the field of study of the algebraist. Topological problems are converted into algebraic problems. In this respect, homology theory parallels analytic geometry. How ever, unlike analytic geometry, it is not reversible. The derived algebraic system represents only an aspect of the given topological system, and is usually much simpler. This has the advantage that the geometric problem is stripped of inessential features and replaced by a familiar type of problem which one can hope to solve. It has the disadvantage that some essential feature may be lost. In spite of this, the subject has proved its value by a great variety of successful applications. Our axioms are statements of the fundamental properties of this assignment of an algebraic system to a topological system. The axioms are categorical in the sense that two such assignments give isomorphic algebraic systems.
I don't know one that avoids diagrams as much as possible, but the book "Abstract and Concrete Categories" is not too diagram heavy. I think the PDF is legally available online.
Maybe I should write my own book about categories, that might keep me awake long enough.
Oh, maybe that is why there are so many books about category theory!
If you want to take it a little slower, you can start with Lawvere's
"Conceptual Mathematics"
https://api.pageplace.de/preview/DT0400.9780511590092_A23569...
Is there a short definition somewhere out there of what a topos is in terms of first-order predicate logic? Something I can understand without reading through 200 pages of preliminary material first?
I've seen statements that such a formulation in first-order logic would be misguided, because category theorists have their own notion of logic, but I'd like to understand it using my own notion of logic first.
Topos theory is a branch of mathematics which applies what programmers would call an aggressive refactoring to the category of sets and functions - a foundational workspace within which almost all of conventional math is conducted (whether the practitioners realise it or not). Math is refactored in a way reminiscent of how HOL refactors math (but constructively).
Set theory is a legacy platform like MS-DOS (!) with many limitations and anomalies which topos theory can explain and perhaps alleviate. A topos is a "virtual machine, for math ... Definitions, constructions, theorems "run" in a topos just as apps run on a VM, or SQL statements run on a database. The promise of topos theory is to cleanly separate language from implementation (just as webdesigners separate HTML from business logic) A lot of math can easily be "refactored" to apply in a much wider context.
The steps in building this refactoring are:
• Define the concept of category, a workspace of dots and composable arrows between them
• Identify the category Sets as fundamental
• Abstract out the operations and laws that make Sets useful (think CCCs (= cartesian closed categories) with some extras)
• Axiomatise a topos as a category equipped with operations obeying these laws
• Find other naturally occurring examples of topoi
• Via internal categories, understand topoi as a complete foundation for math
• Specify a language for describing constructions and deductions in a topos
Note, if you don't care about foundations, then topos theory gives you nothing new, except labour of re-learning what you already know in a new form that is awkward if you heavily rely on non-constructive reasoning.
An elementary topos is a cartesian closed category, with finite limits and a subojbect classifier.
Cartesian closedness means that for objects A and B there is an exponential object A^B of functions from B to A. Cartesian closedness is the right intuition on functions being first class citizens and is at the center of the equivalence CCC-lambda calculus-functional programming. Limits are bread and butter categorical stuff, and the pesky subobject classifier is sort of a pain of what Category Theory understands as classifying things.
In Set, monos into X (inyections into X, subsets of X), determine the characteristic function of the subset, subset of say, U. The characteristic function is U->Bool={True, False}. Summing up, the subobject classifier in Set is Bool and provides the correspondence of functions U->X and X->Bool. In an arbitrary elementary topos, the subobject classifier would be an Ω such that monic arrows ?->X corresponds to arrows X->Ω. I would agree that this has bad digestion, maybe delving in applications one just grow accustomed.
There is an idea of one doing mathematics in an "ambient" set theory, and categorists want to look at that as an ambient category of sets. But then they asked, what are the miminum features I am really using of this ambient category of sets? The list is the requirements of a category to be an elementary topos. So the category of sets is a topos (the topos of sets) very by design. But other categories also do. When one changes Sets to other topos is when weird intepretations emerge. Topos requirements don't let you recover the axiom of choice, for instance. Excluded middle is not available anymore.
"For instance, the definition of elementary topos (with operations defined by universal properties up to isomorphism, not specified as univalued operations) can be given as a finite set of sentences in FOLDS."
https://www.math.mcgill.ca/makkai/folds/foldsinpdf/FOLDS.pdf
Sounds interesting. Could someone elaborate on that?
An easy example is a function on a set. If you have function defined on the whole set, you can shrink it to give you a function defined on a subset. If you have functions defined on several subsets, and those functions agree on the overlaps of the subsets, then you can use that to define a function on the union of the subsets. More interesting examples arise in topology and related fields.
You take this property of a function that’s only defined in an arbitrarily small neighborhood of a point, and from it you can determine the function’s value anywhere else. That is, you take infinitesimally small changes (e.g. velocity) and add them up in the right way and get finite changes (e.g. distance).
It’s more interesting than it sounds because you aren’t computing a sum or something with numbers when you add up infinitesimal change. Local/infinitesimal change is in some ways a different beast than finite/global change.
It consists of a grid of overlapping slots and for each slot there is a clue.
The question that sheaf theory addresses is what constraints do you have to put on the clues to ensure that the overall puzzle has a single solution.
