2  What Is Mathematics?

You can differentiate a function, invert a matrix, and run gradient descent, and still not be able to answer a question a first-year graduate student would consider basic: what is the thing you just did that to? Is a vector an arrow, a list of numbers, or neither? Is a matrix a grid of numbers or a transformation wearing a grid as a costume? When you write \(Ax = b\), what kind of object is \(x\), really?

This isn’t a trick question, and the gap it exposes isn’t a gap in your computational skill. Most STEM curricula teach mathematics as a sequence of toolkits — differentiate this, integrate that, solve this system — because that’s what’s useful fast. What they often skip is the layer underneath: why these particular tools were invented, what problem forced each one into existence, and how apparently unrelated tools turn out to be the same idea wearing different clothes. That layer is what this site is for.

2.1 Objects, structures, theories, applications

One habit does more than any other to organize the rest of this site: keep four different kinds of thing separate, even when they get called by similar names.

  • A mathematical object is a specific thing: the number 7, the vector \((1, 2, 3)\), the function \(\sin(x)\).
  • A mathematical structure is a pattern an object can carry: group, vector space, metric space. A structure is a set equipped with operations or relations that satisfy some fixed list of axioms. Objects instantiate structures.
  • A theory is the body of results about a structure or a family of problems: linear algebra studies vector spaces and the maps between them; probability theory studies probability spaces.
  • An application is a domain that borrows a theory to solve a problem that isn’t, itself, mathematics: physics, engineering, machine learning, economics.

\(\sin(x)\) is an object. It lives inside a vector space of functions (a structure). Functional analysis is the theory that studies that structure in generality. A physicist using \(\sin(x)\) to model a pendulum is an application. Four different levels, one sentence — and conflating any two of them is exactly the kind of confusion that leaves a computationally fluent reader unable to say what a vector is. What Is a Mathematical Structure? develops this distinction in full; it’s named here first because every later chapter leans on it.

2.2 The pattern behind every chapter

Nearly every idea on this site follows the same arc, and it’s worth seeing once, explicitly, before it starts appearing everywhere:

Problem:       we want to describe quantities that can be combined and undone
                    ↓
Need:          a set, plus an operation on it
                    ↓
Problem:       different operations have wildly different useful properties
                    ↓
Abstraction:   Group — the minimal list of properties worth naming
                    ↓
Generalization: study every structure satisfying those properties at once,
                regardless of what its elements "really are"

or:

Problem:       we want to talk about "close" without a fixed idea of distance
                    ↓
Abstraction:   Metric space
                    ↓
Generalization: drop the requirement that "close" be measured by a number at all
                    ↓
Further abstraction: Topological space

Problem, then abstraction, then generalization. Mathematics doesn’t accumulate definitions arbitrarily; each one exists because a specific question couldn’t be answered without it, and each generalization exists because someone noticed the same answer working in a setting nobody had originally built it for. This site tries to always show the problem before the definition, and the definition before its consequences — never a definition presented as if it fell from the sky.

2.3 What this site is not trying to do

It is not going to re-teach you how to compute a derivative or row-reduce a matrix — you already can. It is not going to be proof-heavy: formal definitions appear because they clarify what’s actually being claimed, but full proofs are mostly omitted unless the proof itself is the insight. And it is not going to pretend mathematics is a strict hierarchy with pure logic at the bottom and machine learning at the top — the map already made the case that history, formal foundations, conceptual dependency, and pedagogical order are four different graphs, and this site is built to keep them that way.

What it will do is start at the most primitive layer that’s actually worth starting at — not “where mathematics historically began,” but where the conceptual dependency graph has its root — and build upward: logic, then sets, then the functions and numbers built from them, then the general idea of a mathematical structure, with two structures (groups and vector spaces) worked through in enough depth to show what “understanding a structure” actually looks like. Logic and proof is next.