Day 44 · 2026.08.05

Functional Analysis

When a "vector" has infinitely many coordinates, what is left of geometry?
"Good mathematicians see analogies between theorems; the best see analogies between analogies." — Stefan Banach

Functions as Vectors

Function Spaces · Moving analysis into geometry
Analysis
Intuition

The vector $(3,1,4)$ is three numbers in a row. Stretch it to a million coordinates, then replace the integer index by a real number $x$, and what you hold is a function $f(x)$: a function is a vector whose coordinates are laid out continuously. Addition and scaling are unchanged, so "the distance between two functions" and "the subspace spanned by a family of functions" suddenly mean something. That is the founding move of functional analysis — treat functions, signals, and solutions of equations as points in a space, and handle them geometrically.

But infinite dimensions are not finite dimensions enlarged. In finite dimensions closed and bounded implies compact; in infinite dimensions the unit ball is not compact — take an orthonormal sequence $e_n$, any two of them a fixed distance $\sqrt{2}$ apart, wandering forever with no convergent subsequence. The existence tool that carries all of finite-dimensional analysis fails here.

Formal definition

A Banach space is a complete normed linear space. The norm $\|\cdot\|$ assigns each vector a "length" satisfying the triangle inequality; complete means every Cauchy sequence converges inside the space — the space has no holes. Canonical example $L^2[0,1]$: all square-integrable functions, $\|f\|_2=\left(\int_0^1|f|^2\right)^{1/2}$.

Completeness is not a technicality: constructing an object almost always means building a sequence of ever-better approximations and passing to the limit, and completeness guarantees the limit is still in the space — just as the rationals are incomplete, which is exactly why $\sqrt{2}$ is not among them.

Why it is beautiful

Completeness turns existence into something you can build. The Banach fixed-point theorem: on a complete space every contraction has a unique fixed point, and iterating from any starting point converges to it. So existence and uniqueness for ODEs (Picard–Lindelöf) requires no formula — rewrite the equation as an integral operator, check that it contracts, and the solution exists. Existence follows from the structure of the space rather than from finding the answer: ask first which space the object lives in, then what it looks like.

Applications

Sobolev spaces are where "weak solutions" of PDEs live: first guarantee a solution exists in a large enough space, then prove after the fact that it is smooth; the finite element method turns exactly this framework into an engineering algorithm. In machine learning, the reproducing kernel Hilbert space lets kernel methods run regression in an infinite-dimensional feature space without ever writing the features down.

In one line · A question
Function spaces turn "solving an equation" into "locating a point in a geometric space."
Question: non-compactness of the unit ball kills "extract a convergent subsequence," and the compensation is to invent a weaker notion of convergence that buys compactness back. In engineering problems you know well, is there an isomorphic move — loosening a criterion to recover a guarantee?

Hilbert Space

Hilbert Space · The inner product gives angles back
Geometry
Intuition

A Banach space has length but no angle — you can say how far apart two functions are, but not what angle they make. Add an inner product $\langle f,g\rangle=\int f\bar g$ and "perpendicular" immediately means something: orthogonal is inner product zero, i.e. neither function contains any of the other.

Orthogonality brings projection: project a function onto a closed subspace and you get the point of that subspace nearest to it, with the residual perpendicular to the whole subspace. This picture, drawable in three dimensions, holds verbatim in infinite dimensions — and it is simultaneously the theorem behind least squares, Fourier series, and conditional expectation.

closed subspace M f Pf f − Pf ⊥ M Best approximation = dropping a perpendicular
Formal definition

Projection theorem: if $M$ is a closed subspace of a Hilbert space, every $f$ decomposes uniquely as $f=Pf+(f-Pf)$ with $Pf\in M$ and $f-Pf\perp M$; $Pf$ is the unique $g\in M$ minimising $\|f-g\|$. If $\{e_n\}$ is an orthonormal basis of $M$:

$$Pf=\sum_n \langle f,e_n\rangle\, e_n,\qquad \|f\|^2=\sum_n |\langle f,e_n\rangle|^2$$

Reading the symbols: $\langle f,e_n\rangle$ is the length of $f$'s projection onto the axis $e_n$, and the sum reassembles the components. The right-hand identity is Parseval — the Pythagorean theorem in infinite dimensions: squared length equals the sum of squared coordinates. Take $e_n(x)=e^{2\pi i n x}$ and the expansion is the Fourier series; Fourier is not a trick, it is a change of orthonormal basis.

Why it is beautiful

Three apparently unrelated facts are one sentence. The residual of a linear regression is orthogonal to the column space of the design matrix; decomposing a signal into frequencies is projection onto one orthonormal basis; the conditional expectation $E[X\mid\mathcal G]$ is the projection of $X$ onto the closed subspace of $\mathcal G$-measurable functions, and the tower property $E[E[X\mid\mathcal G]]=E[X]$ is nothing but "projecting twice equals projecting once." This is not a rhetorical analogy — they share one theorem and one proof.

Applications

The normal equations of least squares come straight from projection; each Kalman filter update projects a new observation onto the subspace spanned by existing information; for PCA and every "keep the first $k$ components" compression, Parseval gives the error exactly — the sum of the squares of the discarded coefficients — so the trade-off is computed, not guessed.

In one line · A question
The inner product reduces "best approximation," a problem of analysis, to "dropping a perpendicular," a gesture of geometry.
Question: why is "residual orthogonal to the subspace" equivalent to "error minimised"? If the residual still had a component inside the subspace, could you say concretely how to improve the current solution?

Operators and Spectrum

Operators and Spectrum · The fate of eigenvalues in infinite dimensions
Operator Theory
Intuition

In finite dimensions matrices act on vectors; in infinite dimensions operators act on functions — differentiation $Df=f'$, multiplication by $x$, convolution are all operators. The eigenvalue question carries over unchanged: which functions are merely stretched by a factor when acted upon? The solutions of $-f''=\lambda f$ are sines, so vibrational modes are the eigenvectors of a differential operator — which is why strings, drumheads, and quantum wells share one piece of mathematics.

But infinite dimensions grow a new phenomenon: an operator can have no eigenvectors at all and still fail to be invertible. So the definition is relaxed to the spectrum $\sigma(T)=\{\lambda: T-\lambda I\ \text{not invertible}\}$. The position operator $(Xf)(x)=xf(x)$ on $L^2[0,1]$ has spectrum all of $[0,1]$ yet not a single eigenfunction — $xf=\lambda f$ forces $f$ to vanish almost everywhere. Continuous spectrum is not a generalisation of eigenvalues but a replacement for them.

Finite-dimensional matrix: finitely many eigenvalues Self-adjoint operator: discrete plus continuous spectrum bound states: discrete levels scattering: continuous energy
Formal definition

Self-adjoint means $\langle Tf,g\rangle=\langle f,Tg\rangle$ — moving the operator from one side of the inner product to the other changes nothing, the infinite-dimensional version of a symmetric matrix $A=A^{\top}$. The spectral theorem: every self-adjoint operator can be written

$$T=\int_{\sigma(T)} \lambda\, dE_\lambda$$

where $E_\lambda$ is a projection-valued measure assigning a projection to each piece of the spectrum. How to read it: finite-dimensional diagonalisation is $T=\sum\lambda_i P_i$, a sum of projections onto eigendirections weighted by eigenvalues; infinite dimensions merely replace the sum by an integral over the spectrum, because the spectrum may form a continuum with no countable set of eigendirections.

Why it is beautiful

The spectral theorem states the philosophy of infinite-dimensional analysis in one line: a single algebraic property (self-adjointness) forces an entire geometric structure (diagonalisability), even when eigenvectors do not exist. A second depth comes from unbounded operators, which force the notion of a domain — an operator is not just a formula but also the set of functions it acts on. The same $-d^2/dx^2$ with different boundary conditions is a different operator with a different spectrum: a particle in a box has discrete levels, a free particle a continuum, and the difference lies in the domain, not the formula.

Applications

Spectral methods for PDEs: expand in an eigenfunction basis and the differential operator becomes diagonal multiplication, so a PDE degenerates into a family of independent ODEs — in the heat equation's solution $\sum c_n e^{-\lambda_n t}\phi_n$ one sees at a glance that high frequencies decay fastest. The spectrum of the graph Laplacian underpins spectral clustering and the filter design of graph neural networks. The Koopman operator rewrites a nonlinear dynamical system as a linear operator in infinite dimensions, whose spectrum drives data-driven prediction.

In one line · A question
The spectrum is an operator's fingerprint — it may be continuous, it may have no eigenvectors, and it still determines the operator's behaviour completely.
Question: Kac asked whether one can hear the shape of a drum. The answer is no — isospectral drums of different shapes exist. So which part of the information does this fingerprint lose?

The Mathematics of Quantum Mechanics

Von Neumann's Axiomatisation · Physics read off a spectrum
Mathematical Physics
Intuition

In the 1920s quantum mechanics had two mutually suspicious formalisms — Heisenberg's matrix mechanics and Schrödinger's wave mechanics — which produced identical spectral lines for reasons nobody could state. Von Neumann's 1932 answer: they are coordinate representations of the same Hilbert space in different orthonormal bases, two ways of writing one vector.

What deserves memorising is his dictionary, translating physical notions line by line into operator theory:

  • state = a unit vector $\psi$ in a Hilbert space
  • observable = a self-adjoint operator $A$ (self-adjoint ⟹ real spectrum ⟹ real measured values)
  • possible outcomes = the spectrum of $A$; probability of an outcome = the squared length of the spectral projection $\|E_\lambda\psi\|^2$
  • expectation = $\langle\psi,A\psi\rangle$; time evolution = the unitary group $U_t=e^{-iHt}$ generated by a self-adjoint $H$
Formal definition

The general (Robertson) uncertainty relation:

$$\sigma_A\,\sigma_B \ \ge\ \tfrac{1}{2}\left|\langle\psi,[A,B]\psi\rangle\right|$$

Reading the symbols: $\sigma_A$ is the standard deviation of $A$ in the state $\psi$, i.e. how spread out the outcomes are; $[A,B]=AB-BA$ is the commutator, measuring how much the order of two operations matters, and vanishing when it does not. The whole inequality says: the less two operations commute, the less their values can be pinned down at once. Substituting $[X,P]=i\hbar$ gives $\sigma_X\sigma_P\ge\hbar/2$.

Why it is beautiful

The uncertainty principle is usually told as the experimental fact that measurement disturbs a particle — yet the derivation above uses only the Cauchy–Schwarz inequality and no physical assumption: any two non-commuting self-adjoint operators satisfy this bound automatically. It is a theorem of mathematics disguised as a law of physics. Likewise "quantisation" literally means discreteness of a spectrum — bound states have discrete spectrum, hence energy levels one by one; free particles have continuous spectrum, hence energy as a continuum. The most counter-intuitive part of physics is a direct reading of an operator's spectral structure.

Applications

Quantum computing runs entirely on this dictionary: gates are unitary operators, measurement is projection, and the whole formalism is spectral theory on a finite-dimensional Hilbert space. The variational method of quantum chemistry restricts the ground-state problem to a finite-dimensional subspace and minimises a Rayleigh quotient — the projection theorem in disguise. The classical counterpart of Stone's theorem (unitary groups ↔ self-adjoint generators) is the pairing of Markov semigroups with their generators.

In one line · A question
Quantum mechanics does not "use" mathematics; its axioms are the spectral theory of self-adjoint operators.
Question: if an observable's operator has continuous spectrum, the probability of measuring any exact value is strictly zero. What, then, is the number read off in the laboratory — does "measurement" in fact refer to a spectral interval rather than a spectral point?

Going Deeper

Why is $L^2$ special among all the $L^p$ spaces?
Only for $p=2$ is the norm induced by an inner product, so only $L^2$ is a Hilbert space and only it has angles, orthonormal bases, and a projection theorem. This explains a statistical choice often mistaken for mere convention: least squares dominates not because squared error is more "correct" than absolute error, but because it turns estimation into projection and therefore admits a closed form. Switching to $L^1$ (LASSO, quantile regression) forfeits the projection structure in exchange for sparsity and robustness, at the price of iterative solvers. Choosing a loss function is choosing a geometry.
Is the universal approximation theorem a result of functional analysis?
Yes, but it gives far less than people assume. It asserts that single-hidden-layer networks are dense in a space of continuous functions — only that "there exists a network arbitrarily close to the target." It offers no useful bound on the width required, is entirely silent on how to find those weights, and provides no generalisation guarantee. The real difficulties of deep learning are optimisation and generalisation, and density touches neither. Mistaking existence for reachability is the most common misreading of a theorem.
What is the real link between operator spectra and distributed systems?
A Markov chain's transition matrix is an operator, and its spectral gap (the distance between the largest and second-largest eigenvalue) directly controls mixing time. The conclusion transfers verbatim to gossip protocols — the second-smallest eigenvalue of the graph Laplacian sets how many rounds information needs to reach the whole network, so choosing a topology is choosing a spectrum. The same quantity explains over-smoothing in graph neural networks: repeated aggregation is repeated application of one operator, non-dominant spectral components decay as powers of their eigenvalues, and node representations collapse onto the leading eigenvector. "How fast does it converge" is usually a question about a spectrum.