Riesz representation (Hilbert)

A linear functional is a rule that takes a vector and returns a number, linearly. In R^n every such rule is just dotting with some fixed vector — that vector is its gradient. The Riesz representation theorem says this stays exactly true in any Hilbert space, no matter how infinite-dimensional: every continuous linear functional is secretly an inner product with one fixed vector. There is nothing more general lurking.

Precisely: let H be a Hilbert space and phi a bounded (continuous) linear functional on H. Then there is a unique vector y in H such that phi(x) = <x, y> for all x, and moreover the norm of the functional equals ||y||. The map phi -> y is a bijection from the dual space onto H itself, conjugate-linear and norm-preserving.

Why it is so consequential: it identifies the dual of a Hilbert space with the space itself. Whereas a general Banach space and its dual can be very different objects, a Hilbert space is, up to a conjugation, its own dual — Hilbert spaces are reflexive in the strongest possible way. This self-duality is what makes Hilbert-space geometry so symmetric and so much easier than general Banach-space theory.

Why it powers the rest of operator theory: Riesz representation is exactly the tool that lets you define the adjoint T^* — you apply it to the functional x -> <T x, y> to manufacture the vector T^* y. It also underlies weak formulations of differential equations (Lax-Milgram) and the whole reproducing-kernel Hilbert space machinery used in modern machine learning.