John Ellipsoid¶
The unique maximum-volume ellipsoid inscribed in a full-dimensional compact convex body, providing a centered geometric approximation whose bound depends on symmetry.
Core Idea¶
For a full-dimensional compact convex body \(K\subset\mathbb R^n\), the John ellipsoid \(E(K)\) is its unique maximum-\(n\)-volume inscribed ellipsoid. The candidate is contained in \(K\), and its volume is compared with every other ellipsoid contained there. The minimum-volume enclosing Löwner ellipsoid is a different extremal object, not an alternative spelling of this one.[ref-99cc9658e8e0][ref-bb3309838db4]
If \(c\) is John's center, \(E\subseteq K\subseteq c+n(E-c)\). Only when \(K\) is centrally symmetric does the stronger centered factor \(\sqrt n\) apply. These are derived approximation guarantees: neither factor defines the optimizer, and dilation must be about \(c\), not an arbitrary origin.[^ref-99cc9658e8e0]
Scope of Application¶
For a regular simplex, the unique maximal inner ellipsoid is centered at the simplex barycenter, and the general factor \(n\) is sharp; the planar equilateral triangle gives the inscribed-circle instance. The simplex is not centrally symmetric, so applying the \(\sqrt n\) norm-ball bound would be an error.[^ref-99cc9658e8e0]
For the infinity-norm unit cube \([-1,1]^n\), the Euclidean unit ball is John. The cube is centrally symmetric, its corners meet the \(\sqrt n\) dilation, and the arrangement underlies the \(\sqrt n\) Banach–Mazur comparison with Euclidean space. Unlike the simplex, the stronger bound's symmetry premise is met.[ref-99cc9658e8e0][ref-71fc545bc662]
Clarity¶
To identify a John ellipsoid, name the container \(K\), verify that the ellipsoid lies inside it, and establish global maximal \(n\)-volume among inscribed ellipsoids. An arbitrary inner ellipse, an enclosing Löwner ellipse, ellipsoid packing and an algorithm for finding a shape answer different questions. Contact-point conditions or affine John position can characterize or analyze the optimum but do not replace its definition.[ref-99cc9658e8e0][ref-bb3309838db4]
Manages Complexity¶
A complicated convex shape can be compared with one uniquely determined inner affine ball. The same ball gives dimension-controlled outer containment, with the sharper factor restricted to symmetric bodies. Affine normalization makes a symmetric norm ball's John ellipsoid the Euclidean unit ball and exposes a weighted contact-point condition; conclusions about original coordinate lengths require mapping back.[ref-99cc9658e8e0][ref-bb3309838db4]
Abstract Reasoning¶
First check compactness, convexity and nonempty interior in the stated \(n\) dimensions. Then check inner containment and volume optimality. Use \(c+n(E-c)\) for the general body; claim \(c+\sqrt n(E-c)\) only after establishing central symmetry. The contrast between simplex and cube shows why a bound from normed-space notes cannot simply be transferred to all convex bodies.[ref-99cc9658e8e0][ref-71fc545bc662]
Knowledge Transfer¶
Across the simplex and cube, the roles remain container, feasible inner ellipsoids, volume maximizer, unique \(E(K)\) and center-based enclosure. What changes is the symmetry premise and thus the sharp factor. The staged DAG proposes Convex body as a strict genus of the resulting \(E\). Portable constrained optimization may suggest a future-prime question, but the John object itself requires ellipsoids, Euclidean volume and convex containment.[ref-99cc9658e8e0][ref-bb3309838db4]
[^ref-99cc9658e8e0]: Ralph Howard, “The John Ellipsoid Theorem”, University of South Carolina original proof notes (1997), Theorems 1–3, Remarks 1.1 and 3.1–3.2, §4.1 (PDF pp.1–10), directly inspected 2026-10-01. [^ref-bb3309838db4]: Joel A. Tropp, Lectures on Convex Geometry, original Caltech lecture notes (2018), Lecture 14 §§14.4–14.6, printed pp.120–127 (PDF pp.126–133), directly inspected 2026-10-01; this lecture's direct proof treats norm balls. [^ref-71fc545bc662]: Jonathan Kelner and Chaithanya Bandi, “Lecture 11: John Ellipsoid”, MIT 18.409 original lecture notes (2009), §§5–6 and Theorems 7–8 (PDF pp.6–8), directly inspected 2026-10-01.
Relationships to Other Abstractions¶
Current abstraction John Ellipsoid Domain-specific
Parents (1) — more general patterns this builds on
-
John Ellipsoid is a kind of Convex body Domain-specific
A nondegenerate John ellipsoid is a convex body with an additional relative extremal property.
Hierarchy path (1) — routes to 1 parentless root
- John Ellipsoid → Convex body → Convexity → Optimization
Neighborhood in Abstraction Space¶
John Ellipsoid sits in a sparse region of the domain-specific corpus (83rd percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.
Family — Convex Geometry & Measure Constructions (12 abstractions)
Nearest neighbors
- Milman's reverse Brunn–Minkowski inequality — 0.83
- Ellipsoid packing — 0.82
- Information projection — 0.82
- Feasible Region — 0.82
- Equilateral Dimension — 0.82
Computed from structural-signature embeddings · 2026-10-08