Linear matrix inequality¶

A convex constraint requiring an affine combination of symmetric or Hermitian matrices to be positive semidefinite.

Version: v2 · 2026-09-19 · History
Domain-specific #: 5343
Domain group: Formal Sciences
Origin domain: Mathematics
Subdomain: Convex Optimization → Mathematics

Core Idea¶

An LMI has the form A0 plus sum y_i A_i semidefinite greater than or equal to zero; its feasible set is a spectrahedron and semidefinite programming optimizes linear objectives over intersections of such constraints. Decision variables linearly weight fixed matrices, eigenvalue nonnegativity enforces all quadratic-direction inequalities simultaneously and convexity permits global optimization and dual certificates. The abstraction is therefore identified by a declared carrier, a transformation or constraint over that carrier, and an invariant that tells an analyst whether the named structure is genuinely present.

How would you explain it like I'm…

No faithful explanation at this level. A kid-level version turns the condition into 'this number, or every entry, is at least zero.' That misses that semidefiniteness requires the whole matrix to be nonnegative in every quadratic direction at once.

Never-Negative Grid Rule

A matrix is a square grid of numbers. In a Linear matrix inequality you have some adjustable numbers, and each one turns up or down how much of a fixed grid gets added in. The rule is that the final grid must pass a special 'never negative' test. This test does not just check whether each entry is positive; it checks the grid's behavior in every direction at once. All the settings that pass form a single blob with no dents, so computers can reliably find the best settings inside it.

Semidefinite Matrix Constraint

A Linear matrix inequality says A₀ + y₁A₁ + … + yₙAₙ ⪰ 0, where the Aᵢ are fixed symmetric matrices and the yᵢ are unknowns. The symbol ⪰ 0 means positive semidefinite: all eigenvalues are nonnegative, which is the same as vᵀMv ≥ 0 for every vector v. So one LMI packs infinitely many ordinary inequalities into a single statement. The set of y's that satisfy it is convex (any segment between two solutions stays inside) and is called a spectrahedron. Semidefinite programming optimizes a linear objective over such sets, and convexity means you can find the true global best and prove it with a dual certificate.

A Linear matrix inequality has the form F(y) = A₀ + Σ yᵢAᵢ ⪰ 0, with Aᵢ fixed real symmetric (or complex Hermitian) matrices and an affine dependence on the decision variables y. Semidefiniteness means all eigenvalues of F(y) are nonnegative, equivalently vᵀF(y)v ≥ 0 for all v, enforcing every quadratic-direction inequality simultaneously. The feasible set is a spectrahedron, which is convex. Semidefinite programming minimizes linear objectives over intersections of LMIs, and convexity gives global optimality and duality-based certificates of optimality or infeasibility. One must specify whether the inequality is strict (≻) or nonstrict (⪰), and nonlinear-looking constraints are often converted into LMIs via Schur complements, which should be stated explicitly. LMIs are a specific constraint class within convex optimization, not a synonym for it.

Scope of Application¶

Linear matrix inequality belongs to convex optimization and is useful where the analyst can specify the typed convex optimization carrier, defining objects and relations, parameters, conventions, evidence, boundary cases, and comparison targets, then evaluate the real symmetric or complex Hermitian matrices, affine variable map, semidefinite ordering, feasible variables, strict or nonstrict convention and any Schur-complement equivalence are explicit. The scope is broad within that domain but bounded by the need for the real symmetric or complex Hermitian matrices, affine variable map, semidefinite ordering, feasible variables, strict or nonstrict convention and any Schur-complement equivalence are explicit. Conceptual optimization identity only; safety-critical control applications require validated models and qualified engineering.

Clarity¶

The abstraction clarifies a crowded vocabulary by making the real symmetric or complex Hermitian matrices, affine variable map, semidefinite ordering, feasible variables, strict or nonstrict convention and any Schur-complement equivalence are explicit the center of the account. A claim should name the carrier, the governing operation or relation, the applicable assumptions, and the recognition test. A bare label is insufficient because the name Linear matrix inequality can be used for a formal identity, an implementation, or a neighboring result unless carrier and convention are stated.

Manages Complexity¶

Without the abstraction, an analyst must reason directly over many local details: the carrier roles, admissibility assumptions, competing conventions, derived invariants, boundary cases, and proof or validation obligations specific to Linear matrix inequality. Linear matrix inequality compresses them into the roles in the structural signature. That compression permits comparison across instances without erasing the variables that determine validity. It also exposes which details may be varied safely and which are constitutive.

Abstract Reasoning¶

Identify the carrier. State what the elements, states, objects, or observations are: the typed convex optimization carrier, defining objects and relations, parameters, conventions, evidence, boundary cases, and comparison targets. Reject examples whose alleged carrier belongs to a different problem.
Lock the constitutive rule. Express the real symmetric or complex Hermitian matrices, affine variable map, semidefinite ordering, feasible variables, strict or nonstrict convention and any Schur-complement equivalence are explicit independently of one notation or implementation.

Knowledge Transfer¶

Knowledge transfers strongly among subfields of convex optimization because they reuse the typed convex optimization carrier, defining objects and relations, parameters, conventions, evidence, boundary cases, and comparison targets, Decision variables linearly weight fixed matrices, eigenvalue nonnegativity enforces all quadratic-direction inequalities simultaneously and convexity permits global optimization and dual certificates., and type the carrier, state every parameter and convention in the definition, test that the real symmetric or complex Hermitian matrices, affine variable map, semidefinite ordering, feasible variables, strict or nonstrict convention and any Schur-complement equivalence are explicit, compare the nearest accepted identity, and report counterexamples, uncertainty, and limiting cases.

Relationships to Other Abstractions¶

Current abstraction Linear matrix inequality Domain-specific

Parents (1) — outgoing typed relations

Linear matrix inequality is a kind of Constraint Prime

It remains its own entry because its identity is fixed by the real symmetric or complex Hermitian matrices, affine variable map, semidefinite ordering, feasible variables, strict or nonstrict convention and any Schur-complement equivalence.

Hierarchy path (1) — routes to 1 parentless root

Linear matrix inequality → is a kind of Constraint

Neighborhood in Abstraction Space¶

Linear matrix inequality sits in a crowded region of the domain-specific corpus (17^th percentile for distinctiveness): several abstractions share nearly its structure, so a description that fits it tends to fit its neighbors too.

Family — Convex Analysis & Optimization Structures (8 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-10-25