Skip to content

Matrix equivalence

The two-sided change-of-basis equivalence B=Q⁻¹AP for same-sized rectangular matrices, classifying representations of one linear map under independent domain and codomain bases and determined solely by rank.

Version
v1 · 2026-09-28 · History
Domain-specific #
10613
Domain group
Formal Sciences
Origin domain
Mathematics
Subdomain
Linear Algebra → Mathematics

Core Idea

Matrix equivalence treats a matrix as coordinates for a linear map between two vector spaces. Since the input and output spaces may have different bases, re-expression acts independently on the columns and rows. This produces the two-sided invertible transformation B=Q⁻¹AP.

For a fixed matrix size over a field, rank is the complete invariant. Row and column operations reduce every matrix to an identity block padded by zeros. The simplicity is precisely what distinguishes equivalence from similarity or congruence, which preserve more specialized structure.

Structural Signature

Sig role-phrases:

  • Linear map V→W — Provides a transformation between possibly different vector spaces. It is carrier. Counterfactual: An endomorphism-only frame wrongly forces one basis change on both sides.
  • Domain basis change P — Re-expresses input coordinates by an invertible right action. It is transformation. Counterfactual: A singular column operation can change rank and leaves the equivalence group.
  • Codomain basis change Q — Re-expresses output coordinates by an independent invertible left action. It is transformation. Counterfactual: Requiring Q=P would collapse toward similarity and exclude rectangular cases.
  • Two-sided orbit — Groups matrices reachable under the GL(m)×GL(n) action. It is invariant. Counterfactual: One-sided row equivalence is strictly narrower.
  • Rank — Completely classifies the orbit for matrices of fixed size over a field. It is classifier. Counterfactual: Equal determinant or trace cannot classify rectangular equivalence.
  • Rank normal form — Places every class at a block matrix with an identity block and zeros. It is output. Counterfactual: The canonical form records rank, not the eigenstructure of an endomorphism.

What It Is Not

  • It is not equality of matrices entry by entry.
  • It is not matrix similarity.
  • It is not row equivalence alone.
  • It does not preserve eigenvalues, trace, characteristic polynomial, or a bilinear form in general.
  • Closest near-miss. Matrix similarity uses the linked transformation P⁻¹AP for one square endomorphism; matrix equivalence permits independent P and Q because domain and codomain bases are separate.

Scope of Application

  • Linear-map classification. Identifies coordinate presentations that differ only by bases in source and target.
  • Normal forms. Reduces a matrix to the rank block form for proof and computation.
  • Systems of linear equations. Uses row and column transformations while tracking which solution semantics they preserve.
  • Algebraic geometry and modules. Provides a basic orbit model later refined by equivalence over rings or parameter families.

Clarity

State matrix dimensions, scalar field, left and right transformation conventions, and whether the basis changes are independent and invertible. Use rank as the complete invariant only over the stated field and fixed size; over rings, additional invariants can appear.

Manages Complexity

The relation compresses all same-sized matrices into rank-indexed orbits under a product of general linear groups. It exposes how much structure is discarded by independent coordinate choices and tells an analyst when a stronger equivalence notion is necessary.

Abstract Reasoning

  1. Interpret the matrix as a map from a declared domain to codomain.
  2. Allow independent invertible basis changes on the two spaces.
  3. Compute rank or reduce by elementary row and column operations.
  4. Compare canonical rank forms to decide equivalence.
  5. Recover transformation matrices if a concrete change of coordinates is needed.
  6. Reject the relation when the problem requires similarity, congruence, or one-sided invariants.

Knowledge Transfer

The transferable cargo is orbit classification under independent invertible left and right actions. It transfers across coordinate representations of linear maps when field and dimensions are fixed; it stops where a shared basis, form, spectrum, or module structure must be preserved.

Examples

Canonical

A rank-k m×n matrix is reduced by invertible row and column operations to a block identity I_k with all remaining entries zero; every same-sized rank-k matrix reaches the same form.

Mapped back: size → m by n; operations → invertible two-sided; invariant → rank k; form → Ik plus zeros.

Applied / In Practice

Two square matrices can have equal rank and therefore be equivalent but different characteristic polynomials, so they need not be similar.

Mapped back: rank → equal; eigenstructure → different.

Applied / In Practice

Row-equivalent matrices are matrix-equivalent with P equal to identity, but a general column change can create matrix-equivalent matrices outside one row-equivalence class.

Mapped back: row action → subset; column action → additional.

Structural Tensions

T1 — Coordinate Freedom versus Retained Structure. Independent basis changes remove nearly all information except rank.

Diagnostic: Is rank the intended invariant, or does the application need eigenvalues, bilinear form, or row-space identity?

T2 — Canonical Simplicity versus Transformation Provenance. Normal form clarifies classification while hiding which domain and codomain bases produced it.

Diagnostic: Must the actual basis maps be retained for interpretation?

Structural–Framed Character

Matrix Equivalence is hybrid: structurally a two-sided group action and framed by linear-algebraic coordinate conventions.

Structural Core vs. Domain Accent

The core is one representation changed independently at input and output while a complete invariant remains. Linear algebra supplies vector spaces, fields, bases, rank, elementary operations, and distinctions from endomorphism and bilinear-form equivalences.

This entry is a kind of Mathematical Relation.

  • Approved root. The frozen graph preserves this specific two-sided equivalence without a forced parent.

  • Related — equivalence relation, matrix similarity, matrix congruence, row equivalence, rank, and canonical form. These supply generic form, nearest relations, invariant, and representative.

Relationships to Other Abstractions

Local relationship map for Matrix equivalenceParents appear above the current abstraction, mutual partners to the right, and children below. Node labels state whether each abstraction is prime or domain-specific; colors identify relation types.Matrix equivalenceDOMAINDomain-specific abstraction: Mathematical Relation — is a kind ofMathematicalRelationDOMAIN

Current abstraction Matrix equivalence Domain-specific

Parents (1) — more general patterns this builds on

  • Matrix equivalence is a kind of Mathematical Relation Domain-specific

    Matrix equivalence satisfies the defining boundary of Mathematical Relation: A mathematical relation is a formally specified subset of a Cartesian product, predicate over typed objects, equivalence or order structure, or equation constraining quantities, with arity, domain, parameters, and satisfaction conditions declared.

Hierarchy path (1) — routes to 1 parentless root

Neighborhood in Abstraction Space

Matrix equivalence sits in a crowded region of the domain-specific corpus (27th percentile for distinctiveness): several abstractions share nearly its structure, so a description that fits it tends to fit its neighbors too.

Family — Physical & Geometric Dynamical Quantities (29 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-10-08

Not to Be Confused With

  • Matrix Similarity. Tell: Similarity links both sides through one basis change and preserves endomorphism eigenstructure; equivalence uses independent bases and preserves only rank.
  • Matrix Congruence. Tell: Congruence uses a transpose-linked action appropriate to bilinear forms, not independent domain and codomain changes.
  • Row Equivalence. Tell: Row operations form a one-sided special case and preserve row-space information absent from full equivalence.
  • Equivalent Impedance Transforms. Tell: That engineering phrase concerns network behavior and is unrelated to the matrix group action.

References

  • Frozen Wikipedia discovery revision: https://en.wikipedia.org/wiki/Matrix_equivalence (revision 1317296789).
  • Preserved source candidate: https://hefferon.net/linearalgebra/

The frozen Wikipedia revision is discovery provenance. The retained source set was reviewed for identity, formal or operational relation, and scope. The encyclopedia's structural synthesis is bounded to those claims; a thin authority surface is recorded as a nonblocking source-strengthening repair rather than concealed.