Skip to content

Nonnegative Matrix

A real matrix constrained entrywise to the nonnegative orthant, forming a convex cone closed under addition and multiplication and linking directed-graph structure to Perron–Frobenius spectral behavior.

Version
v2 · 2026-09-06 · History
Domain-specific #
2389
Origin domain
mathematics
Subdomain
matrix theory
Aliases
Entrywise nonnegative matrix, Non-negative matrix

Core Idea

A real matrix \(A=(a_{ij})\) is nonnegative, written \(A\ge0\), when

\[ a_{ij}\ge0\qquad\text{for every }i,j. \]

The order is entrywise. It places \(A\) in the closed nonnegative orthant of matrix space, makes the set of fixed-size nonnegative matrices a convex cone, and ensures that addition, nonnegative scaling, and conformable multiplication preserve nonnegativity.[1]

For square matrices, the constraint connects algebra to a directed weighted graph and enables Perron–Frobenius conclusions: the spectral radius is an eigenvalue with a nonnegative eigenvector, while irreducibility and primitivity strengthen positivity, uniqueness, and asymptotic behavior.[2]

The recognition invariant is real matrix + entrywise partial order + every entry nonnegative + cone-preserving matrix operations.

Structural Signature

  • A finite rectangular real matrix.
  • Entrywise comparison with the zero matrix.
  • Every entry in \([0,\infty)\).
  • Boundary points allowed through zero entries.
  • Strict positivity distinguished by every entry being greater than zero.
  • Closure under addition and multiplication by nonnegative scalars.
  • Closure under conformable matrix multiplication.
  • A convex polyhedral cone for each fixed shape.
  • Support pattern represented by a bipartite or directed graph.
  • Reducibility of a square matrix read from graph connectivity.
  • Spectral radius paired with a nonnegative eigenvector.
  • Stronger Perron conclusions gated by irreducibility or primitivity.

What It Is Not

Entrywise nonnegative does not mean positive semidefinite. Positive semidefiniteness applies to a symmetric quadratic form \(x^TAx\ge0\) and permits negative entries. A matrix satisfying both conditions is often called doubly nonnegative.[3]

A nonnegative matrix need not be invertible, symmetric, stochastic, irreducible, or positive. Even when invertible, its inverse is generally not nonnegative; inverse positivity is a separate restrictive property.

Scope of Application

Nonnegative matrices encode Markov transition weights, Leontief input–output systems, population projection, graph adjacency, consensus and positive dynamical systems, ranking, queueing, and network flows. Stochastic and doubly stochastic matrices add normalization constraints. M-matrices and inverse-positive matrices connect order preservation with stability and linear systems. Nonnegative matrix factorization seeks nonnegative factors of nonnegative data but is a separate approximation problem.[4]

Clarity

State that the inequality is entrywise and whether the matrix is real. Declare square versus rectangular shape and distinguish nonnegative \(A\ge0\) from positive \(A>0\). Spectral claims requiring irreducibility or primitivity must not be attributed to every nonnegative matrix.

Manages Complexity

One local sign constraint produces global monotonicity: multiplying a nonnegative vector cannot create negative components, paths in powers retain nonnegative weights, and cone methods replace unrestricted cancellation arguments. The support graph separates combinatorial zeros from numerical magnitudes.

Abstract Reasoning

  1. Verify every entry against zero under the real entrywise order.
  2. Record the zero-support pattern and matrix shape.
  3. For square matrices, construct the associated directed graph.
  4. Test irreducibility from strong connectivity.
  5. Compute or bound the spectral radius.
  6. Apply only the Perron–Frobenius conclusion licensed by reducibility and periodicity.
  7. Distinguish entrywise, Loewner, and inverse-positive orders.
  8. Preserve the cone explicitly under factorization, iteration, or optimization.

Knowledge Transfer

The portable pattern is constraining coordinates to an order cone so admissible operations preserve positivity and cancellation is controlled. The proposed immediate parent is Matrix.

Examples

Adjacency matrix. A directed graph's zero-one adjacency matrix is nonnegative; powers count walks and the support reveals reducibility.

Transition matrix. A row-stochastic matrix is nonnegative with each row summing to one, adding probabilistic normalization.

Not semidefinite. \(\begin{pmatrix}0&1\\1&0\end{pmatrix}\) is entrywise nonnegative but has eigenvalues \(1\) and \(-1\), so it is not positive semidefinite.

Structural Tensions

  • Entrywise order versus spectral order.
  • Zeros preserving sparsity versus strict positivity strengthening theorems.
  • Cone closure versus inverse nonclosure.
  • Support graph versus entry magnitudes.
  • Reducible blocks versus global Perron behavior.
  • Exact nonnegativity versus noisy estimated entries.

Structural–Framed Character

Coordinatewise cone restriction and closure are structural. Matrix entries, spectra, graph support, and Perron–Frobenius theory are linear-algebra frame.

Structural Core vs. Domain Accent

The portable core is an array constrained to an operation-preserving positive cone. The constitutive accent is the entrywise order on real matrices and its spectral-combinatorial consequences.

Matrix is the proposed immediate parent. Constraint, Cone, Closure, Order, Network, Factorization, and Iteration are related.

The prospective queue contains one strict edge to domain_specific:matrix. No live DAG mutation is authorized.

Relationships to Other Abstractions

Local relationship map for Nonnegative MatrixParents 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.Nonnegative MatrixDOMAINDomain-specific abstraction: Matrix — is a kind ofMatrixDOMAIN

Current abstraction Nonnegative Matrix Domain-specific

Parents (1) — more general patterns this builds on

  • Nonnegative Matrix is a kind of Matrix Domain-specific

    Matrix is the proposed immediate parent.

Hierarchy paths (5) — routes to 5 parentless roots

Neighborhood in Abstraction Space

Nonnegative Matrix sits in a sparse region of the domain-specific corpus (86th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Unclustered & Miscellaneous (1565 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-09-08

Not to Be Confused With

  • Positive-semidefinite matrix.
  • Positive matrix with all entries strictly positive.
  • Stochastic matrix.
  • M-matrix.
  • Nonnegative matrix factorization.
  • Matrix whose eigenvalues are all nonnegative.

References

[1] Abraham Berman and Robert J. Plemmons, Nonnegative Matrices in the Mathematical Sciences, SIAM Classics in Applied Mathematics 9 (1994), doi:10.1137/1.9781611971262. registry ↩

[2] Eugene Seneta, Non-negative Matrices and Markov Chains, 2nd ed. (Springer, 1981). registry ↩

[3] Roger A. Horn and Charles R. Johnson, Matrix Analysis, 2nd ed. (Cambridge University Press, 2013), chapters on Hermitian order and nonnegative matrices. registry ↩

[4] Richard S. Varga, Matrix Iterative Analysis, 2nd ed. (Springer, 2000), treatment of regular splittings, M-matrices, and nonnegative iteration matrices. registry ↩