Matrix Norm Calculator

Enter a matrix to compute its Frobenius norm, 1-norm, infinity norm, and 2-norm (spectral norm) — the four most common ways to measure how "large" a matrix is.

Quick Facts

Frobenius norm
‖A‖_F = √(Σ aᵢⱼ²)
Treats the matrix like a flattened vector and takes its Euclidean length.
1-norm
‖A‖₁ = max column sum
Largest sum of absolute values down any single column.
Infinity norm
‖A‖∞ = max row sum
Largest sum of absolute values across any single row.
2-norm (spectral)
‖A‖₂ = σ_max(A) = √(λ_max(AᵀA))
The matrix's largest singular value; found here by power iteration.

Your Results

Calculated
Frobenius Norm
-
‖A‖_F = √(Σ aᵢⱼ²)
1-Norm (Max Column Sum)
-
‖A‖₁ = max_j Σᵢ |aᵢⱼ|
Infinity Norm (Max Row Sum)
-
‖A‖∞ = max_i Σⱼ |aᵢⱼ|
2-Norm (Spectral Norm)
-
‖A‖₂ = σ_max(A), via power iteration

Ready

Enter matrix rows (comma or space separated), then press Calculate.

How Matrix Norms Are Calculated

A matrix norm is a single non-negative number that measures the "size" or magnitude of a matrix, generalizing the idea of absolute value from ordinary numbers to matrices. Different norms weight the entries differently, so the same matrix can have several different, equally valid norm values. This calculator computes the four norms used most often in linear algebra and numerical computing: the Frobenius norm, the 1-norm, the infinity norm, and the 2-norm (spectral norm).

Formula and method

For an m×n matrix A with entries aᵢⱼ, this calculator computes:

  • Frobenius norm: ‖A‖_F = √(Σᵢ Σⱼ |aᵢⱼ|²) — square every entry, add them all up, and take the square root, exactly like the Euclidean length of a vector formed by stacking the matrix's entries.
  • 1-norm: ‖A‖₁ = maxⱼ Σᵢ |aᵢⱼ| — sum the absolute values down each column, then take the largest column total.
  • Infinity norm: ‖A‖∞ = maxᵢ Σⱼ |aᵢⱼ| — sum the absolute values across each row, then take the largest row total.
  • 2-norm (spectral norm): ‖A‖₂ = σ_max(A) = √(λ_max(AᵀA)) — the largest singular value of A, equal to the square root of the largest eigenvalue of AᵀA. Because computing eigenvalues in closed form gets hard past 2×2 or 3×3, this calculator finds λ_max numerically with power iteration: it repeatedly multiplies a vector by AᵀA and renormalizes, and the result converges to the dominant eigenvalue within a few dozen steps for any well-conditioned matrix.

Common sources of error

  • Ragged rows: every row you type must have the same number of entries — a matrix cannot have three values in one row and two in the next.
  • Confusing entrywise and induced norms: the Frobenius norm weighs every entry equally, while the 1-, infinity-, and 2-norms are "induced" (operator) norms tied to how the matrix stretches vectors — they typically give different numbers even for the same matrix.
  • Row vs. column mix-up: it is easy to swap the 1-norm (column sums) with the infinity norm (row sums). A quick mnemonic: "1" is a single upright stroke like a column, "∞" lies on its side like a row.

Checking your result

Two inequalities make handy sanity checks: the spectral norm never exceeds the Frobenius norm (‖A‖₂ ≤ ‖A‖_F), and it also never exceeds the geometric mean of the 1-norm and infinity norm (‖A‖₂ ≤ √(‖A‖₁ · ‖A‖∞)). If your computed 2-norm is larger than either bound, recheck the matrix entries. For a symmetric matrix, the 2-norm equals the largest absolute eigenvalue directly, which is another quick way to cross-check small examples by hand.

Applications

Matrix norms bound how much a matrix can amplify a vector, since ‖Ax‖ ≤ ‖A‖ · ‖x‖ for every choice of vector norm and its matching induced matrix norm. That inequality underlies convergence analysis for iterative solvers, condition-number estimates (κ(A) = ‖A‖ · ‖A⁻¹‖) used to judge how sensitive a linear system is to rounding error, error bounds in numerical and perturbation analysis, and regularization terms such as Frobenius-norm weight decay in machine learning.

Frequently Asked Questions

What is the difference between the Frobenius norm and the 2-norm (spectral norm)?
The Frobenius norm treats the matrix like a flattened vector and takes the square root of the sum of the squares of every entry. The 2-norm (spectral norm) is the largest singular value of the matrix — the maximum factor by which it can stretch a unit vector. For any matrix, the 2-norm is always less than or equal to the Frobenius norm, with equality only for rank-1 matrices.
How do I compute the 1-norm and infinity norm of a matrix by hand?
For the 1-norm, add the absolute values down each column and take the largest column total. For the infinity norm, add the absolute values across each row and take the largest row total. For example, in the matrix [[2,-1],[-1,2]], both column sums and both row sums equal 3, so the 1-norm and infinity norm are both 3.
Does a matrix need to be square to have a norm?
No. The Frobenius norm, 1-norm, and infinity norm are defined for any m×n matrix, since they only require summing absolute values or squared values across rows or columns. The 2-norm (spectral norm) is also defined for rectangular matrices because it comes from the eigenvalues of AᵀA, which is always a square, symmetric matrix regardless of the shape of A.
How is the spectral norm (2-norm) actually computed?
The spectral norm equals the square root of the largest eigenvalue of AᵀA (equivalently, the largest singular value of A). This calculator finds that dominant eigenvalue numerically using power iteration: it repeatedly multiplies a vector by AᵀA and renormalizes until the result converges to the dominant eigenvector, which typically takes well under a hundred iterations for small matrices.

Related calculators