Condition Number Calculator - Matrix Sensitivity Solver

Condition Number Calculator evaluates how much a 2x2 matrix A amplifies small input changes, by returning the condition number κ in the 1, 2, infinity, and Frobenius norms. Use it to decide whether a linear system Ax = b is safe to solve in floating-point arithmetic, whether a covariance matrix is reliable for sampling, or whether a regression design matrix X will give stable normal equations.

• • Sensitivity of linear systems: Run a small perturbation through Ax = b and read the condition number to see whether floating-point arithmetic can recover the true solution or whether the answer is dominated by roundoff.
• • Linear regression diagnosis: Compute κ for the X^T X matrix from ordinary least squares to decide whether the regression coefficients are stable, especially when the predictors are highly collinear.
• • Covariance and kernel checks: Compute κ for a 2x2 covariance matrix before plugging it into a multivariate Gaussian sampler or a kernel method, so an ill-conditioned input is caught before it propagates.
• • Singular matrix detection: Use the determinant line to spot a singular A where the inverse does not exist; the calculator then reports κ = Infinity in every norm and the system Ax = b has either no solution or infinitely many.

The condition number is one of the few numbers that links the abstract definition of invertibility to the practical question of how many digits of the answer are trustworthy.

When the matrix is symmetric positive definite, the Cholesky Decomposition Calculator produces the same L factor that LAPACK's POCON routine uses to estimate the 1-norm condition number, so the two tools pair naturally for SPD inputs.

The Condition Number Calculator builds the inverse of A from the adjugate-divided-by-determinant formula, computes the four norms of A and the four norms of A⁻¹, and reports the products as κ₁, κ∞, κF, and κ₂. The 2-norm is the largest singular value of A, obtained as the square root of the larger eigenvalue of A^T A.

• A: The 2x2 input matrix [[a₁₁, a₁₂], [a₂₁, a₂₂]]. The calculator reads all four entries from the form.
• det(A): Determinant of A, equal to a₁₁·a₂₂ − a₁₂·a₂₁. When |det(A)| drops below 1e-10 the calculator treats A as singular and skips the inverse step.
• A⁻¹: The matrix inverse, equal to the adjugate [[a₂₂, -a₁₂], [-a₂₁, a₁₁]] divided by the determinant. The four norms of A⁻¹ are the building blocks of the condition number.
• κ_n(A): The condition number in norm n, equal to the norm of A times the norm of A⁻¹. The calculator reports κ₁, κ∞, κF, and κ₂.

The 1-norm is the maximum absolute column sum, the infinity-norm is the maximum absolute row sum, the Frobenius norm is the square root of the sum of squared entries, and the spectral 2-norm comes from A^T A. The 2-norm value drives the interpretation banner because it is the singular-value ratio textbooks treat as canonical.

According to Wikipedia Condition number, the condition number of a square invertible matrix A in a given norm is the product of the norm of A and the norm of its inverse, and in the 2-norm it equals the ratio of the largest singular value of A to the smallest.

The natural next step after κ is to fit the regression itself, and the Linear Regression Calculator consumes the X^T X matrix whose condition number you just computed.

Four ideas show up in every condition number calculation.

These four ideas are the conceptual vocabulary that makes the rest of the page work. Knowing which norm a downstream algorithm uses turns the four κ values from a number salad into a decision tool.

The characteristic polynomial of a 2x2 matrix expands to λ² − tr(A)·λ + det(A) = 0, and the Characteristic Polynomial Calculator produces the eigenvalues that feed the κ₂ ratio on any symmetric input.

Type the four entries of A into the form. The calculator detects singular inputs, computes the four condition numbers, and renders the verdict banner above the breakdown table.

1. 1 Enter the four matrix entries: Type a₁₁, a₁₂, a₂₁, and a₂₂ into the top row and bottom row. The placeholder values are the standard SPD example [[2, 1], [1, 3]].
2. 2 Read the verdict banner: The black banner at the top of the result panel reports well-conditioned, moderately conditioned, ill-conditioned, severely ill-conditioned, or singular, based on κ₂.
3. 3 Inspect det(A) first: If det(A) is zero, the calculator marks the matrix as singular and reports κ = Infinity in every norm. Treat the linear system as not solvable in the usual sense.
4. 4 Compare the four κ values: When the four κ values disagree, the larger one reflects a norm the downstream algorithm does not use, and the smallest one is the one most relevant to the final answer.
5. 5 Use κ₂ as the canonical diagnostic: For floating-point arithmetic, κ₂ < 10 is comfortable, κ₂ around 1e3 is the practical ill-conditioning threshold, and κ₂ above 1e6 means several digits of the answer are unreliable.
6. 6 Rescale when κ₂ is large: If κ₂ explodes, rescale the columns of A so the largest absolute entries are within an order of magnitude. The κ values drop together and the system becomes easier to solve.

For A = [[1, 1], [1, 1.0001]], det = 0.0001, κ₂ ≈ 40001, and the calculator flags the matrix as severely ill-conditioned. A small regularization epsilon added to a₂₂ recovers a usable condition number.

When κ₂ is large, the corresponding covariance matrix should not be used to draw correlated normal samples, and the Normal Distribution Calculator shows the uncorrelated standard-normal side of the same workflow.

The main benefit is replacing a hand-written norm routine with a tool that returns the four condition numbers, the underlying matrix norms, and a verbal verdict in a single panel.

• • Catches ill-conditioned inputs early: The interpretation banner surfaces a 1e3 κ₂ value before the user feeds A into a downstream solve, so a near-singular matrix never silently corrupts the answer.
• • Reports κ in the four standard norms: The calculator returns κ₁, κ∞, κF, and κ₂ in the same panel, so a user who needs the 2-norm value and a user who needs the 1-norm or infinity-norm estimate can both stop here.
• • Pairs with linear regression workflows: Plug the X^T X matrix from ordinary least squares into the calculator and the κ₂ value answers the collinearity question the regression standard errors depend on.
• • Flags singular matrices explicitly: When det(A) is zero or below 1e-10, the calculator returns κ = Infinity in every norm with a clear singular warning.
• • Shows the underlying norm breakdown: The result panel also lists ||A||₁, ||A||∞, ||A||F, and ||A||₂ so the user can verify the condition number by hand.

Use this tool as a quick sensitivity check, not as a replacement for a tuned numerical solver. Read det(A) first; a zero determinant is a stronger statement than a large condition number.

The 2x2 inverse inside this calculator is built from the adjugate, and the Adjoint Matrix Calculator returns that adjugate directly when the user wants to see the unnormalized inverse by hand.

Five factors determine how large the condition number is and which norm gives the largest value.

• • The calculator only handles 2x2 matrices. The same formulas apply to larger matrices, but the singular values would have to come from an SVD routine.
• • A floating-point κ₂ above 1e6 does not mean the answer is wrong; it means the answer is sensitive to small changes in the input.
• • The Frobenius condition number is generally larger than the 2-norm condition number, so a verdict based on κF alone is more conservative.

According to Wolfram MathWorld Condition number, the condition number satisfies κ(A) ≥ 1 in every norm with equality when A is a scalar multiple of a unitary matrix, and the 2-norm condition number of a real matrix equals the ratio of the largest singular value to the smallest.

According to Higham What Is a Condition Number, a matrix with spectral 2-norm condition number near 1 is well-conditioned, a value around 1e3 is the practical threshold for ill-conditioning in floating-point arithmetic, and a singular matrix has an infinite condition number.