Why is matrix inversion considered numerically unstable in practice?

Updated May 16, 2026

Short answer

Matrix inversion amplifies numerical errors, especially for ill-conditioned matrices.

Deep explanation

Computing A⁻¹ explicitly is unstable because small floating-point errors in A get amplified, especially when A has a high condition number. Instead of inversion, numerical systems solve Ax=b directly using factorization methods like LU or QR decomposition to avoid instability.

Real-world example

Used in ML training where stability is more important than exact inversion.

Common mistakes

  • Explicitly computing inverse instead of solving linear systems.

Follow-up questions

  • Why does condition number matter?

More Linear Algebra interview questions

View all →