Power iteration with a dominant eigenvalue
For diagonalizable A with |λ₁|>|λ₂|≥… and a start having nonzero λ₁-eigenvector coefficient, repeated x_(k+1)=Ax_k/‖Ax_k‖₂ approaches the dominant eigendirection, up to phase or sign.
Key knowledge
For diagonalizable A with |λ₁|>|λ₂|≥… and a start having nonzero λ₁-eigenvector coefficient, repeated x_(k+1)=Ax_k/‖Ax_k‖₂ approaches the dominant eigendirection, up to phase or sign. Directional error is O((|λ₂|/|λ₁|)ᵏ); normalization limits overflow but does not supply the required spectral gap.
Conditions and scope
- The stated convergence is in exact arithmetic and requires λ₁≠0; n≥2.
- A tied dominant magnitude, missing start component or defective matrix requires another analysis.
Sources · 1
References
Sources retain their recorded scope. A source link alone does not establish verification of every statement.
Evidence and review
Section8.2.1 equations8.2.1–5, Algorithm8.2.1 and section8.2.3 expansion8.2.10; no unqualified error-ratio limit adopted when the leading coefficient vanishes.
2026-10-10 · Independent AI editorial review: All241 compact claims, hypotheses, displayed equations/conventions and all596 origin semantic links read; entire241-entry author exact-source inspection matrix and declared limitations reviewed. Prior105 actual inherited Mathematics/Information records read for whole-inventory breadth. Targeted independent primary sections are listed below; this does NOT claim every cited source body or every proof was independently re-read. This is AI reference-content review, not human disciplinary certification, exhaustive textbook reproduction or experimental replication. Receipt: mathematics-information-independent-content-review.1.
Entry details
Permanent link to this entry ↗- Record identity
- math-power-eigenvalue-iteration
- Revision
- university-core.1
- Structured reference record
- University core expansion