# Stability of the Lanczos Method for Matrix Function Approximation

arXiv: https://arxiv.org/abs/1708.07788 · HTML: https://mlsanity.com/paper/1708.07788/

## Key facts

- Cited by 16 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 25 Aug 2017.

## Built on by (most cited first, 16 of 16)

- [Solving Dense Linear Systems Faster Than via Preconditioning](https://mlsanity.com/paper/2312.08893/) (2023)
- [Optimal Randomized Approximations for Matrix based Renyi's Entropy](https://mlsanity.com/paper/2205.07426/) (2022)
- [Efficient Graph Field Integrators Meet Point Clouds](https://mlsanity.com/paper/2302.00942/) (2023)
- [Fine-grained Analysis and Faster Algorithms for Iteratively Solving Linear Systems](https://mlsanity.com/paper/2405.05818/) (2024)
- [Faster Linear Systems and Matrix Norm Approximation via Multi-level Sketched Preconditioning](https://mlsanity.com/paper/2405.05865/) (2024)
- [Approaching Optimality for Solving Dense Linear Systems with Low-Rank Structure](https://mlsanity.com/paper/2507.11724/) (2025)
- [Sublinear Time Quantum Algorithm for Attention Approximation](https://mlsanity.com/paper/2602.00874/) (2026)
- [Towards Universal Convergence of Backward Error in Linear System Solvers](https://mlsanity.com/paper/2604.16075/) (2026)
- [Scaling Up Graph Propagation Computation on Large Graphs: A Local Chebyshev Approximation Approach](https://mlsanity.com/paper/2412.10789/) (2024)
- [Theoretically and Practically Efficient Resistance Distance Computation on Large Graphs](https://mlsanity.com/paper/2601.11159/) (2026)
- [SWING: Unlocking Implicit Graph Representations for Graph Random Features](https://mlsanity.com/paper/2602.12703/) (2026)
- [Well-Conditioned Oblivious Perturbations in Linear Space](https://mlsanity.com/paper/2604.23193/) (2026)
- [Don't Get Your Kroneckers in a Twist: Gaussian Processes on High-Dimensional Incomplete Grids](https://mlsanity.com/paper/2605.08036/) (2026)
- [The Value of Covariance Matching in Gaussian DDPMs and the Lanczos Sampler](https://mlsanity.com/paper/2605.22723/) (2026)
- [Smoothed Score Queries and the Complexity of Sampling](https://mlsanity.com/paper/2605.27769/) (2026)
- [Eigenbasis-Independent Learnable Spectral Positional Encodings for Directed Graphs via Hermitian Block Krylov Subspaces](https://mlsanity.com/paper/2607.07032/) (2026)

## Abstract (arXiv)

The ubiquitous Lanczos method can approximate f(A)x for any symmetric n × n matrix A, vector x, and function f. In exact arithmetic, the method's error after k iterations is bounded by the error of the best degree-k polynomial uniformly approximating f(x) on the range [λmin(A), λmax(A)]. However, despite decades of work, it has been unclear if this powerful guarantee holds in finite precision. We resolve this problem, proving that when max_x ∈ [λmin, λmax]|f(x)| ≤ C, Lanczos essentially matches the exact arithmetic guarantee if computations use roughly log(nC|A|) bits of precision. Our proof extends work of Druskin and Knizhnerman [DK91], leveraging the stability of the classic Chebyshev recurrence to bound the stability of any polynomial approximating f(x). We also study the special case of f(A) = A-1, where stronger guarantees hold. In exact arithmetic Lanczos performs as well as the best polynomial approximating 1/x at each of A's eigenvalues, rather than on the full eigenvalue range. In seminal work, Greenbaum gives an approach to extending this bound to finite precision: she proves that finite precision Lanczos and the related CG method match any polynomial approximating 1/x in a tiny range around each eigenvalue [Gre89]. For A-1, this bound appears stronger than ours. However, we exhibit matrices with condition number kappa where exact arithmetic Lanczos converges in polylog(kappa) iterations, but Greenbaum's bound predicts Omega(kappa1/5) iterations. It thus cannot offer significant improvement over the O(kappa1/2) bound achievable via our result. Our analysis raises the question of if convergence in less than poly(kappa) iterations can be expected in finite precision, even for matrices with clustered, skewed, or otherwise favorable eigenvalue distributions.