Abstract:Lanczos-based methods have become standard tools for tasks involving matrix functions. Progress on these algorithms has been driven by several largely disjoint communities, resulting many innovative and important advancements which would not have been possible otherwise. However, this also has resulted in a somewhat fragmented state of knowledge and the propagation of a number of incorrect beliefs about the behavior of Lanczos-based methods in finite precision arithmetic. This monograph aims to provide an accessible introduction to Lanczos-based methods for matrix functions. The intended audience is scientists outside of numerical analysis, graduate students, and researchers wishing to begin work in this area. Our emphasis is on conceptual understanding, with the goal of providing a starting point to learn more about the remarkable behavior of the Lanczos algorithm. Hopefully readers will come away from this text with a better understanding of how to think about Lanczos for modern problems involving matrix functions, particularly in the context of finite precision arithmetic.

What problem does this paper attempt to address?

### Problems the paper attempts to solve The paper aims to address the behavior of the Lanczos algorithm related to matrix functions in finite - precision arithmetic. Specifically, the paper focuses on the following aspects: 1. **Fragmented knowledge state**: Although the Lanczos algorithm and its variants are becoming increasingly important in tasks involving matrix functions, the development in different research communities has led to fragmented knowledge in this area. This has allowed some wrong views to spread, especially regarding the behavior of the Lanczos algorithm in finite - precision arithmetic. 2. **Gap between theory and practice**: There is a significant difference between the theoretical behavior of the Lanczos algorithm in exact arithmetic and its behavior in finite - precision arithmetic in practical applications. This difference may lead to misunderstandings and incorrect applications. 3. **Providing clear understanding**: The goal of the paper is to provide an easy - to - understand introductory guide for scientists outside the field of numerical analysis, graduate students, and researchers who wish to enter this field, helping them better understand the application of the Lanczos algorithm in modern matrix function problems, especially in the context of finite - precision arithmetic. ### Overview of the main content - **Introduction**: Introduces the importance of the Lanczos algorithm and its applications in different tasks, such as solving linear systems, matrix functions acting on vectors, quadratic forms, and trace estimation. - **Arnoldi and Lanczos algorithms**: Describes in detail how these two algorithms work, especially how the Lanczos algorithm uses symmetry and orthogonal polynomials to reduce computational complexity. - **Orthogonal polynomials**: Explores the relationship between orthogonal polynomials and the Lanczos algorithm, especially their behavior in finite - precision arithmetic. - **Lanczos algorithm in finite - precision arithmetic**: Discusses the theoretical and practical performance of the Lanczos algorithm in finite - precision arithmetic, including the theories of Paige, Greenbaum, and Knizhnerman, etc. - **Application examples**: Demonstrates the application of the Lanczos algorithm in different tasks through specific examples, such as solving linear systems, matrix functions acting on vectors, quadratic forms, and trace estimation. ### Target audience The target audience of the paper includes: - Scientists outside the field of numerical analysis - Graduate students - Researchers who wish to enter this field The paper emphasizes conceptual understanding and aims to provide a starting point for these readers to enable them to gain a deeper understanding of the behavior of the Lanczos algorithm in modern matrix function problems.

The Lanczos algorithm for matrix functions: a handbook for scientists

Nearly Optimal Approximation of Matrix Functions by the Lanczos Method

Error Bounds for Lanczos-Based Matrix Function Approximation

On the Lanczos Method for Computing Some Matrix Functions

A posteriori error bounds for the block-Lanczos method for matrix function approximation

LAW: A Tool for Improved Productivity with High-Performance Linear Algebra Codes. Design and Applications

A low-memory Lanczos method with rational Krylov compression for matrix functions

Gradients of Functions of Large Matrices

The Linear Algebra Mapping Problem. Current state of linear algebra languages and libraries

Generalized Block Lanczos Methods for Large Unsymmetric Eigenproblems

Stability of the Lanczos algorithm on matrices with regular spectral distributions

The Convergence of Generalized Lanczos Methods for Large Unsymmetric Eigenproblems

Sparse Matrix Implementation in Octave

Matrix computations (2nd edition) , by G. H. Golub and C. F. Van Loan. Pp 642. £38. 1989. ISBN 0-8018-3772-3 (John Hopkins Press)

A randomized small-block Lanczos method for large-scale null space computations

A literature survey of matrix methods for data science

Lanczos for lattice QCD matrix elements

Lanczos, the transfer matrix, and the signal-to-noise problem

Improving the numerical stability of fast matrix multiplication