Algorithms for Large-Scale Eigenvalue Problems

Explore algorithms designed to solve large-scale eigenvalue problems in this comprehensive lecture from the Complexity and Linear Algebra Boot Camp. Learn how these specialized methods differ from dense matrix approaches by focusing on computing only small subsets of the spectrum relevant to specific applications, such as eigenvalues with largest or smallest magnitude or those positioned rightmost in the complex plane. Begin with foundational concepts including the power method and its block generalization through subspace iteration, then advance to more sophisticated techniques like the Lanczos and Arnoldi methods along with their restarted variants. Discover why eigenvalues of large symmetric matrices can typically be computed with high reliability, while understanding the ongoing challenges in convergence theory for nonsymmetric cases. Gain insight into how different algorithms strategically guide computations toward desired eigenvalues, making this essential knowledge for anyone working with large-scale linear algebra problems in computational mathematics and scientific computing applications.