Parlett The Symmetric Eigenvalue Problem Pdf ✧

The Soul of a Matrix: Why Parlett’s "Symmetric Eigenvalue Problem" is Still Must-Read

  • Option A: accumulate Q explicitly (cost O(n^2) memory/time) if eigenvectors needed.
  • Option B: store compact reflectors (vectors and scalars) and apply later to backtransform.

Limitations (to be aware of):

  • No parallel computing: The book predates MPI, GPU, or distributed memory models. You must adapt the algorithms yourself.
  • Dense matrices dominate: Sparse direct methods (e.g., nested dissection) are barely mentioned.
  • No mention of randomized SVD or modern randomized eigendecomposition (post-2000).
  • Complexity notations are classical: Flop counts are given, but cache-efficiency and memory hierarchies are not discussed.

Who Should Avoid It?