Lanczos Algorithm
The Lanczos method is an iterative algorithm that reduces a symmetric matrix of size to a tridiagonal matrix of size , where . The eigenvalues of approximate the extremal eigenvalues of , and the corresponding eigenvectors can be reconstructed.
Key Steps of the Lanczos Method
Initialization:
- Choose a random starting vector with unit norm.
- Set and .
Iteration: For :
- Compute .
- Compute .
- Compute .
- Compute .
- If , stop; otherwise, set .
Tridiagonal Matrix: After iterations, the matrix is constructed as:
Diagonalization of :
- Diagonalize using standard dense matrix techniques (e.g., QR algorithm).
- The eigenvalues of approximate the extremal eigenvalues of .
- The corresponding eigenvectors of can be reconstructed from the Lanczos vectors .
Advantages of the Lanczos Method
Efficiency:
- Only matrix-vector products are required, making it suitable for sparse matrices.
- Memory usage is instead of .
Scalability:
- Works well for very large matrices where dense methods are infeasible.
Focus on Extremal Eigenvalues:
- The Lanczos method is particularly effective at finding the largest or smallest eigenvalues and their eigenvectors.
Challenges and Considerations
Loss of Orthogonality:
- In finite-precision arithmetic, the Lanczos vectors can lose orthogonality, leading to spurious eigenvalues.
- Remedies include reorthogonalization or using more advanced variants like the Implicitly Restarted Lanczos Method.
Choice of :
- The number of iterations must be chosen carefully to balance accuracy and computational cost.
Convergence:
- Convergence to the extremal eigenvalues is typically fast, but interior eigenvalues may require many iterations.