跳至内容

Lanczos Algorithm

The Lanczos method is an iterative algorithm that reduces a symmetric matrix AA of size N×NN \times N to a tridiagonal matrix TT of size m×mm \times m, where mNm \ll N. The eigenvalues of TT approximate the extremal eigenvalues of AA, and the corresponding eigenvectors can be reconstructed.

Key Steps of the Lanczos Method

  1. Initialization:

    • Choose a random starting vector v1v_1 with unit norm.
    • Set β0=0\beta_0 = 0 and v0=0v_0 = 0.
  2. Iteration: For j=1,2,,mj = 1, 2, \dots, m:

    • Compute w=Avjβj1vj1w = A v_j - \beta_{j-1} v_{j-1}.
    • Compute αj=vjw\alpha_j = v_j^\top w.
    • Compute w=wαjvjw = w - \alpha_j v_j.
    • Compute βj=w\beta_j = \|w\|.
    • If βj=0\beta_j = 0, stop; otherwise, set vj+1=w/βjv_{j+1} = w / \beta_j.
  3. Tridiagonal Matrix: After mm iterations, the matrix TT is constructed as:

    T=(α1β100 β1α2β20 0β2α30 βm1 000βm1αm) T = \begin{pmatrix} \alpha_1 & \beta_1 & 0 & \dots & 0 \\\ \beta_1 & \alpha_2 & \beta_2 & \dots & 0 \\\ 0 & \beta_2 & \alpha_3 & \dots & 0 \\\ \vdots & \vdots & \vdots & \ddots & \beta_{m-1} \\\ 0 & 0 & 0 & \beta_{m-1} & \alpha_m \end{pmatrix}
  4. Diagonalization of TT:

    • Diagonalize TT using standard dense matrix techniques (e.g., QR algorithm).
    • The eigenvalues of TT approximate the extremal eigenvalues of AA.
    • The corresponding eigenvectors of AA can be reconstructed from the Lanczos vectors vjv_j.

Advantages of the Lanczos Method

  1. Efficiency:

    • Only matrix-vector products are required, making it suitable for sparse matrices.
    • Memory usage is O(Nm)O(N \cdot m) instead of O(N2)O(N^2).
  2. Scalability:

    • Works well for very large matrices where dense methods are infeasible.
  3. Focus on Extremal Eigenvalues:

    • The Lanczos method is particularly effective at finding the largest or smallest eigenvalues and their eigenvectors.

Challenges and Considerations

  1. Loss of Orthogonality:

    • In finite-precision arithmetic, the Lanczos vectors vjv_j can lose orthogonality, leading to spurious eigenvalues.
    • Remedies include reorthogonalization or using more advanced variants like the Implicitly Restarted Lanczos Method.
  2. Choice of mm:

    • The number of iterations mm must be chosen carefully to balance accuracy and computational cost.
  3. Convergence:

    • Convergence to the extremal eigenvalues is typically fast, but interior eigenvalues may require many iterations.