Help System (web edition)

MAT #EIGLAN(A,k,S,L,n_iter,L_file)
computes k largest eigenvalues L and
their eigenvectors S of a symmetric matrix A by the Lanczos method.
n_iter (by default 10) is the number of Lanczos iterations.
If L_file is given, Lanczos vectors are saved as L_file.

This is an efficient method for computing a few largest eigenvalues
and their eigenvectors when the dimension of A is large,
say, more than 200.

Reference: Golub and van Loan: Matrix computations,
           Chapter 9.2

MAT #EIGFEW(A,k,S,L,tol,iter)
computes k largest eigenvalues L and corresponding eigenvectors S
of matrix A. Parameters tol and iter are optional.
The simple power method is used and this command is intended for
computing a few (k) eigenvalues and vectors when A is a large
square matrix. tol gives the accuracy (default tol=1e-12) and
iter the maximium number of iterations for each eigenvalue
(default iter=100).
When more accurate results are needed and A is symmetric,
MAT SPECTRAL DECOMPOSITION or MAT #EIGLAN should be used.
In non-symmetric cases MAT #EIGEN should be preferred to.

A = More about additional MAT #operations 
M = More about MAT operations