Lanczos's method

Please type the dimension of the matrix:
n= important: 3 <= n <= 1000 !

How many eigenvalues should be found?
numew= important: 1 <= numew <= min{n,300}

Give a limit for the number of steps :
numsteps = important: numew <= numstep <= 300

choose the type of reorthogonalization:
none at all
selective
complete

choose epsaccept for Parlett's acceptance criterion
epsaccept = important: 1.0e-16 <= epsaccept <= 1.0e-2

type your initial vector: n values are required. Values separated by comma or white space. Input can extend over several lines.
If you want the vector generated internally at random, then type "xxx" here.
x0:

write here the piece of code which performs y=A*x. The result must be a vector y. Input is the vector x and n is the dimension. You might use: a (full) matrix A,(i.e. you address it as A(i,j), i,j=1,..,n which you must initialize here then , a temporary storage vector zeisum with n components, integer variables i,j,k (for loops e.g.), logical variables bool1, bool2, bool3, constants pi, e=exp(1), sqrt2=sqrt(2.0) . It is impossible to define new variables yourself.

Click on "evaluate", in order to submit your input.

back to the theory page

 Back to the top!

14.01.2013