I want to use lanczos method to calculate the few lowest eigenvalue and eigen-vector of a large sparse symmetric matrix(~50k x ~50k). In http://www.netlib.org/lanczos/index.html I found the codes leval and levec are suitable for my situation. However, the instruction is too complicated to find how to use it and I don't know how to set the input and output files. Who used it and please gives some advice. Thanks!
This does not directly answer your question, but maybe you want to use something like ARPACK, for example
scipy.sparse.linalg.eigsh in python.