TNPACK

Nonlinear unconstrained minimization of large-scale separable problems

A truncated Newton method for unconstrained minimization has been specifically developed for large-scale separable problems. TNPACK uses a preconditioned conjugate gradient method to solve the Newton equations approximately at every step. Modifications are incorporated to handle indefiniteness of both the Hessian and the preconditioner. The preconditioning matrix (usually a sparse approximation to the Hessian) is provided by the user. It is factored by a sparse modified Cholesky factorization based on the Yale Sparse Matrix Package. TNPACK is intended to solve complex problems that arise in practical applications, such as computational chemistry and biology, where a natural separability or hierarchy in complexity exists among the different functional components. The user can adapt details of the algorithm to suit the problem at hand (for example, by preconditioning and variable reordering).

Software is writen in double precision ANSI Fortran 77.

The code is available from from Netlib.

Need more info?

Contact:

Tamar Schlick 
Courant Institute of Mathematical Sciences
251 Mercer Street 
New York, NY 10012  
Phone: (212) 998-3116
schlick@acfclu.nyu.edu

References:

T. Schlick and A. Fogelson, TNPACK -- A truncated Newton minimization package for large-scale problems. I: Algorithm and usage, ACM. Trans. Math. Software 18 (1992), pp. 46--70.

T. Schlick and A. Fogelson, TNPACK -- A truncated Newton minimization package for large-scale problems. II: Implementation examples, ACM. Trans. Math. Software 18 (1992), pp. 71--111.

T. Schlick and W. K. Olson, Supercoiled DNA structure and dynamics by computer simulations, J. Mol. Biol. 223 (1992), pp. 1089--1119.

T. Schlick and M. L. Overton, A powerful truncated Newton method for potential energy functions, J. Comp. Chem., 8 (1987), pp. 1025--1039.

[ Optimization Software Guide | OTC Home Page | NEOS Server | NEOS Guide ]