Trustable Algorithms for Nonlinear General Optimization The University of Campinas and University of São Paulo joint project for optimization software development.

# Code

## Methods

ALGENCAN
Fortran code for general nonlinear programming that does not use matrix manipulations at all and, so, is able to solve extremely large problems with moderate computer time. The general algorithm is of Augmented Lagrangian type and the subproblems are solved using GENCAN. GENCAN (included in ALGENCAN) is a Fortran code for minimizing a smooth function with a potentially large number of variables and box-constraints.
SPG
Fortran code for minimizing a smooth function with convex constraints with a potentially very large number of variables.
OTHER METHODS
In addition to the methods above, there are also other methods that use ALGENCAN and/or SPG as subalgorithms. Those other methods are not being updated. So, for the newest versions of ALGENCAN and/or SPG follow the links above. The available methods are: (i) ALABB: Augmented Lagrangian for Global Optimization that uses the αBB method to find the global solution of linearly constraint subproblems. (ii) ALBETRA: Augmented Lagrangian that uses BETRA to solve bound-constrained subproblems. BETRA is an active set method for bound constraint minimization that uses the classical Euclidian trust-region method inside the faces. SPG directions are used to abandon faces. (iii) ALGENCAN-NEWTON: This method is an attempt to improve the local convergence of ALGENCAN. The Newton′s method is used to solve a KKT system identified by the Augmented Lagrangian algorithm. (iv) ALSPG: Augmented Lagrangian that uses the SPG method to solve convex constrained subproblems. (v) IVM: Inexact Variable Metric method for convex constrained optimization. (vi) SCG: Spectral Conjugate Gradient method for unconstrained optimization. (vii) GENLIN: Partial Spectral Projected Gradient Method with Active-Set Strategy for Linearly Constrained Optimization. (viii) ALGENCAN-OTR: Outer Trust-Region method for Constrained Optimization.

## Applications

PUMA
C code for the estimation of the thickness and the optical constants of thin films. It is based on unconstrained minimization and uses the Spectral Gradient method.
PACKMOL
Fortran code to create initial configurations for molecular dynamics. The problem is modeled as a bound-constrained minimization problem and GENCAN is used to solve it.
CUTTING AND PACKING
Large variety of nonlinear models for packing problems.

 Page last modified: Jan 31, 2007. Page URL: http://www.ime.usp.br/~egbirgin/tango/ Contact: egbirgin@ime.usp.br, egbirgin@gmail.com