|
|
|
|
|
|
|
|
|
|
|
|
The Toolkit for Advanced Optimization (TAO) focuses on large-scale optimization software, including nonlinear least squares, unconstrained minimization, bound constrained optimization, and general nonlinear optimization. There is a variety of software tools for solving the aforementioned problems; however, only TAO offers an object-oriented based solution that provides a flexible optimization toolkit capable of addressing issues of portability, versatility and scalability in many computational environments.
The algorithms in the toolkit place strong emphasis on the reuse of external tools where appropriate. TAO's design enables bidirectional connection to lower level linear algebra support (e.g., parallel sparse matrix data structures) that is available in toolkits like PETSc, as well as higher level application frameworks like POOMA.
TAO's design has been strongly motivated by the needs of many scientist who develop codes for high performance computing environments and in many cases work with legacy codes. Figure 1 illustrates the functionality of TAO and the way it interfaces with with external libraries and application codes.
One component of the TAO is the GPCG algorithm for solving bound constrained, convex quadratic problems. Originally developed by Moré and Toraldo, 1991, this algorithm was designed for large-scale problems but had only been implemented for a single processor. Algorithm GPCG is especially suitable for large scale problems. It combines the advantages of low computational complexity offered by first order gradient methods and rapid local convergence offered by higher order methods.
The image in the left panel correspond to an application
of TAO. Click on the small icon to
display the full image (a new window will open).
- Related Tools:
TAO has been developed at Argonne National Laboratory, in the Mathematics and Computer Science Division; its principal developers are Steve Benson, Lois Curfman McInnes and Jorge More.
| Tools | News | Home |