Dr. Ampl — a meta solver for optimization

Home
News
Optimization
What
How
Examples
FAQ
Benchmarks
Installation
Documentation
Mailing list
Project page
Download
To do
Links
Who

To perform the tasks described in the what section, it is fundamental to parse the data structures in which the AMPL engine has stored the information about the problem. Most important is the recursive data structure in which nonlinear functions are stored — a tree referred to as a Directed Acyclic Graph, or DAG for short, which essentially represents the computational graph in the Kantorovitch sense. Nodes of the DAG represent composed functions f o g, where f is the operator represented by the current node and g its operands — the child nodes of f.

An online presentation explaining these mechanisms may prove a good introduction.

A more in-depth paper in under way.