ANTIGONE

1. Reformulation of user input
2. Detection of special structure
3. Selection of best algorithm for the detected structure
4. Solution of the problem using the selected algorithm

With the exception of special cases of optimization problems (for instance convex NLPs) which can be solved at the root node of the branch-and-bound algorithm, ANTIGONE will initiate a branch and bound procedure. This procedure involves the following steps:

• Generating/updating convex relaxations (lower bounding problems)
• Domain reduction
• Search for feasible solutions (upper bounds)
• Calculation of rigorous lower bounds by solving the convex lower bounding problems.

ANTIGONE employs classic factorable programming techniques to generate relaxations, e.g., McCormick Relaxations, as well as contemporary algorithms, such as the Reformulation-Linearisation Technique (RLT), Edge Convex/Concave relaxations, and

Limitations

Like all deterministic global optimization software, ANTIGONE requires the user to provide the explicit mathematical expressions for all the functions used in the problem, as well as initial bounds for all variables. If initial bounds are not supplied, ANTIGONE will attempt to infer bounds, but global optimality is not guaranteed. ANTIGONE can only solve differentiable functions, and can not solve trigonometric problems.

Interfaces

ANTIGONE is part of the

Licensing