The Relevance of Optimization to Functional Modelling
The task faced by the functional modeller essentially consists of satisfying a very large number of (often incompatible) equality and inequality constraints. The best approach found so far to solving these is to express the unsatisfied constraints as a penalty function, and to treat the problem as a non-linear optimization task. Simulated Annealing appears to offer the best type of optimization approach, as its restricted use of heuristics allows it to explore a promising solution area in detail without the risk of getting trapped in a local minimum.
Back to my Home Page