COSA: Cooperative Simulated Annealing
-
This is the home of our current (and forthcoming) applications of Cooperative
Simulated Annealing (COSA), a populational extension of Simulated Annealing,
meant to find good solutions for large combinatorial optimization problems.
-
COSA has been developed by Oliver
Wendt as part of his dissertation
thesis.
JAVA-Applet of COSA
Tim Stockheim
converted my COSA sources to JAVA and developed a really nice graphical
interface! Well, in the beginning it was even slower than my initial CLOS
(Common Lisp Object System) implementation, but... thanks to JIT
compilation.... those days are gone...
Just load a TSP problem instance and then click on optimize and run
COSA. Hint: Do NOT try Grötschel's 442 problem without having refilled
your fridge first! Rather run the Christofides&Eilon 51 problem with
the default parameters.
SHOW
ME RIGHT HERE AND NOW!
HINT: Just load a TSP problem instance and then click OPTIMIZE (although
it does not look to be clickable!! [JAVA-Bug]) to run COSA. Do NOT try
Grötschel's 442 problem without having refilled your fridge first!
Rather run the Christofides&Eilon 51 problem with the default parameters.
What's it all about ?
When does it work ?
Publications
Any real-world applications ?
From Sept.'97 to August '98 we have been co-operating with the "Zentrum
für Dermatologie und Venerologie" in order to develop a
personnel scheduling application. The objective of planning (and re-planning)
is threefold:
-
To ensure that required personnel is available at every shift
-
To minimize excess hours (and hereby cost)
-
To minimize the deviation from the old plan (since other plans of the employees
may rely on its stability)
Last modification: 98/08/21 Author: Oliver
Wendt