[Coin-announce] Introducing METSlib Metaheuristic library.

Mirko Maischberger mirko.maischberger at gmail.com
Wed Jul 22 11:55:50 EDT 2009

Hello COIN-OR developers,

I'm pleased to announce METSlib, a metaheuristics modeling framework and
optimization toolkit in C++ released under the GPLv3 or, at your 
opinion, the CPL 1.0.


In METSlib model and agorithms are modular: *any provided search
algorithm* can be applied to the *same model*. On the other hand *no
assumption is made on the model*, you can work on any problem type:
timetabling, assignment problems, vehicle routing, bin-packing and so on.

Once you have implemented your model in the problem framework, the
library makes easy testing different Tabu Search strategies or even
different algorithms (Simulated Annealing or other local search based
algorithms) with a few lines of code.

METSlib hides the nuts and bolts of the algorithms from your code, you
are only supposed to code the problem logic, not the search algorithm.
This results in clearer code, which is also easier to read: your search
strategy can be better understood and modified with less effort.

The toolkit implements the basics of some metaheuristic algorithms:

* Local search
* Simulated Annealing
* Last but not least, *Tabu Search*.

It's really easy to code other metaheuristics using the previous ones as
building blocks:

* Random Restart Local Search
* Variable Neighborhood Search
* Iterated Local Search

You can get the latest METSlib 0.4 sources and examples using the
following subversion commands:

$ svn co https://projects.coin-or.org/svn/metslib/stable/0.4/ metslib-0.4
$ svn co https://projects.coin-or.org/svn/metslib/Examples/stable/0.4/

You can build it with ./autogen.sh; make; make install

There is still no release, but 0.4.3 is due in a few weeks: please help 
in testing the code.

Mirko Maischberger

More information about the Coin-announce mailing list