### Supervisors:

Faramarz Khosravi, Behnaz Pourmohseni

### Module description:

**ECTS information**:

2 SWS (5 ECTS)

**Fields of study**:

Informatik, CE und I&K

### Initial Meeting

Friday 12.05.2017, 10:00 – 12:00 at the room number 02.112-128, Chair of Computer Science 12 (Cauerstr. 11).

### Documents:

All Documents can be found in StudOn.

### Useful Links

Optimizer Submission

Submitted Optimizers

Optimizer Ranking

Extended Results

#### Description:

Meta-heuristic optimization techniques like Evolutionary Algorithms or Simulated Annealing have gained a huge popularity whenever problems are too complex to be reasonably tackled with complete or brute-force approaches. Over the years, a smorgasbord of meta-heuristics have been developed in both the scientific community as well the industry. At this juncture, the landscape of meta-heuristics is vast and, particularly within the scientific community, there are numerous variations of these techniques which have been tailored, extended, and/or tweaked to solve very specific problems more efficiently.

The purpose of the *Black Box Challenge* is to compare the performance of different meta-heuristic optimization techniques by applying them to arbitrary problem instances, about which only minimal information is exposed. I.e. no one knows what kind of problem is „in the box“. Opposed to the trend of tailoring optimization techniques to a particular problem, we want to find out which approaches perform best in a fair comparison over a wide range of different problems. Such a comparison provides useful information for everyone who needs to use a meta-heuristic simply as a tool. In short, we seek for the meta-heuristic that features flexibility instead of specialization.

In this seminar, each student will be provided with an existing meta-heuristic optimization algorithm from literature. This algorithm shall be implemented in the Java-based meta-heuristic optimization framework *Opt4J*. This basic implementation will already take part in the *Black Box Challenge* automatically. Afterwards, each student may start improving this algorithm to achieve better results in the competition. Depending on the number of registrations, students may work in small groups.

The seminar finishes with a session of talks where each student introduces both the optimization algorithm from literature as well as the applied enhancements to the other participants.

#### Literature:

- M. Lukasiewycz, M. Glaß, F. Reimann and J. Teich. Opt4J – A Modular Framework for Meta-heuristic Optimization. Proceedings of the Genetic and Evolutionary Computing Conference (GECCO 2011), pp. 1723–1730, Dublin, Ireland, Jul. 12–16, 2011.

[doi:10.1145/2001576.2001808]