Design and Application of Metaheuristics

Metaheuristics like evolutionary algorithms, genetic programming, variable neighborhood search, tabu search, simulated annealing, and others are applied to large, difficult, or realistic optimization problems, for which efficient classical optimization methods are not available or applicable. This includes the optimization of other AI systems like neural networks.  Many text books teach such methods by providing detailed descriptions of the functionality of single examples of metaheuristics neglecting the underlying and common concepts. As a result, the systematic design and application of metaheuristics is often not a systematic engineering task but a result of repeated trial and error. Applicants apply textbook approaches and are surprised that they do not perform well when used for problems of realistic size or complexity. 

This course at hand takes a different approach. It teaches the basic, method-independent principles and design guidelines of metaheuristics and how they can be used to systematically develop superior heuristic optimization methods for problems of choice. Consequently, the course focuses on the application side and deals with four fundamental questions: 

1. It tells the participants on which problems metaheuristics are expected to perform well, and what are problems where other optimization paradigms are a better choice. 
2. Participants learn to systematically design an appropriate metaheuristic for a particular problem using a coherent view on design elements and working principles of metaheuristics. 
3. Participants learn how to make use of problem-specific knowledge for the design of efficient and effective metaheuristics that solve not only small toy problems but also perform well on large and real-world problems.
4. Participants learn what are the different application domains of metaheuristics and neural network models as well as how both approaches can be combined.  

The course consists of two parts: the first part interactively teaches basic underlying concepts; the second discusses application examples or application problems provided by the participants. Thus, participants are encouraged to bring in their own application problems. It is not expected that all participants bring in own examples. If the participants do not have any problems they want to solve, representative example problems are provided by the lecturer. 

The course is aimed at PhD-students from various disciplines, including operations, production and logistics, supply chain management, finance, marketing and management science. Of course, the course fits well for PhD-students working in the field of neural networks, artificial intelligence, or data science. The elements of the course range from lectures on theory, to the discussion of best practices and case studies, and finally to the presentation of work done by the students.


Anmeldefrist: 30. August 2026

Sprache / Language

Englisch

Ort / Location

Tickets

Noch 20 Plätze verfügbar.