Introduction to Computational Intelligence
Prof. Dr. Oliver Wendt
Dr. habil. Mahdi Moeini
Summer term 2018
Important notice: T.A.
The lectures take place as a block during June 18-22, 2018 in 42-421a.
Please note that there might be some slight modifications in the following planning:
- Monday (18.06.18): 09:00-12:00 and 13:00-16:00
- Tuesday (19.06.18): 09:00-12:00 and 13:00-16:00
- Wednesday (20.06.18): 14:00-17:00
- Thursday (21.06.18): 09:00-12:00 and 14:00-17:00
- Friday (22.06.18): 09:00-12:00 and 14:00-17:00
If you need further information, please contact Dr. habil. Mahdi Moeini via: mahdi.moeini[at]wiwi.uni-kl.de
The module "Computational Intelligence" comprises the two parts "Optimization of Logistics Systems" and "Introduction to Computational Intelligence". Both lectures are given in English. The whole module has 6 (2x3) ECTS.
Part: Introduction to Computational Intelligence:
For many assignment and permutation problems an exponential growth of the number of solutions prohibits the application of optimization algorithms known from Operations Research. Rather, literature and practitioners resort to the application of heuristics. Heuristics come with much lower computational effort but as a downside - cannot provide a guarantee for the optimality of the solutions found. First, the course focuses on local search heuristics inspired by analogies to nature (Genetic Algorithms and Simulated Annealing) and Tabu Search and compares their applicability for different classes of planning problems. Furthermore, most decision processes do not only confront us with a high number of alternatives but also with uncertainty. We will show how Machine Learning (esp. Reinforcement Learning) can address this uncertainty in complex decision processes, when an appropriate representation of the search space and the value functions can be found. Artificial Neural Networks are introduced (as another paradigm in analogy to nature) as a computational solution of this representational problem.
- Planning: Easy versus hard optimization problems
- Problem Space Search: A* - Optimal Search for an Optimal Solution
- Solution Space Search
- Populational Search I
- Populational Search II
- Artificial Neural Networks
- Optimizing Stochastic Decision Processes: Reinforcement Learning
- Integrating Application Scenario: Dynamic Pricing of Services
- Representing & Learning Stochastic Knowledge: Bayesian Belief Networks
- Representing & Learning Relational Knowledge: Statistical Relational Learning
Lise Getoor's video lecture can be found at videolectures.net/ecml07_getoor_isr/.