Karsten Weicker

Evolutionary multiobjective optimization for base station transmitter placement with frequency assignment

We propose a new solution to the problem of positioning base station transmitters of a mobile phone network and assigning frequencies to the transmitters, both in an optimal way. Since an exact solution cannot be expected to run in polynomial time for all interesting versions of this problem (they are all NP-hard), our algorithm follows a heuristic approach based on the evolutionary paradigm. For this evolution to be efficient, i.e., goal-oriented and sufficiently random at the same time, problem-specific knowledge is embedded in the operators. The problem requires both the minimization of the cost and of the channel interference. We examine and compare two standard multiobjective techniques and a new algorithm - the steady-state evolutionary algorithm with Pareto tournaments. One major finding of the empirical investigation is a strong influence of the choice of the multiobjective selection method on the utility of the problem-specific recombination leading to a significant difference in the solution quality.

Performance Measures for Dynamic Environments

This article investigates systematically the utility of performance measures in non-stationary environments. Three characteristics for describing the goals of a dynamic adaptation process are proposed: accuracy, stability, and recovery. This examination underpins the usage of the best fitness value as a basis for measuring the three characteristics in scenarios with moderate changes of the best fitness value. However, for dynamic problems without coordinate transformations all considered fitness based measures exhibit severe problems. In case of the recovery, a newly proposed window based performance measure is shown to be best as long as the accuracy level of the optimization is rather high.

On the improvement of coevolutionary optimizers by learning variable interdependencies

During the last years, cooperating coevolutionary algorithms could improve the convergence of several optimization benchmarks significantly by placing each dimension of the search space in its own subpopulation. However, their general applicability is restricted by problems with epistatic links between problem dimensions, a major obstacle in cooperating coevolutionary function optimization. The work presents first preliminary studies on a technique to recognize epistatic links in problems and self-adapt the algorithm in such a way that populations with interrelated dimensions are merged to a common population.

On evolution strategy optimization in dynamic environments

The article analyzes the behavior of evolution strategies and their current mutation variants on a simple rotating dynamic problem. The degree of rotation is a parameter for the involved dynamism which enables systematic examinations. As a result, the complex covariance matrix adaptation proves to be superior with slow rotation but with increasing dynamism whose adaptation mechanism seldom finds the optimum where the simple uniform adaptation produces stable results. Moreover, this examination gives rise to questioning the principle of small mutation changes with high probability in the dynamic context.

Automated Solution of a Highly Constrained School Timetabling Problem - Preliminary Results

This work introduces a highly constrained school timetabling problem which was modeled from the requirements of a German high school. The concept for solving the problem uses a hybrid approach. On the one hand an evolutionary algorithm searches the space of all permutations of the events from which a timetable builder generates the school timetables. Those timetables are further optimized by local search using specific mutation operators. Thus, only valid (partial) timetables are generated which fulfill all hard constraints.

Burden and Benefits of Redundancy

Abstract This work investigates the effects of two techniques introducing redundancy, diploid representations and decoders. The effects of redundancy on mutation, recombination, the structure of the landscape and the interplay between these ingredients are of particular interest. Resulting neutral mutations are found to be the reason for the transformation of local optima into platenu points. Additional neutral recombinations enable a more efficient search on those neutral plateaus. Furthermore, both kinds of neutral operations are shown to have a positive impact on the diversity of the population over time. Lastly, the diploid representation is compared empirically to a macromutation which models the working principles of diploidity. However, this control experiment shows that blind exploring mutations cannote approximate diploidity.

Dynamic rotation and partial visibility

This article generalizes a previously presented dynamic fitness function with two different concepts, namely a coordinate rotation and the concept of partial visibility. Those concepts define different classes of test problems. A set of standard evolution strategies and genetic algorithms with and without hypermutation are tested on two of the dynamic problem classes. They give insight into certain properties of the presented concepts and dynamic optimization in general.

Analysis of Local Operators Applied to Discrete Tracking Problems

This article investigates various local operators in a discrete (1, λ)-setting applied to tracking problems, a specific class of non-stationary problems. In the first instance, the influence of operator properties on the tracking performance is examined. Both the enforcement of bigger steps and, especially, directed mutations are found to increase the tracking accuracy considerably. For the examination of highly time restricted problems, a correlation between the population size and the severity of the problem dynamics is assumed. Relatively large population sizes are found to be advantageous if the number of evaluations has a big influence on the severity. All results are obtained using a fixpoint analysis of a worst-case model as well as simulations within a two-dimensional Markov model.

Development of Problem-Specific Evolutionary Algorithms

It is a broadly accepted fact that evolutionary algorithms (EA) have to be developed problem-specifically. Usually this is based on experience and experiments. Though, most EA environments are not suited for such an approach. Therefore, this paper proposes a few basic concepts which should be supplied by modern EA simulators in order to serve as a toolkit for the development of such algorithms.

Maximum flow networks for stability analysis of LEGO®Structures

To determine the stability of LEGO® structures is an interesting problem because the special binding mechanism prohibits the usage of methods of structural frame design or dynamic physics engines. We propose a new two-phase approach where instances of maximum-flow networks are constructed. In a first phase, the distribution of compressive and tensile forces is computed which is used in a second phase to model the moments within the structure. By solving the maximum-flow networks we can use the resulting flow as a sufficient criterion for the stability of the structure. The approach is demonstrated for two exemplary structures which outperform previous results using a multi-commodity flow network.