While most theoretical run time analyses of discrete randomized search heuristics provide bounds on the expected number of evaluations to find the global optimum, Jurek Sander and Dr. habil Timo Kötzing consider in the paper Anytime Analysis on BinVal: Adaptive Parameters Help the anytime performance of evolutionary algorithms. They showed that especially a (1+1) EA with self-adjusting mutation rate is well-suited to optimize the most significant bits.
In the project Analysis of Search Heuristics in the Multi-Armed Bandit Setting, Jurek Sander and Dr. habil Timo Kötzing combined their knowledge of randomized search heuristics with the research field of Jasmin Brandt: Multi-Armed Bandits. For the "Dueling Bandits" setting, they showed that while the standard (1+1) EA struggles to find the Condorcet Winner - the arm that beats every other arm with a probability strictly higher than 1/2 -, a simple EDA chooses it with a larger probability.
The paper Gray-Box Optimization and the Vertex Coloring Problem, a BSc-student project coordinated by Dr. habil Timo Kötzing analyzes different heuristic algorithms for coloring the vertices of a graph such that no two neighbors are colored the same. The paper shows that using problem knowledge for the operators and fitness function used can greatly reduce the optimization time of standard search heuristics.
The papers will be presented at Gecco 2026 on July 13-17 in San José, Costa Rica.
