Direkt zum InhaltDirekt zur SucheDirekt zur Navigation
▼ Zielgruppen ▼

Humboldt-Universität zu Berlin - Mathematisch-Naturwissenschaftliche Fakultät - Software Engineering

Master's Thesis Proposal: Sebastian Müller

Wann 21.01.2020 von 14:00 bis 14:30 (Europe/Berlin / UTC100) iCal
Wo 12489 Berlin, Rudower Chaussee 25, Raum IV.410

Sebastian Müller will give the proposal presentation for his master's thesis on the topic "Bet and Run Strategy in the Context of Test Case Generation".



Probably anyone working in the technology sector is familiar with the question: “Have you tried turning it off and on again?”, as it is usually the default question asked by tech support. Similarly, researchers noted that meta-heuristics – as employed in the field of search-based software-engineering (SBSE) – tend to get trapped in a plateau at some point during computation. As a human, one can look at the gradient of the fitness curve and decide that it is time to restart the computation with a slightly different set of initial parameters, so as to hopefully get stuck in another (and in terms of the fitness metric better) plateau. When trying to automate this process however, the problem becomes to decide whether the meta-heuristic has actually encountered such a plateau yet. This reduces directly to the Halting Problem, which is undecidable [Tur37]. To overcome this fact, Tobias Friedrich et al. [FKW17] looked at a strategy called Bet-and-Run for two classical NP-complete problems, where they first started a number of instances with randomized initial conditions, and after some time they only kept running the one instance that showed the most promise in terms of fitness values. In the proposed thesis we will consider and evaluate the Bet And Run stRategy In the context of teST casE geneRation (BARRISTER).

[Tur37] Alan Mathison Turing. On computable numbers, with an application to the entscheidungsproblem. Proceedings of the London mathematical society, 2(1):230–265, 1937.

[FKW17] Tobias Friedrich, Timo Kötzing, and Markus Wagner. A generic bet-and-run strategy for speeding up stochastic local search. In Satinder P. Singh and Shaul Markovitch, editors, Proceedings of the Thirty-First AAAI Conference on Artificial Intelligence, February 4-9, 2017, San Francisco, California, USA, pages 801–807. AAAI Press, 2017.