Affiliations: Institut für Softwaretechnik und Theoretische Informatik, TU Berlin Sekr. MAR 5-5, Marchstr. 23, D-10587 Berlin, Germany hartmut.ehrig@tu-berlin.de | Institut für Softwaretechnik und Theoretische Informatik, TU Berlin Sekr. MAR 5-5, Marchstr. 23, D-10587 Berlin, Germany claudia.ermel@tu-berlin.de | Institut für Softwaretechnik und Theoretische Informatik, TU Berlin Sekr. TEL 5-1, Ernst-Reuter-Platz 7, D-10587 Berlin, Germany falk.hueffner@tu-berlin.de | Institut für Softwaretechnik und Theoretische Informatik, TU Berlin Sekr. TEL 5-1, Ernst-Reuter-Platz 7, D-10587 Berlin, Germany rolf.niedermeier@tu-berlin.de | Institut für Softwaretechnik und Theoretische Informatik, TU Berlin Sekr. MAR 5-5, Marchstr. 23, D-10587 Berlin, Germany olga.runge@tu-berlin.de
Note: [] An extended abstract of this paper appears in Proceedings of the 8th Conference on Computability in Europe (CiE 2012), volume 7318 in Lecture Notes in Computer Science, pages 193–202, Springer, 2012. The present version contains additional details in Section 4 and all proofs.
Abstract: Kernelization is a core tool of parameterized algorithmics for coping with computationally intractable problems. A kernelization reduces in polynomial time an input instance to an equivalent instance whose size is bounded by a function only depending on some problem-specific parameter k; this new instance is called problem kernel. Typically, problem kernels are achieved by performing efficient data reduction rules. So far, there was little systematic study in the literature concerning the mutual interaction of data reduction rules, in particular whether data reduction rules for a specific problem always lead to the same reduced instance, no matter in which order the rules are applied. This corresponds to the concept of confluence from the theory of rewriting systems. We argue that it is valuable to study whether a kernelization is confluent, using the NP-hard graph problems (EDGE) CLIQUE COVER and PARTIAL CLIQUE COVER as running examples. We apply the concept of critical pair analysis from graph transformation theory, supported by the AGG software tool. These results support the main goal of our work, namely, to establish a fruitful link between (parameterized) algorithmics and graph transformation theory, two so far unrelated fields.
