Fast algorithms for special instances of hard computational problems are essential to many computer applications. In this project we focus on combinatorial problems from data mining and automatic inference, most of them originating from bioinformatics. Fast computations are facilitated by the small size of certain input parameters. Building upon our recently published results we pursue the following research directions: speed up parameterized algorithms by an input separation technique; develop a systematic theory of compact solution space descriptions and their complexities, that unifies existing ad-hoc results; develop a repository of state-of-the-art group testing strategies for realistic input parameter values; and, as a concrete application, further develop our combinatorial approach to protein mixture inference. The first point is basic research in algorithms and problem complexity. Compact solution space descriptions are relevant if all solutions to a combinatorial problem are needed (e.g., all possible interpretations of a data set) but their sheer number prohibits explicit lists. In nonadaptive group testing, which is a classical problem in chemical testing, we aim at combining own and others´ theoretical results to a tool that outputs the best known test strategy given a set of parameters. We also work with protein digestion simulation data to evaluate the practicality of our inference algorithms developed in an established theoretical framework.
Biträdande professor vid Chalmers, Computer Science and Engineering (Chalmers), Computing Science (Chalmers)
Funding Chalmers participation during 2011–2013