Search Results

Knowledge compilation (KC) is a research topic which aims to investigate the possibility of circumventing the computational intractability of hard tasks, by preprocessing part of the available information, common to a number of instances. Pioneered almost three decades ago, KC is nowadays a very active research field, transversal to several areas within computer science. Among others, KC intersects knowledge representation, constraint satisfaction, algorithms, complexity theory, machine learning, and databases. The results obtained so far take various forms, from theory (compilability settings, definition of target languages for KC, complexity results, succinctness results, etc.) to more practical results (development and evaluation of compilers and other preprocessors, applications to diagnosis, planning, automatic configuration, etc.). Recently, KC has been positioned as providing a systematic method for solving problems beyond NP, and also found applications in machine learning. The goal of this Dagstuhl Seminar was to advance both aspects of KC, and to pave the way for a fruitful cross-fertilization between the topics, from theory to practice. The program included a mixture of long and short presentations, with discussions. Several long talks with a tutorial flavor introduced the participants to the variety of aspects in knowledge compilation and the diversity of techniques used. System presentations as well as an open problem session were also included in the program.