Propositional logic provides a convenient language for expressing real-world originated problems such as AI planning, automated test pattern generation, bounded model checking and cryptanalysis. The interest in SAT solving has increased mainly due to improvements in the solving algorithms, which recently have increasingly focused on using parallelism offered by multi-CPU computers.
Partly orthogonally to these improvements this work studies several novel approaches to parallel solving of SAT instances in a grid of widely distributed "virtual" computers instead of workstations or supercomputers.
The doctoral dissertation of Licentiate of Science (Technology) Antti Hyvärinen examines the solving of hard structured problems using cloud computing. According to Hyvärinen, cloud computing can significantly speed up problem solving and save energy. Cloud computing means decentralising IT services so that tens, hundreds or thousands of distributed computers can be used simultaneously.
Hyvärinen says that solving practical problems often requires going through large amounts of data efficiently and performing automatic inference based on the material. For instance, when the link between certain genes and the onset of a disease is studied, the material from which correlations should be found can be extremely extensive.
− Cloud computing speeds up solving a problem because it is possible to use thousands of computers instead of just one. Instead of spending ten years looking for a solution, the computers may solve the problem in a matter of hours.
Hyvärinen explains that decentralising computers also saves energy.
− Computers produce a lot of heat and there is a major demand for computing capacity in large cities in the south. Due to the warm climate in these areas, cooling the heat produced by the computers consumes a lot of electricity.
Speed requires parallel computing
Hyvärinen says that the speed of computer processors, which execute program commands, will no longer increase significantly. If we want faster results in the future, all computing should be performed in parallel by several computers.
However, parallel programming is difficult. The dissertation presents several analytical and experimental results that offer solutions to the problems related to parallel processes. These new methods can be used to solve several previously unsolved problems for the first time. In practice, certain algorithms have been developed so that they can more efficiently make use of several processors simultaneously, decreasing the computational time.
Hyvärinen uses a propositional logic to model structured problems. Due to its general nature, this logic enables the modelling of different types of problems from bioinformatics to artificial intelligence design.
The doctoral dissertation of Antti Hyvärinen, “Grid Based Propositional Satisfiability Solving”, will be examined at the School of Science Department of Information and Computer Science on 28 November 2011 at noon (lecture hall T2, Konemiehentie 2, Espoo).The dissertation is available online at: http://lib.tkk.fi/Diss/
Terhi Arvela | alfa
Construction of practical quantum computers radically simplified
05.12.2016 | University of Sussex
UT professor develops algorithm to improve online mapping of disaster areas
29.11.2016 | University of Tennessee at Knoxville
16.11.2016 | Event News
01.11.2016 | Event News
14.10.2016 | Event News
09.12.2016 | Life Sciences
09.12.2016 | Ecology, The Environment and Conservation
09.12.2016 | Health and Medicine