Exact Probing of Glassy States by Survey Propagation
Author(s) -
Demian Battaglia,
Alfredo Braunstein,
Joël Chavas,
Riccardo Zecchina
Publication year - 2005
Publication title -
progress of theoretical physics supplement
Language(s) - English
Resource type - Journals
ISSN - 0375-9687
DOI - 10.1143/ptps.157.330
Subject(s) - lossy compression , forcing (mathematics) , coding (social sciences) , statistical physics , set (abstract data type) , computer science , algorithm , theoretical computer science , mathematics , statistics , physics , mathematical analysis , artificial intelligence , programming language
After briefly reviewing the theoretical set up underlying the Survey Propagation (SP) equations, we show how SP can be generalized to include external forcing - external surveys - which allow to address selectively glassy states (which may be exponentially numerous). Taking as working model the random K-SAT problem, we show that the geometrical nature of its different glassy phases can be probed efficiently. A preliminary application of this new algorithm to source coding (lossy data compression) is discussed.
Accelerating Research
Robert Robinson Avenue,
Oxford Science Park, Oxford
OX4 4GP, United Kingdom
Address
John Eccles HouseRobert Robinson Avenue,
Oxford Science Park, Oxford
OX4 4GP, United Kingdom