Stochastic Systems Group  

Discussion of survey propagation
Venkat Chandrasekaran
SSG, MIT
Survey propagation (SP) has been shown to solve difficult instances of an important class of constraint satisfaction problems. Recent work has focused on interpreting SP as an algorithm that performs belief propagation on a particular augmented factor graph derived from the factor graph of the original problem. We will briefly overview these developments, and discuss further connections between SP and inference in graphical models.
Problems with this site should be emailed to jonesb@mit.edu