next up previous
Next: 2.2 Weight Learning Up: 2 Architecture Overview Previous: 2 Architecture Overview

2.1 Inference

All classes representing inference algorithms are subclasses of Inference. At the lowest level are the classes MaxWalkSat, UnitPropagation, BP, Gibbs Sampler, MCSAT and SimulatedTempering. The last three are subclasses of MCMC. This contains member functions and attributes which are common to all MCMC sampling methods, such as number of chains, maximum number of steps, etc.

MaxWalkSat is a subclass of SAT which abstractly represents any weighted SAT solver.

Marc Sumner 2008-09-17