Documentation‎ > ‎API‎ > ‎

Inference

Support for performing inference on certain variables and storing the results of inference. Inferencer can be requested to perform inference on certain variables. For example, BPInferencer and SamplingInferencer.

MCMC

Sampling-based inference. Sampler, with sampling requests optionally guided by an arbitrary "context". SettingsSampler considers an arbitrary number of value settings of a portion of the possible world. ProposalSampler represents these changes to the current possible world as aProposal object, and selects a Proposal according to its associated change in score. VariableSettingsSampler is a SettingsSampler in which the proposals are all possible settings of a given variable. MHSampler is a SettingsSampler in which one proposal is a noop and the other is defined by a (possibly user-defined) proposal function.

Belief Propagation

Marginal Inference (Sum Product)

Loopy BP

val summary = new BPSummary(variables, model)
BP.inferLoopy(summary, iterations)

Forward-Backward

val summary = BP.inferChainSum(chains, model)

Map Inference (Max Product)