What is Inference

Inference is a term that does not have a consistent definition. To many statisticians, inference includes all aspects of model formulation, including choosing a model, choosing a neighborhood structure, estimating the corresponding model parameters, and the final step of inferring the class label of a sample inference. In the areas of probabilistic graphical models and machine learning, inference is defined as answering queries using a probabilistic model of the problem domain. That is, inference refers to computing the probability distribution over the variables of interest (the query) given the observed values of other random variables (the evidence). Marginal inference then refers to computing the distribution of a subset of variables. More specifically, given a graphical model, we often wish to answer probabilistic queries of the form P( y | x ), where y is the “label” variable and x is the “input”' variables.

Conditional Random Fields for Modeling Structured Data

Tom Burr (Statistical Sciences, Los Alamos National Laboratory, USA) and Alexei Skurikhin (Space Data Systems, Los Alamos National Laboratory, USA)