API introduction

The entire model is represented by the FactorSet, which is a collection of Factor objects. The FactorSet also encapsulates most of the AI algorithms available in this library.

The structure of each Factor is defined by VarSet, which is a set of variables. Logically, the VarSet for a Factor is divided into "Head" and "Tail" portions. For example, the Factor described by relationship P(D|AB) has VarSet (A,B,D), a "Tail" VarSet {A,B} and a "Head" VarSet {D}.

Factor contains the table which assigns the real value of each combination of values in VarSet. In most cases, this value is interpreted as a conditional probability P(Head Varset | Tail VarSet) for this combination, but in other cases it could have different meanings. Each VarSet combination is called a Clause. The Clause can be assigned an integer value corresponding to the raw in the Factor. This integer value is called an Instance. Instances and Clauses are interchangeable in the context of a single Factor, but can be used in different contexts to improve algorithm performance.

Also, though not shown on the picture, all Variables are entered into the Domain DB, which provides convenience mappings between Variable names and the numeric variable IDs that are used throughout the implementation.