Probabilistic similarity networks

I address practical issues concerning the construction of normative expert systems, and examine the influence diagram as a potential framework for representing knowledge in such systems. I introduce an extension of the influence-diagram representation called a similarity network. A similarity network is a tool for constructing large and complex influence diagrams. The representation allows a user to construct independent influence diagrams for subsets of a given domain. A valid influence diagram for the entire domain can then be constructed from the individual diagrams. Similarity networks represent forms of conditional independence that are not represented conveniently in an ordinary influence diagram. I discuss in detail one such conditional independence, called subset independence, and examine how similarity networks exploit this form of independence to facilitate the construction of an influence diagram. Also, I describe the assessment of probability distributions for influence diagrams. I show that similarity networks exploit subset independence to simplify such probability assessments. I introduce a representation that is closely related to similarity networks, called a partition. This representation further exploits subset independence to simplify probability assessment. Finally, I examine a real-world normative expert system for the diagnosis of lymph-node pathology, called Pathfinder. The similarity-network and partition representations played a crucial role in the construction of this expert system.