Cutset conditioning and clique-tree propagation are two popular methods for performing exact probabilistic inference in Bayesian belief networks. Cutset conditioning is based on decomposition of a subset of network nodes, whereas clique-tree propagation depends on aggregation of nodes. We describe a means to combine cutset conditioning and clique- tree propagation in an approach called aggregation after decomposition (AD). We discuss the application of the AD method in the Pathfinder system, a medical expert system that offers assistance with diagnosis in hematopathology.

Two different approaches to dealing with probabilistic knowledge are examined -models and inductive inference. Examples of the first are: influence diagrams [1], Bayesian networks [2], log-linear models [3, 4]. Examples of the second are: games-against nature [5, 6] varieties of maximum-entropy methods [7, 8, 9], and the author's min-score induction [10]. In the modeling approach, the basic issue is manageability, with respect to data elicitation and computation. Thus, it is assumed that the pertinent set of users in some sense knows the relevant probabilities, and the problem is to format that knowledge in a way that is convenient to input and store and that allows computation of the answers to current questions in an expeditious fashion. The basic issue for the inductive approach appears at first sight to be very different. In this approach it is presumed that the relevant probabilities are only partially known, and the problem is to extend that incomplete information in a reasonable way to answer current questions. Clearly, this approach requires that some form of induction be invoked. Of course, manageability is an important additional concern. Despite their seeming differences, the two approaches have a fair amount in common, especially with respect to the structural framework they employ. Roughly speaking, this framework involves identifying clusters of variables which strongly interact, establishing marginal probability distributions on the clusters, and extending the subdistributions to a more complete distribution, usually via a product formalism. The product extension is justified on the modeling approach in terms of assumed conditional independence; in the inductive approach the product form arises from an inductive rule.

In this paper we explore representations of temporal knowledge based upon the formalism of Causal Probabilistic Networks (CPNs). Two different ?continuous-time? representations are proposed. In the first, the CPN includes variables representing ?event-occurrence times?, possibly on different time scales, and variables representing the ?state? of the system at these times. In the second, the CPN describes the influences between random variables with values in () representing dates, i.e. time-points associated with the occurrence of relevant events. However, structuring a system of inter-related dates as a network where all links commit to a single specific notion of cause and effect is in general far from trivial and leads to severe difficulties. We claim that we should recognize explicitly different kinds of relation between dates, such as ?cause?, ?inhibition?, ?competition?, etc., and propose a method whereby these relations are coherently embedded in a CPN using additional auxiliary nodes corresponding to "instrumental" variables. Also discussed, though not covered in detail, is the topic concerning how the quantitative specifications to be inserted in a temporal CPN can be learned from specific data.

In previous work (Fertig and Breese, 1989; Fertig and Breese, 1990) we defined a mechanism for performing probabilistic reasoning in influence diagrams using interval rather than point-valued probabilities. In this paper we extend these procedures to incorporate decision nodes and interval-valued value functions in the diagram. We derive the procedures for chance node removal (calculating expected value) and decision node removal (optimization) in influence diagrams where lower bounds on probabilities are stored at each chance node and interval bounds are stored on the value function associated with the diagram's value node. The output of the algorithm are a set of admissible alternatives for each decision variable and a set of bounds on expected value based on the imprecision in the input. The procedure can be viewed as an approximation to a full e-dimensional sensitivity analysis where n are the number of imprecise probability distributions in the input. We show the transformations are optimal and sound. The performance of the algorithm on an influence diagrams is investigated and compared to an exact algorithm.

The most difficult task in probabilistic reasoning may be handling directed cycles in belief networks. To the best knowledge of this author, there is no serious discussion of this problem at all in the literature of probabilistic reasoning so far.

We introduce a new heuristic algorithm for the problem of finding minimum size loop cutsets in multiply connected belief networks. We compare this algorithm to that proposed in [Suemmondt and Cooper, 1988]. We provide lower bounds on the performance of these algorithms with respect to one another and with respect to optimal. We demonstrate that no heuristic algorithm for this problem cam be guaranteed to produce loop cutsets within a constant difference from optimal. We discuss experimental results based on randomly generated networks, and discuss future work and open questions.

In this paper, optimum decomposition of belief networks is discussed. Some methods of decomposition are examined and a new method - the method of Minimum Total Number of States (MTNS) - is proposed. The problem of optimum belief network decomposition under our framework, as under all the other frameworks, is shown to be NP-hard. According to the computational complexity analysis, an algorithm of belief network decomposition is proposed in (Wee, 1990a) based on simulated annealing.

Within diagnostic reasoning there have been a number of proposed definitions of a diagnosis, and thus of the most likely diagnosis, including most probable posterior hypothesis, most probable interpretation, most probable covering hypothesis, etc. Most of these approaches assume that the most likely diagnosis must be computed, and that a definition of what should be computed can be made a priori, independent of what the diagnosis is used for. We argue that the diagnostic problem, as currently posed, is incomplete: it does not consider how the diagnosis is to be used, or the utility associated with the treatment of the abnormalities. In this paper we analyze several well-known definitions of diagnosis, showing that the different definitions of the most likely diagnosis have different qualitative meanings, even given the same input data. We argue that the most appropriate definition of (optimal) diagnosis needs to take into account the utility of outcomes and what the diagnosis is used for.

We show an approach to automated control of machine vision systems based on incremental creation and evaluation of a particular family of influence diagrams that represent hypotheses of imagery interpretation and possible subsequent processing decisions. In our approach, model-based machine vision techniques are integrated with hierarchical Bayesian inference to provide a framework for representing and matching instances of objects and relationships in imagery and for accruing probabilities to rank order conflicting scene interpretations. We extend a result of Tatman and Shachter to show that the sequence of processing decisions derived from evaluating the diagrams at each stage is the same as the sequence that would have been derived by evaluating the final influence diagram that contains all random variables created during the run of the vision system.

Decision tree induction systems are being used for knowledge acquisition in noisy domains. This paper develops a subjective Bayesian interpretation of the task tackled by these systems and the heuristic methods they use. It is argued that decision tree systems implicitly incorporate a prior belief that the simpler (in terms of decision tree complexity) of two hypotheses be preferred, all else being equal, and that they perform a greedy search of the space of decision rules to find one in which there is strong posterior belief. A number of improvements to these systems are then suggested.