A sum-product network (SPN) is a probabilistic model, based on a rooted acyclic directed graph, in which terminal nodes represent probability distributions and non-terminal nodes represent convex sums (weighted averages) and products of probability distributions. They are closely related to probabilistic graphical models, in particular to Bayesian networks with multiple context-specific independencies. Their main advantage is the possibility of building tractable models from data, i.e., models that can perform several inference tasks in time proportional to the number of edges in the graph. They are somewhat similar to neural networks and can address the same kinds of problems, such as image processing and natural language understanding. This paper offers a survey of SPNs, including their definition, the main algorithms for inference and learning from data, several applications, a brief review of software libraries, and a comparison with related models.

OpenMarkov is a Java open-source tool for creating and evaluating probabilistic graphical models, including Bayesian networks, influence diagrams, and some Markov models. With more than 100,000 lines of code, it offers some features for interactive learning, explanation of reasoning, and cost-effectiveness analysis, which are not available in any other tool. OpenMarkov has been used at universities, research centers, and large companies in more than 30 countries on four continents. Several models, some of them for real-world medical applications, built with OpenMarkov, are publicly available on Internet.

Bayesian networks (BNs) and influence diagrams (IDs) are probabilistic graphical models that are widely used for building diagnosis- and decision-support expert systems. Explanation of both the model and the reasoning is important for debugging these models, alleviating users' reluctance to accept their advice, and using them as tutoring systems. This paper describes some explanation options for BNs and IDs that have been implemented in Elvira and how they have been used for building medical models and teaching probabilistic reasoning to pre- and postgraduate students.

Díez's algorithm for the noisy MAX is very efficient for polytrees, but when the network has loops, it has to be combined with local conditioning, a suboptimal propagation algorithm. Other algorithms, based on several factorizations of the conditional probability of the noisy MAX, are not as efficient for polytrees but can be combined with general propagation algorithms such as clustering or variable elimination, which are more efficient for networks with loops. In this article we propose a new factorization of the noisy MAX that amounts to Díez's algorithm in the case of polytrees and at the same time is more efficient than previous factorizations when combined with either variable elimination or clustering. © 2003 Wiley Periodicals, Inc.