Inference propagation engine for belief networks.

Due to significant limitations of rule-based extensional decision-support systems researchers are looking for new theories, methods and semantics to efficiently encode causality. Artificial Intelligence community demonstrates significant interest for the approaches based on theory of probability. Graphical model approach offers significant benefits and leans on sound theoretical basement. Paper discusses benefits of Intentional (declarative or model based) vs. Extensional (rule-based or production rules) approaches. Probability Propagation in Trees of Clusters (PPTC) algorithm is one of the most efficient algorithm inspired by generalized distributive law. Paper focuses on details of this recently adapted algorithm. Applet written in Java culminates the research. Algorithm implementation as an applet opens new horizons of system use, since Java™ is supported by nearly all platforms nowadays. Optimized inference propagation algorithm was devised based on high-level description of algorithm, making the applet highly fit for real life applications. Object oriented language implementation is beneficial over other approaches, since it makes package reuse simple and handy. Graphical user interface was designed with idea of maximal ease of the applet use. Program can run both in applet mode and as a standalone application. Autonomous on-board computers, enabling intelligence for mobile devices, could run the inference propagation engine which is essential part of newly created software