Morphing the Hugin and Shenoy-Shafer architectures

Abstract

The Hugin and Shenoy-Shafer architectures are two variations on the jointree algorithm, which exhibit different tradeoffs with respect to efficiency and query answering power. The Hugin architecture is more time-efficient on arbitrary jointrees, avoiding some redundant computations performed by the Shenoy-Shafer architecture. This efficiency, however, comes at the price of limiting the number of queries the Hugin architecture is capable of answering. In this paper, we present a simple algorithm which retains the efficiency of the Hugin architecture and enjoys the query answering power of the Shenoy-Shafer architecture.