Exact Loopy Belief Propagation on Euler Graphs
نویسنده
چکیده
Belief propagation is widely used in inference of graphical models. It yields exact solutions when the underlying graph is singly connected. When the graph contains loops, double-counting of evidence may degrade the accuracy of converged beliefs or even prohibit the convergence of messages. In this paper, we propose an exact belief propagation algorithm on Euler graphs. An Euler graph contains paths (Euler paths) which traverse the entire graph by visiting each edge exactly once. By exploiting this property, messages are propagated along an Euler path without double counting local evidence. Unlike standard belief propagation, it postpones marginalization over a variable until all edges incident to the variable have been traversed along the Euler path. The primary bottleneck of the algorithm is the marginalization over multiple variables which may have exponential time complexity. We propose two conditions when it is tractable. If O(log n) nodes have loops and each node has loop size O(c), then the marginalization of almost every variable requires polynomial time. Furthermore, if all nodes have loop size no greater than O(log n), then only O(log n) variables require marginalization over an exponential number of configurations. Empirical results demonstrate the Euler belief propagation can efficiently evaluate the exact marginal probabilities when the graph has a clustered structure.
منابع مشابه
Exact and Approximate Inference for Annotating Graphs with Structural SVMs
Training processes of structured prediction models such as structural SVMs involve frequent computations of the maximum-aposteriori (MAP) prediction given a parameterized model. For specific output structures such as sequences or trees, MAP estimates can be computed efficiently by dynamic programming algorithms such as the Viterbi algorithm and the CKY parser. However, when the output structure...
متن کاملBelief propagation algorithm for computing correlation functions in finite-temperature quantum many-body systems on loopy graphs
Belief propagation—a powerful heuristic method to solve inference problems involving a large number of random variables—was recently generalized to quantum theory. Like its classical counterpart, this algorithm is exact on trees when the appropriate independence conditions are met and is expected to provide reliable approximations when operated on loopy graphs. In this paper, we benchmark the p...
متن کاملCorrectness of Belief Propagation in Gaussian Graphical Models of Arbitrary Topology Correctness of Belief Propagation in Gaussian Graphical Models of Arbitrary Topology
Graphical models, such as Bayesian networks and Markov random elds represent statistical dependencies of variables by a graph. Local \belief propagation" rules of the sort proposed by Pearl [20] are guaranteed to converge to the correct posterior probabilities in singly connected graphs. Recently good performance has been obtained by using these same rules on graphs with loops, a method known a...
متن کاملAn empirical analysis of loopy belief propagation in three topologies: grids, small-world networks and random graphs
Recently, much research has been devoted to the study of loopy belief propagation algorithm. However, little attention has been paid to the change of its behavior in relation with the problem graph topology. In this paper we empirically study the behavior of loopy belief propagation on different network topologies which include grids, small-world networks and random graphs. In our experiments, ...
متن کامل