Partial abductive inference in Bayesian belief networks by simulated annealing

Abductive inference in Bayesian belief networks (BBN) is intended as the pr ocess of generating the K most probable configurations given observed evide nce, When we are only interested in a subset of the network variables, this problem is called partial abductive inference, Due to the noncommutative b ehaviour of the two operators (summation and maximum) involved in the compu tational process of solving partial abductive inference in BBNs, the proces s can be unfeasible by exact computation even for networks in which other t ypes of probabilistic reasoning are not very complicated, This paper descri bes an approximate method to perform partial abductive inference in BBNs ba sed on the simulated annealing (SA) algorithm, The algorithm can be applied to multiple connected networks and for any value of K, The evaluation func tion is based on clique tree propagation, and allow to evaluate neighbour c onfigurations by means of local computations, in this way the efficiency wi th respect to previous algorithms (based on the use of genetic algorithms ( GAs)) is improved, (C) 2001 Elsevier Science Inc, All rights reserved.