wolffd@0: function engine = quickscore_inf_engine(inhibit, leak, prior)
wolffd@0: % QUICKSCORE_INF_ENGINE Exact inference for the QMR network
wolffd@0: % engine = quickscore_inf_engine(inhibit, leak, prior)
wolffd@0: %
wolffd@0: % We create an inference engine for QMR-like networks.
wolffd@0: % QMR is a bipartite graph, where the top layer contains hidden disease nodes,
wolffd@0: % and the bottom later contains observed finding nodes.
wolffd@0: % The diseases have Bernoulli CPDs, the findings noisy-or CPDs.
wolffd@0: % The original QMR (Quick Medical Reference) network has specific parameter values which we are not
wolffd@0: % allowed to release, for commercial reasons.
wolffd@0: %
wolffd@0: % inhibit(f,d) = inhibition probability on f->d arc for disease d, finding f
wolffd@0: % If inhibit(f,d) = 1, there is effectively no arc from d->f
wolffd@0: % leak(j) = inhibition prob. on leak node -> finding j arc
wolffd@0: % prior(i) = prob. disease i is on
wolffd@0: %
wolffd@0: % We use exact inference, which takes O(2^P) time, where P is the number of positive findings.
wolffd@0: % For details, see
wolffd@0: % - Heckerman, "A tractable inference algorithm for diagnosing multiple diseases", UAI 89.
wolffd@0: % - Rish and Dechter, "On the impact of causal independence", UCI tech report, 1998.
wolffd@0: % Note that this algorithm is numerically unstable, since it adds a large number of positive and
wolffd@0: % negative terms and hopes that some of them exactly cancel.
wolffd@0: %
wolffd@0: % For an interesting variational approximation, see
wolffd@0: % - Jaakkola and Jordan, "Variational probabilistic inference and the QMR-DT network", JAIR 10, 1999.
wolffd@0: %
wolffd@0: % See also 
wolffd@0: % - "Loopy belief propagation for approximate inference: an empirical study",
wolffd@0: %      K. Murphy, Y. Weiss and M. Jordan, UAI 99.
wolffd@0: 
wolffd@0: engine.inhibit = inhibit;
wolffd@0: engine.leak = leak;
wolffd@0: engine.prior = prior;
wolffd@0: 
wolffd@0: % store results here between enter_evidence and marginal_nodes
wolffd@0: engine.post = [];
wolffd@0: 
wolffd@0: engine = class(engine, 'quickscore_inf_engine'); % not a child of the inf_engine class!