Compression of General Bayesian Net CPTs

Non-Impeding Noisy-AND NIN-AND Tree NAT models offer a highly expressive approximate representation for significantly reducing the space of Bayesian Nets BNs. They can also significantly improve efficiency of BN inference, as shown for binary NAT models. To enable these advantages for general BNs, advancements on three technical challenges are made in this work. We overcome the limitation of well-defined Pairwise Causal Interaction PCI bits and present a flexible PCI pattern extraction from general target Conditional Probability Tables CPTs. We extend parameter estimation for binary NAT models to constrained gradient descent for compressing target CPTs into multi-valued NAT models. The effectiveness of the compression is demonstrated experimentally. A novel framework is also developed for PCI pattern extraction when persistent leaky causes exist.

[1]  Y. Xiang.  Bayesian Network Inference With NIN-AND Tree Models , 2012 .

[2]  Steffen L. Lauritzen,et al.  Bayesian updating in causal probabilistic networks by local computations , 1990 .

[3]  Jirí Vomlel,et al.  Exploiting tensor rank-one decomposition in probabilistic inference , 2007, Kybernetika.

[4]  Yang Xiang,et al.  Towards Effective Elicitation of NIN-AND Tree Causal Models , 2009, SUM.

[5]  Yang Xiang,et al.  Non-impeding noisy-AND tree causal models over multi-valued variables , 2012, Int. J. Approx. Reason..

[6]  Bruce D'Ambrosio,et al.  A Factorized Representation of Independence of Causal Influence and Lazy Propagation , 1999, Int. J. Uncertain. Fuzziness Knowl. Based Syst..

[7]  Marek J. Druzdzel,et al.  Knowledge Engineering for Bayesian Networks: How Common Are Noisy-MAX Distributions in Practice? , 2006, IEEE Transactions on Systems, Man, and Cybernetics: Systems.

[8]  Yang Xiang,et al.  Compression of Bayesian Networks with NIN-AND Tree Modeling , 2014, Probabilistic Graphical Models.

[9]  Yang Xiang.  Acquisition and Computation Issues with NIN-AND Tree Models , .

[10]  Nevin Lianwen Zhang,et al.  Exploiting Causal Independence in Bayesian Network Inference , 1996, J. Artif. Intell. Res..

[11]  Judea Pearl,et al.  Probabilistic reasoning in intelligent systems - networks of plausible inference , 1991, Morgan Kaufmann series in representation and reasoning.

[12]  Bruce D'Ambrosio,et al.  Multiplicative Factorization of Noisy-Max , 1999, UAI.

[13]  Yang Xiang,et al.  Acquisition of Causal Models for Local Distributions in Bayesian Networks , 2014, IEEE Transactions on Cybernetics.

[14]  Anders L. Madsen,et al.  LAZY Propagation: A Junction Tree Inference Algorithm Based on Lazy Evaluation , 1999, Artif. Intell..