Mixing ICI and CSI Models for More Efficient Probabilistic Inference

Conditional probability tables (CPTs) in Bayesian Networks (BNs) have exponential space on the family size. Local models based on independence of causal influence (ICI) or context-specific independence (CSI) have been applied separately to improve the efficiency. We propose a framework to mix both local models in the same BN for improved efficiency. In particular, we show that ICI and CSI are orthogonal, and each is unable to express the other efficiently and accurately. We propose a formalism to encode both types of local models in the same BN, and to convert it into a homogenous representation to support exact inference. We report experimental evaluation where significant efficiency gain is obtained in exact inference.

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

[2]  Nir Friedman,et al.  Learning Bayesian Networks with Local Structure , 1996, UAI.

[3]  Yang Xiang,et al.  De-Causalizing NAT-Modeled Bayesian Networks for Inference Efficiency , 2018, Canadian Conference on AI.

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

[5]  Yang Xiang,et al.  Modeling Causal Reinforcement and Undermining for Efficient CPT Elicitation , 2007, IEEE Transactions on Knowledge and Data Engineering.

[6]  Yang Xiang,et al.  Compressing Bayesian Networks: Swarm-Based Descent, Efficiency, and Posterior Accuracy , 2018, Canadian Conference on AI.

[7]  Huaiyu Zhu.  On Information and Sufficiency , 1997 .

[8]  Ralf Eggeling,et al.  Learning Bayesian networks with local structure, mixed variables, and exact algorithms , 2019, Int. J. Approx. Reason..

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

[10]  David Poole,et al.  Probabilistic Partial Evaluation: Exploiting Rule Structure in Probabilistic Inference , 1997, IJCAI.

[11]  Adnan Darwiche,et al.  Compiling Bayesian Networks Using Variable Elimination , 2007, IJCAI.

[12]  Yang Xiang,et al.  Multiplicative Factorization of Multi-Valued NIN-AND Tree Models , 2016, FLAIRS.

[13]  Max Henrion.  Practical issues in constructing a Bayes belief network , 1988, Int. J. Approx. Reason..

[14]  F. Cozman,et al.  Generalizing variable elimination in Bayesian networks , 2000 .

[15]  Yang Xiang,et al.  PROBABILISTIC REASONING IN MULTIAGENT SYSTEMS: A GRAPHICAL MODELS APPROACH, by Yang Xiang, Cambridge University Press, Cambridge, 2002, xii + 294 pp., ISBN 0-521-81308-5 (Hardback, £45.00). , 2002, Robotica.

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

[17]  Nir Friedman,et al.  Context-specific Bayesian clustering for gene expression data , 2001, J. Comput. Biol..

[18]  David Poole,et al.  Context-specific approximation in probabilistic inference , 1998, UAI.

[19]  Nir Friedman,et al.  On the Sample Complexity of Learning Bayesian Networks , 1996, UAI.

[20]  Kristian G. Olesen,et al.  HUGIN - A Shell for Building Bayesian Belief Universes for Expert Systems , 1989, IJCAI.

[21]  Marek J. Druzdzel,et al.  An Independence of Causal Interactions Model for Opposing Inuences , 2008 .

[22]  Linda C. van der Gaag,et al.  An intercausal cancellation model for Bayesian-network engineering , 2015, Int. J. Approx. Reason..

[23]  Max Henrion,et al.  Some Practical Issues in Constructing Belief Networks , 1987, UAI.

[24]  Randy Goebel,et al.  Computational intelligence - a logical approach , 1998 .

[25]  Craig Boutilier,et al.  Context-Specific Independence in Bayesian Networks , 1996, UAI.