Incremental integration of probabilistic models learned from data
Document Type
Conference Proceeding
Publication Date
12-1-2007
Abstract
This paper addresses multi-source information integration in stochastic environments where information from sources consists of probabilistic domain models (represented as joint distributions or Bayesian networks) learned from data. We extend the batch algorithm proposed by Maynard-Reid II and Chajewska [9] to accommodate incremental integration so as to support 'anytime' querying. Experimental results verify that our algorithms compare well with the batch algorithm in accuracy and efficiency. Extensions for integrating joint distributions are independent of the order in which sources arrive, but the Bayesian network integration extension is only approximately so. This is due to bias introduced by the algorithm's use of heuristic optimization, and an 'inertial' effect that makes this bias difficult to undo over time. © 2007 IEEE.
Publication Source (Journal or Book title)
Proceedings - IEEE International Conference on Data Mining, ICDM
First Page
519
Last Page
524
Recommended Citation
Xu, J., Maynard-Zhang, P., & Chen, J. (2007). Incremental integration of probabilistic models learned from data. Proceedings - IEEE International Conference on Data Mining, ICDM, 519-524. https://doi.org/10.1109/ICDMW.2007.16