Academic
Publications
A ``Microscopic'' Study of Minimum Entropy Search in Learning Decomposable Markov Networks

A ``Microscopic'' Study of Minimum Entropy Search in Learning Decomposable Markov Networks,Machine Learning,Yang Xiang,S. K. Michael Wong,Nick Cercone

A ``Microscopic'' Study of Minimum Entropy Search in Learning Decomposable Markov Networks   (Citations: 40)
BibTex | RIS | RefWorks Download
Journal: Machine Learning - ML , vol. 26, no. 1, pp. 65-92, 1997
Cumulative Annual
View Publication
The following links allow you to view full publications. These links are maintained by other sources not affiliated with Microsoft Academic Search.
    • ...At present, scoring-search method [8, 9] is adopted in...

    Shuangcheng Wanget al. Conditional Markov Network Hybrid Classifiers Using on Client Credit S...

    • ...Graphical models such as Bayesian networks (BNs) (Pearl 1988), decomposable Markov networks (DMNs) (Xiang, Wong, & Cercone 1997), and chain graphs (Lauritzen 1996) have been applied successfully to probabilisticreasoning in intelligent systems...
    • ...Given the usefulness of graphicalmodels, one way to construct them is by discovery from data, as demonstrated by pioneer work such as (Chow & Liu 1968; Rebane & Pearl 1987; Herskovits & Cooper 1990; Fung & Crawford 1990; Spirtes & Glymour 1991; Cooper & Herskovits 1992; Lam & Bacchus 1994; Heckerman, Geiger, & Chickering 1995; Xiang, Wong, & Cercone 1997)...
    • ...For the single-linklookahead heuristic,we consider an algorithmthat scores a graphicalstructure usingthe K-L cross entropy and starts with G0. It can be shown (Xiang, Wong, & Cercone 1997) that the score KLS(G0) is...

    Yang Xiang. A Decision Theoretic View on Choosing Heuristics for Discovery of Grap...

    • ...Learning belief networks in possibly PI domains requires multi-link lookahead search (Xiang, Wong, and Cercone, 1997)...
    • ...To discover the correct dependence structure in potentially PI domains, multilink lookahead search is necessary (Xiang, Wong, and Cercone, 1997)...
    • ...The following proposition, reformatted from Xiang, Wong, and Cercone (1997), establishes a relation between the computable H (V ) and not directly computable � 2(� |M)...
    • ...Proposition 15 (Xiang, Wong & Cercone, 1997)...
    • ...The class of PI domains was identified by Xiang, Wong, and Cercone (1997) and their practical relevance was demonstrated in Xiang et al. (2000)...
    • ...Alternatively, in this paper and the related work (Xiang, Wong, and Cercone, 1997), junction forest models have been used...

    Y. Xianget al. Learning decomposable markov networks in pseudo-independent domains wi...

    • ...We just mention (Xiang et al., 1997) as a starting point...
    • ...This method has limitations in constructing the correct network for some artificial cases, as is shown in (Xiang et al., 1997)...

    Heinz Mühlenbeinet al. The Estimation of Distributions and the Minimum Relative Entropy Princ...

    • ...To learn PI models, a more sophisticated search method called multi-link lookahead (Xiang, Wong, & Cercone 1997) should be used...
    • ...A pseudo-independent (PI) model is a PDM where proper subsets of a set of collectively dependent variables display marginal independence (Xiang, Wong, and Cercone 1997)...

    Jae-hyuck Leeet al. Model Complexity of Pseudo-independent Models

Sort by: