图模型的维数对于检验、模型选择和分类都至关重要。针对现有文献给出的两种离散无向图模型维数的定义是否一致的问题进行了研究,其中一种定义给出了一个显式的计算模型维数的公式。证明了对任意的离散无向图模型,两种定义下维数的差值为1,即两种定义等价,从而证明了两种定义的一致性。
Dimensions of graphical models are important for test, model selection and classification. There are two definitions of dimension for discrete undirected graphical models in literature, and one of them provides a formula for calculating the dimensions of discrete undirected graphical models. It isshown that the consistency of the two definitions for all discrete undirected graphical models.
[1] LAURITZEN S L. Graphical models[M].Oxford:OxfordClarendon Press, 1996.
[2] GEIGER D, HECKERMAN D, KING H, et al. Stratified exponential families: graphical models and model selection[J].The Annal of Statistics, 2001,29(2):505-529.
[3] HASTIE T, TIBSHIRANI R T, FRIEDMAN J. The elements of statistical learning: data mining,inference, and prediction[M].New York: Springer, 2009.
[4] GEIGER D, MEEK C, STURMFELS B. On the toric algebra of graphical models[J].The Annal of Statistics, 2006,34:1463-1492.
[5] STURMFELS B. Grobner bases and convex polytopes[M].Rhode Island: American Mathematical Society, 1996.
[6] PENA J M. Faithfulness in chain graphs: the discrete case[J].International Journal of Approximate Reasoning, 2009, 50:1306-1313.
[7] NAKAMURA A, SCHMITT M, SCHMITT N, et al. Inner product spaces for Bayesian networks[J].Journal of Machine Learning Research, 2005,6:1383-1403.
[8] VARANDO G, BIELZA C, LARRANAGA P. Decision boundary for discrete Bayesian network classifiers[J].Journal of Machine Learning Research, 2015,16:2725-2749.
[9] YANG Y, WU Y. On the properties of concept classes induced by multivalued Bayesiannetworks[J].Information Sciences, 2012,184:155-165.
[10] LI B C, YANG Y.Complexity of concept classes induced by discrete Markov networks and Bayesian networks[J]. Pattern Recognition,2018,82:31-37.
