A Proof of SPNs as Mixture of Trees
ثبت نشده
چکیده
Proof. Argue by contradiction that T is not a tree, then there must exist a node v 2 T such that v has more than one parent in T . This means that there exist at least two paths R, p1, . . . , v and R, q1, . . . , v that connect the root of S(T ), which we denote by R, and v. Let t be the last node in R, p1, . . . , v and R, q1, . . . , v such that R, . . . , t are common prefix of these two paths. By construction we know that such t must exist since these two paths start from the same root node R (R will be one candidate of such t). Also, we claim that t 6= v otherwise these two paths overlap with each other, which contradicts the assumption that v has multiple parents. This shows that these two paths can be represented as R, . . . , t, p, . . . , v and R, . . . , t, q, . . . , v where R, . . . , t are the common prefix shared by these two paths and p 6= q since t is the last common node. From the construction process defined in Def. 2, we know that both p and q are children of t in S . Recall that for each sum node in S , Def. 2 takes at most one child, hence we claim that t must be a product node, since both p and q are children of t. Then the paths that t! p v and t! q v indicate that scope(v) ✓ scope(p) ✓ scope(t) and scope(v) ✓ scope(q) ✓ scope(t), leading to ? 6= scope(v) ✓ scope(p) \ scope(q), which is a contradiction of the decomposability of the product node t. Hence as long as S is complete and decomposable, T must be a tree.
منابع مشابه
Learning Tractable Graphical Models Using Mixture of Arithmetic Circuits
In recent years, there has been a growing interest in learning tractable graphical models in which exact inference is efficient. Two main approaches are to restrict the inference complexity directly, as done by low-treewidth graphical models and arithmetic circuits (ACs), or introduce latent variables, as done by mixtures of trees, latent tree models, and sum-product networks (SPNs). In this pa...
متن کاملA Unified Approach for Learning the Parameters of Sum-Product Networks
We present a unified approach for learning the parameters of Sum-Product networks (SPNs). We prove that any complete and decomposable SPN is equivalent to a mixture of trees where each tree corresponds to a product of univariate distributions. Based on the mixture model perspective, we characterize the objective function when learning SPNs based on the maximum likelihood estimation (MLE) princi...
متن کاملApoptosis induction and proliferation inhibition by silibinin encapsulated in nanoparticles in MIA PaCa-2 cancer cells and deregulation of some miRNAs
Objective(s): Silibinin, as an herbal compound, has anti-cancer activity. Because of low solubility of silibinin in water and body fluids, it was encapsulated in polymersome nanoparticles and its effects were evaluated on pancreatic cancer cells and cancer stem cells.Materials and Methods: MIA PaCa-2 pancreatic cancer cells were treated ...
متن کاملشناسایی ساختار و فرآیندهای فرسایش ملات قیر به کاررفته در آجرهای یادمان هخامنشی تل آجری در پارسه
In this paper the bituminous water proof mortar, which has been used between the bricks in the Achaemenes monument of Tall-e-Ajori in Persepolis, has been analyzed by different instrumental chemical analysis methods, such as Fourier transform infrared spectroscopy (FTIR), X ray diffraction (XRD and X ray florescence spectroscopies (XRF). Moreover, due to identifying the chemical structure and e...
متن کاملMerging Strategies for Sum-Product Networks: From Trees to Graphs
Learning the structure of sum-product networks (SPNs) – arithmetic circuits over latent and observed variables – has been the subject of much recent research. These networks admit linear time exact inference, and thus help alleviate one of the chief disadvantages of probabilistic graphical models: accurate probabilistic inference algorithms are often computationally expensive. Although, algorit...
متن کامل