Relational Dependency Networks

By Jennifer Neville, David D. Jensen. Journal of machine learning research, 2007

349 citations recorded by the source corpus.

Recent work on graphical models for relational data has demonstrated significant improvements in classification and inference when models represent the dependencies among instances. Despite its use in conventional statistical models, the assumption of instance independence is contradicted by most relational data sets. For example, in citation data there are dependencies among the topics of a paper's references, and in genomic data there are dependencies among the functions of interacting proteins. In this paper, we present relational dependency networks (RDNs), graphical models that are capable of expressing and reasoning with such dependencies in a relational setting. We discuss RDNs in the context of relational Bayes networks and relational Markov networks and outline the relative strengths of RDNs---namely, the ability to represent cyclic dependencies, simple methods for parameter estimation, and efficient structure learning techniques. The strengths of RDNs are due to the use of pseudolikelihood learning techniques, which estimate an efficient approximation of the full joint distribution. We present learned RDNs for a number of real-world data sets and evaluate the models in a prediction context, showing that RDNs identify and exploit cyclic relational dependencies to achieve significant performance gains over conventional conditional models. In addition, we use synthetic data to explore model performance under various relational data characteristics, showing that RDN learning and inference techniques are accurate over a wide range of conditions.

View Paper: https://www.semanticscholar.org/paper/5f8aaefa3c07563cb11884f3f227bd94431544ff

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Statistical Relational Artificial Intelligence: Logic, Probability, and Computation (2016, background): “Naive ways to incorporate such rules into directed models often do not work as expected because the grounding of the statements are typically not consistent, but are sometimes interpreted in terms of the equilibrium of a Markov chain [Neville and Jensen, 2007].”
Using ghost edges for classification in sparsely labeled networks (2008, background): “However, as pointed out in [15], when the labeled data is very sparse, which is quite common in the sparsely labeled networks that we are particularly interested in, the performance of collective classification might be largely degraded due to the lack of sufficient neighbors.”
Lifted Relational Neural Networks: Efficient Learning of Latent Relational Structures (2018, methodology): “The main inspiration for the work presented in this paper are lifted graphical models such as Markov logic networks (Richardson & Domingos, 2006), Bayesian logic programs (Kersting & De Raedt, 2001) or Relational dependency networks (Neville & Jensen, 2007).”
Gradient-based boosting for statistical relational learning: The relational dependency network case (2011, methodology): “To perform inference, Neville et al., unroll the RDN to the ground network (where each predicate is instantiated with all the possible values for the variables in the predicate).”
Column Networks for Collective Classification (2016, background): “A canonical task in learning from this data type is collective classification in which networked data instances are classified simultaneously rather than independently to exploit the dependencies in the data (Macskassy and Provost 2007; Neville and Jensen 2007; Richardson and Domingos 2006; Sen et a”
Collective Classi!cation in Network Data (2008, background): “…past five to seven years, initiated by the work of Jennifer Neville and David Jensen (Neville and Jensen 2000; Jensen, Neville, and Gallagher 2004; Neville and Jensen 2007) and the work of Ben Taskar and colleagues (Taskar, Segal , and Koller 2001; Getoor et al. 2001; Taskar, Abbeel, and Koller…”
Hinge-Loss Markov Random Fields and Probabilistic Soft Logic (2015, background): “L G ] 1 7 M”
Human-Guided Learning for Probabilistic Logic Models (2018, methodology): “…we show that several formalisms such as Markov Logic Networks (MLNs) (Domingos and Lowd, 2009) and Relational Dependency Networks (RDNs) (Neville and Jensen, 2007) and tasks such as relational imitation learning (Natarajan et al., 2011) and relational transfer learning can benefit from…”
Leveraging Label-Independent Features for Classification in Sparsely Labeled Networks: An Empirical Study (2008, background): “, statistical relational learning or SRL) [3,4,5,6,7].”
A Review of Relational Machine Learning for Knowledge Graphs (2015, unknown): “, relational dependency networks [133]).”
Learning relational dependency networks in hybrid domains (2015, background): “…hand, SRL approaches such as logical Bayesian networks (LBNs) (Fierens et al. 2005), probabilistic relational models (Getoor et al. 2001), and relational dependency networks (Neville and Jensen 2007) capture both structure and uncertainty in problems but are generally restricted to discrete data.”
Structure learning for relational logistic regression: an ensemble approach (2018, background): “They have generally ranged from directed models (Kersting and De Raedt 2007; Koller 1999; Heckerman, Meek, and Koller 2007; Kazemi et al. 2014a; Neville and Jensen 2007; Kazemi and Poole 2018) to undirected models (Richardson and Domingos 2006; Taskar et al. 2007; Kimmig et al. 2012).”
Link Prediction in Complex Networks: A Survey (2010, background): “Relational Dependency Networks (RDNs) [115,116]: The RDN is a bi-directed graphical model with a set of conditional probability distributions, which can be used to represent the cyclic dependencies in a relational setting.”
Handling Missing Values when Applying Classification Models (2007, background): “…model trees (Landwehr et al., 2005), and as components of or tools for the development of graphical models such as Bayesian networks (Friedman and Goldszmidt, 1996), dependency networks (Heckerman et al., 2000), and probabilistic relational models (Getoor et al., 2002; Neville and Jensen, 2007).”
Collective Classification in Network Data (2008, background): “Even though collective classification has gained attention only in the past five to seven years, initiated by the work of Jennifer Neville and David Jensen (Neville and Jensen 2000; Jensen, Neville, and Gallagher 2004; Neville and Jensen 2007) and the work of Ben Taskar and colleagues (Taskar, Segal”
Graph Regularized Transductive Classification on Heterogeneous Information Networks (2010, background): “Generally, classification can be categorized into two groups: (1) transductive classification [10] [11] [22] [21] [19]: to predict labels for the given unlabeled data; and (2) inductive classification [9] [15] [12] [17] [3]: to construct a decision function in the whole data space.”