Iterative Classification in Relational Data

By Jennifer Neville, David D. Jensen. 2000

411 citations recorded by the source corpus.

How researchers see this paper

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

In Natural Language Processing

Citations in the corpus

DeepWalk: online learning of social representations (2014, methodology): “Relational classification (or collective classification) methods [15, 25, 32] use links between data items as part of the”
Column Networks for Collective Classification (2016, methodology): “For tractable learning, we often resort to surrogate loss functions such as (structured) pseudo-likelihood (Sutton and McCallum 2007), approximate gradient (Hinton 2002), or iterative schemes, stacked learning (Choetkiertikul et al. 2015; Kou and Cohen 2007; Macskassy and Provost 2007; Neville and J”
Graph Embedding Techniques, Applications, and Performance: A Survey (2017, methodology): “Feature-based models [11], [12], [56] generate features for nodes based on their neighborhood and local network statistics and then apply a classifier like Logistic Regression [57] and Naive Bayes [58] to predict the labels.”
Why collective inference improves relational classification (2004, background): “[1] √ √ √ √ Slattery & Mitchell [14] √ √ Neville & Jensen [10] √ √ Taskar et al.”
Mining newsgroups using networks arising from social behavior (2003, methodology): “The proposal in [ 22 ] used both the text as well as the link structure.”
Graph based anomaly detection and description: a survey (2014, methodology): “[Neville and Jensen, 2000] also use a Naive Bayes model for the attributes, but they use an iterative classification algorithm (ICA) for inference.”
Relational Dependency Networks (2007, methodology): “Examples of this technique include iterative classification (Neville and Jensen, 2000), link-based classification (Lu and Getoor, 2003), and probabilistic relational neighbor models (Macskassy and Provost, 2003, 2004).”
Beyond Homophily in Graph Neural Networks: Current Limitations and Effective Designs (2020, methodology): “[14] D.”
Linkage and Autocorrelation Cause Feature Selection Bias in Relational Learning (2002, background): “Relational autocorrelation represents an extremely important type of knowledge about relational data, one that is just beginning to be explored and exploited for learning statistical models of relational data (Neville and Jensen 2000; Slattery and Mitchell 2000).”
Graph Neural Networks with Heterophily (2020, methodology): “We note that our empirical analysis also conﬁrms the small time complexity overhead of CPGNN: on the Squirrel dataset, the runtimes of CPGNN-MLP-1 and CPGNN-Cheby-1 are 39s and 697s, respectively, while the prior belief estimators, MLP and GCN-Cheby, run in 29s and 592s in our implementation.”
Discriminative Probabilistic Models for Relational Data (2002, methodology): “Belief Propagation (BP) is a local message passing algorithm introduced by Pearl (1988). It is guaranteed to converge to the correct marginal probabilities for each node only for singly connected Markov networks.”
Classification in Networked Data: A Toolkit and a Univariate Case Study (2007, positive): “Iterative classiﬁcation has been used for collective inference elsewhere as well, for example Neville and Jensen (2000) use it in combination with naive Bayes for local and relational classiﬁcation (with a simulated annealing procedure to settle on the ﬁnal labeling).”
Local Modeling of Attributed Graphs: Algorithms and Applications (2016, methodology): “Traditional approaches to relational classification pose the problem as an inference in an undirected Markov network, and then use iterative approximate inference algorithms (such as the iterative classification algorithm [102], Gibbs Sampling [52], or label relaxation [65]) to compute the posterior”
Link Prediction in Relational Data (2003, background): “We could, for example, build a link predictor that considers other links in the graph by converting graph features into flat features [11], as we did in the social network data.”
Mining the network value of customers (2001, methodology): “Neville and Jensen [32] proposed a simple iterative algorithm for labeling nodes in social networks, based on the naive Bayes classifier.”