Leveraging relational autocorrelation with latent group models
By Jennifer Neville, David D. Jensen. 2005
147 citations recorded by the source corpus.
The presence of autocorrelation provides a strong motivation for using relational learning and inference techniques. Autocorrelation is a statistical dependence between the values of the same variable on related entities and is a nearly ubiquitous characteristic of relational data sets. Recent research has explored the use of collective inference techniques to exploit this phenomenon. These techniques achieve significant performance gains by modeling observed correlations among class labels of related instances, but the models fail to capture a frequent cause of autocorrelation - the presence of underlying groups that influence the attributes on a set of entities. We propose a latent group model (LGM) for relational data, which discovers and exploits the hidden structures responsible for the observed autocorrelation among class labels. Modeling the latent group structure improves model performance, increases inference efficiency, and enhances our understanding of the datasets. We evaluate performance on three relational classification tasks and show that LGM outperforms models that ignore latent group structure, particularly when there is little information with which to seed inference.
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
- Discriminative Deep Random Walk for Network Classification (2016, background): “(Neville and Jensen, 2005) propose a latent group model for relational data, which discovers and exploits the hidden structures responsible for the observed autocorrelation among class labels.”
- Statistical predicate invention (2007, methodology): “Most of them only cluster objects, not relations (Popescul & Ungar, 2004; Wolfe & Jensen, 2004; Neville & Jensen, 2005; Xu et al., 2005; Long et al., 2006; Roy et al., 2006).”
- Scalable Learning of Collective Behavior (2012, methodology): “Hence, researchers propose to model network connections or class labels based on latent groups [20], [26].”
- Leveraging social media networks for classification (2011, methodology): “To capture the long-distance correlation, the latent group model [32] and the nonparametric infinite hidden relational model [51] are proposed.”
- Leveraging Label-Independent Features for Classification in Sparsely Labeled Networks: An Empirical Study (2008, background): “Neville and Jensen [12] use spectral clustering to group instances based on link structure and use these groups in learning classifiers.”
- Relational learning via latent social dimensions (2009, methodology): “This classifier follows the idea presented in latent group model [23].”
- metapath2vec: Scalable Representation Learning for Heterogeneous Networks (2017, background): “5 RELATEDWORK Network representation learning can be traced back to the usage of latent factor models for network analysis and graph mining tasks [10, 34], such as the application of factorization models for recommendation systems [14, 16], node classication [32], relational mining [19], and role d”
- Link mining: a survey (2005, background): “[58] D. Jensen and H. Goldberg.”
- A bias/variance decomposition for models using collective inference (2008, background): “formation to seed the inference process; latent group models (Neville and Jensen 2005) have”
- Classification in Networked Data: A Toolkit and a Univariate Case Study (2007, unknown): “Inference can be more effective if these groups are modeled explicitly, such as by using latent group models (LGMs, Neville and Jensen, 2005) to specify joint models of attributes, links, and groups .”
- Scalable learning of collective behavior based on sparse social dimensions (2009, methodology): “A similar idea has been adopted in latent group model [13] for efficient inference.”
- To join or not to join: the illusion of privacy in social networks with mixed public and private user profiles (2009, unknown): “[8] M. Hay, G. Miklau, D. Jensen, and D. Towsley.”
- Local Modeling of Attributed Graphs: Algorithms and Applications (2016, unknown): “The most relevant relational classification algorithms to our work incorporate community information by learning clusters [103], by adding edges between nearby nodes [49], by using PageRank [81], or by extending relational classification to take additional features into account [144].”