Neural Networks for Relational Data

  • 2020-01-13 17:26:00
  • Navdeep Kaur, Gautam Kunapuli, Saket Joshi, Kristian Kersting, Sriraam Natarajan
  • 0

Abstract

While deep networks have been enormously successful over the last decade,they rely on flat-feature vector representations, which makes them unsuitablefor richly structured domains such as those arising in applications like socialnetwork analysis. Such domains rely on relational representations to capturecomplex relationships between entities and their attributes. Thus, we considerthe problem of learning neural networks for relational data. We distinguishourselves from current approaches that rely on expert hand-coded rules bylearning relational random-walk-based features to capture local structuralinteractions and the resulting network architecture. We further exploitparameter tying of the network weights of the resulting relational neuralnetwork, where instances of the same type share parameters. Our experimentalresults across several standard relational data sets demonstrate theeffectiveness of the proposed approach over multiple neural net baselines aswell as state-of-the-art statistical relational models.

 

Quick Read (beta)

loading the full paper ...