Relational Learning Algorithm

A Relational Learning Algorithm is a learning algorithm that can be implemented by a Relational Learning System (to solve a relational learning task - a learning task with a structured dataset).

• Context:
  • output: a Relational Predictive Model, such as a relational classifier or a relational estimator.
  • It can range from being a Non-Statistical Relational Learning Algorithm to being a Statistical Relational Learning Algorithm.
  • It can make use of a Relational Similarity Function.
  • It can be the focus of a Relational Learning Research Question.
• Example(s):
  • FOIL Algorithm.
  • a Collective Classification Algorithm.
  • a Collective Regression Algorithm.
  • a Graph Mining Algorithm.
  • ...
• Counter-Example(s):
  • a Linear Regression Algorithm.
  • a Tuple-based Learning Algorithm.
• See: Multi-Relational Learning Algorithm.

References

2016

• (Nickel et al., 2016) ⇒ Maximilian Nickel, Kevin Murphy, Volker Tresp, and Evgeniy Gabrilovich. (2016). "A Review of Relational Machine Learning for Knowledge Graphs.” In: Proceedings of the IEEE, 104(1). doi:10.1109/JPROC.2015.2483592
  • QUOTE: Relational machine learning studies methods for the statistical analysis of relational, or graph-structured, data. In this paper, we provide a review of how such statistical models can be "trained" on large knowledge graphs, and then used to predict new facts about the world (which is equivalent to predicting new edges in the graph). In particular, we discuss two fundamentally different kinds of statistical relational models, both of which can scale to massive data sets. The first is based on latent feature models such as tensor factorization and multiway neural networks.

1999

• (Friedman et al., 1999) ⇒ Nir Friedman, Lise Getoor, Daphne Koller, and Avi Pfeffer. (1999). "Learning Probabilistic Relational Models." In: Proceedings of IJCAI Conference (IJCAI 1999).

• Concept