Relational Boosted Bandits

TL;DR

Relational Boosted Bandits (RB2) is a novel algorithm designed for relational domains, enabling interpretable and effective decision-making in social network-like environments by leveraging relational boosted trees.

Contribution

Introduces RB2, a relational contextual bandit algorithm based on boosted trees, addressing the limitations of attribute-based contexts in relational domains.

Findings

RB2 outperforms traditional bandit algorithms in relational tasks.

RB2 provides interpretable models for link prediction and recommendations.

Empirical results demonstrate RB2's effectiveness in relational settings.

Abstract

Contextual bandits algorithms have become essential in real-world user interaction problems in recent years. However, these algorithms rely on context as attribute value representation, which makes them unfeasible for real-world domains like social networks are inherently relational. We propose Relational Boosted Bandits(RB2), acontextual bandits algorithm for relational domains based on (relational) boosted trees. RB2 enables us to learn interpretable and explainable models due to the more descriptive nature of the relational representation. We empirically demonstrate the effectiveness and interpretability of RB2 on tasks such as link prediction, relational classification, and recommendations.