Music Recommendations in Hyperbolic Space: An Application of Empirical Bayes and Hierarchical Poincar\'e Embeddings

TL;DR

This paper introduces a novel hyperbolic embedding method for music recommendation hierarchies, utilizing empirical Bayes for link reliability, resulting in improved personalized playlist performance over traditional Euclidean models.

Contribution

It presents a new hyperbolic embedding approach for music hierarchies combined with empirical Bayes estimation, enhancing recommendation accuracy.

Findings

Significant performance improvement in A/B testing

Hyperbolic space better captures music hierarchy structures

Empirical Bayes improves link reliability estimation

Abstract

Matrix Factorization (MF) is a common method for generating recommendations, where the proximity of entities like users or items in the embedded space indicates their similarity to one another. Though almost all applications implicitly use a Euclidean embedding space to represent two entity types, recent work has suggested that a hyperbolic Poincar\'e ball may be more well suited to representing multiple entity types, and in particular, hierarchies. We describe a novel method to embed a hierarchy of related music entities in hyperbolic space. We also describe how a parametric empirical Bayes approach can be used to estimate link reliability between entities in the hierarchy. Applying these methods together to build personalized playlists for users in a digital music service yielded a large and statistically significant increase in performance during an A/B test, as compared to the Euclidean model.