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 Poincaré 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.

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is to improve the performance of music recommendation systems, especially for complex hierarchical structures containing multiple entity types (such as artists, radio stations, tracks, genres, etc.). Traditional Matrix Factorization (MF) methods have limitations when dealing with the hierarchical relationships of such multiple entity types because they are usually embedded in Euclidean space, and Euclidean space is not efficient enough for representing hierarchical relationships. The paper proposes a new method. By using the Poincaré sphere (a high - dimensional hyperbolic space) to embed these hierarchical relationships and combining the parametric empirical Bayes method to estimate the link reliability between entities, a personalized playlist is constructed. This method shows a performance improvement significantly better than the traditional Euclidean model in the A/B test. Specifically, the paper mainly solves the following problems: 1. **Multi - entity type embedding**: How to effectively embed multiple different types of entities in a high - dimensional hyperbolic space, especially when there are complex hierarchical relationships between these entities. 2. **Link reliability estimation**: How to reliably estimate the link strength between entities using the parametric empirical Bayes method, especially in the case of sparse data. 3. **Personalized recommendation**: How to use the above - mentioned embedding and link estimation methods to generate high - quality personalized playlists, thereby increasing the Average Listening Time (ALT) of users. By solving these problems, the paper shows that in practical applications, the hyperbolic space embedding method can significantly improve the performance of the recommendation system, especially when dealing with data sets with complex hierarchical structures.

Music Recommendations in Hyperbolic Space: An Application of Empirical Bayes and Hierarchical Poincaré Embeddings

Recommending More Suitable Music Based on Users' Real Context.

Music Recommendation Via Heterogeneous Information Graph Embedding.

Learning to Embed Music and Metadata for Context-Aware Music Recommendation

Learning Music Embedding with Metadata for Context Aware Recommendation

Music Recommendation via Hypergraph Embedding

Music recommendation by unified hypergraph: combining social media information and music content.

Advancing Cultural Inclusivity: Optimizing Embedding Spaces for Balanced Music Recommendations

Music recommendation using dynamic feedback and content-based filtering

Where are we in embedding spaces? A Comprehensive Analysis on Network Embedding Approaches for Recommender Systems

Using Rich Social Media Information for Music Recommendation Via Hypergraph Model

Of Spiky SVDs and Music Recommendation

Modeling of the Latent Embedding of Music using Deep Neural Network

HyperML: A Boosting Metric Learning Approach in Hyperbolic Space for Recommender Systems

Sequence-based Context-Aware Music Recommendation

Leveraging the structure of musical preference in content-aware music recommendation

Music Recommendation on Spotify using Deep Learning

Learning Recommendation Algorithm Based on Improved BP Neural Network in Music Marketing Strategy

Context-Based Music Recommendation Algorithm Evaluation

Against Filter Bubbles: Diversified Music Recommendation via Weighted Hypergraph Embedding Learning

CAME: Content- and Context-Aware Music Embedding for Recommendation