Of Spiky SVDs and Music Recommendation

TL;DR

This paper explores the phenomenon of spiking formations in music recommendation embeddings derived from truncated SVD, providing a metric, theoretical insights, and practical implications for evolving music embeddings over time.

Contribution

It introduces a metric to quantify spiking formations, proves their origin related to item communities, and discusses implications for dynamic music recommendation systems.

Findings

Spiking formations are linked to item community structures.

A metric to measure spiking strength is proposed.

Implications for evolving music embeddings are analyzed.

Abstract

The truncated singular value decomposition is a widely used methodology in music recommendation for direct similar-item retrieval or embedding musical items for downstream tasks. This paper investigates a curious effect that we show naturally occurring on many recommendation datasets: spiking formations in the embedding space. We first propose a metric to quantify this spiking organization's strength, then mathematically prove its origin tied to underlying communities of items of varying internal popularity. With this new-found theoretical understanding, we finally open the topic with an industrial use case of estimating how music embeddings' top-k similar items will change over time under the addition of data.