# Adversarial Mahalanobis Distance-based Attentive Song Recommender for Automatic Playlist Continuation

## TL;DR

This paper introduces three novel Mahalanobis distance-based deep learning models for automatic playlist continuation, improving recommendation accuracy by modeling complex interactions and incorporating adversarial training.

## Contribution

The paper proposes three innovative Mahalanobis distance-based approaches and a unified model, enhancing playlist recommendation by capturing complex interactions and robustness.

## Key findings

- Outperforms eight state-of-the-art models on large datasets
- Utilizes Mahalanobis distance for better similarity measurement
- Incorporates adversarial training for robustness

## Abstract

In this paper, we aim to solve the automatic playlist continuation (APC) problem by modeling complex interactions among users, playlists, and songs using only their interaction data. Prior methods mainly rely on dot product to account for similarities, which is not ideal as dot product is not metric learning, so it does not convey the important inequality property. Based on this observation, we propose three novel deep learning approaches that utilize Mahalanobis distance. Our first approach uses user-playlist-song interactions, and combines Mahalanobis distance scores between (i) a target user and a target song, and (ii) between a target playlist and the target song to account for both the user's preference and the playlist's theme. Our second approach measures song-song similarities by considering Mahalanobis distance scores between the target song and each member song (i.e., existing song) in the target playlist. The contribution of each distance score is measured by our proposed memory metric-based attention mechanism. In the third approach, we fuse the two previous models into a unified model to further enhance their performance. In addition, we adopt and customize Adversarial Personalized Ranking (APR) for our three approaches to further improve their robustness and predictive capabilities. Through extensive experiments, we show that our proposed models outperform eight state-of-the-art models in two large-scale real-world datasets.