# End-to-end learning for music audio tagging at scale

arXiv: https://arxiv.org/abs/1711.02520 · HTML: https://mlsanity.com/paper/1711.02520/

## Key facts

- Cited by 25 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 7 Nov 2017.

## Built on by (most cited first, 25 of 25)

- [Contrastive Learning of Musical Representations](https://mlsanity.com/paper/2103.09410/) (2021)
- [Supervised and Unsupervised Learning of Audio Representations for Music Understanding](https://mlsanity.com/paper/2210.03799/) (2022)
- [MusCaps: Generating Captions for Music Audio](https://mlsanity.com/paper/2104.11984/) (2021)
- [HumBugDB: A Large-scale Acoustic Mosquito Dataset](https://mlsanity.com/paper/2110.07607/) (2021)
- [Explainability in Music Recommender Systems](https://mlsanity.com/paper/2201.10528/) (2022)
- [A Multimodal Single-Branch Embedding Network for Recommendation in Cold-Start and Missing Modality Scenarios](https://mlsanity.com/paper/2409.17864/) (2024)
- [Tracing Back Music Emotion Predictions to Sound Sources and Intuitive Perceptual Qualities](https://mlsanity.com/paper/2106.07787/) (2021)
- [EfficientLEAF: A Faster LEarnable Audio Frontend of Questionable Use](https://mlsanity.com/paper/2207.05508/) (2022)
- [Perceptual Musical Features for Interpretable Audio Tagging](https://mlsanity.com/paper/2312.11234/) (2023)
- [Learning Music Sequence Representation from Text Supervision](https://mlsanity.com/paper/2305.19602/) (2023)
- [From West to East: Who can understand the music of the others better?](https://mlsanity.com/paper/2307.09795/) (2023)
- [Enhancing Affective Representations of Music-Induced EEG through Multimodal Supervision and latent Domain Adaptation](https://mlsanity.com/paper/2202.09750/) (2022)
- [Adversarial Audio Synthesis with Complex-valued Polynomial Networks](https://mlsanity.com/paper/2206.06811/) (2022)
- [Supervised Chorus Detection for Popular Music Using Convolutional Neural Network and Multi-task Learning](https://mlsanity.com/paper/2103.14253/) (2021)
- [Music Auto-Tagging with Robust Music Representation Learned via Domain Adversarial Training](https://mlsanity.com/paper/2401.15323/) (2024)
- [Towards Training Music Taggers on Synthetic Data](https://mlsanity.com/paper/2407.02156/) (2024)
- [MuSLCAT: Multi-Scale Multi-Level Convolutional Attention Transformer for Discriminative Music Modeling on Raw Waveforms](https://mlsanity.com/paper/2104.02309/) (2021)
- [A Modulation Front-End for Music Audio Tagging](https://mlsanity.com/paper/2105.11836/) (2021)
- [Learning Sparse Analytic Filters for Piano Transcription](https://mlsanity.com/paper/2108.10382/) (2021)
- [Multi-task Learning with Metadata for Music Mood Classification](https://mlsanity.com/paper/2110.04765/) (2021)
- [End-to-End Audio Strikes Back: Boosting Augmentations Towards An Efficient Audio Classification Network](https://mlsanity.com/paper/2204.11479/) (2022)
- [Music Era Recognition Using Supervised Contrastive Learning and Artist Information](https://mlsanity.com/paper/2407.05368/) (2024)
- [Music Genre Classification using Large Language Models](https://mlsanity.com/paper/2410.08321/) (2024)
- [Linear Complexity Self-Supervised Learning for Music Understanding with Random Quantizer](https://mlsanity.com/paper/2601.09603/) (2026)
- [Constructing Composite Features for Interpretable Music-Tagging](https://mlsanity.com/paper/2603.28644/) (2026)

## Abstract (arXiv)

The lack of data tends to limit the outcomes of deep learning research, particularly when dealing with end-to-end learning stacks processing raw data such as waveforms. In this study, 1.2M tracks annotated with musical labels are available to train our end-to-end models. This large amount of data allows us to unrestrictedly explore two different design paradigms for music auto-tagging: assumption-free models - using waveforms as input with very small convolutional filters; and models that rely on domain knowledge - log-mel spectrograms with a convolutional neural network designed to learn timbral and temporal features. Our work focuses on studying how these two types of deep architectures perform when datasets of variable size are available for training: the MagnaTagATune (25k songs), the Million Song Dataset (240k songs), and a private dataset of 1.2M songs. Our experiments suggest that music domain assumptions are relevant when not enough training data are available, thus showing how waveform-based models outperform spectrogram-based ones in large-scale data scenarios.