LC-Protonets: Multi-Label Few-Shot Learning for World Music Audio Tagging
We introduce Label-Combination Prototypical Networks (LC-Protonets) to address the problem of multi-label few-shot classification, where a model must generalize to new classes based on only a few available examples. Extending Prototypical Networks, LC-Protonets generate one prototype per label combi...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Open Journal of Signal Processing |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10839319/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1823859590025969664 |
|---|---|
| author | Charilaos Papaioannou Emmanouil Benetos Alexandros Potamianos |
| author_facet | Charilaos Papaioannou Emmanouil Benetos Alexandros Potamianos |
| author_sort | Charilaos Papaioannou |
| collection | DOAJ |
| description | We introduce Label-Combination Prototypical Networks (LC-Protonets) to address the problem of multi-label few-shot classification, where a model must generalize to new classes based on only a few available examples. Extending Prototypical Networks, LC-Protonets generate one prototype per label combination, derived from the power set of labels present in the limited training items, rather than one prototype per label. Our method is applied to automatic audio tagging across diverse music datasets, covering various cultures and including both modern and traditional music, and is evaluated against existing approaches in the literature. The results demonstrate a significant performance improvement in almost all domains and training setups when using LC-Protonets for multi-label classification. In addition to training a few-shot learning model from scratch, we explore the use of a pre-trained model, obtained via supervised learning, to embed items in the feature space. Fine-tuning improves the generalization ability of all methods, yet LC-Protonets achieve high-level performance even without fine-tuning, in contrast to the comparative approaches. We finally analyze the scalability of the proposed method, providing detailed quantitative metrics from our experiments. The implementation and experimental setup are made publicly available, offering a benchmark for future research. |
| format | Article |
| id | doaj-art-0880e47d41d742e9ba97a68cf07ddff0 |
| institution | Kabale University |
| issn | 2644-1322 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Open Journal of Signal Processing |
| spelling | doaj-art-0880e47d41d742e9ba97a68cf07ddff02025-02-11T00:01:46ZengIEEEIEEE Open Journal of Signal Processing2644-13222025-01-01613814610.1109/OJSP.2025.352931510839319LC-Protonets: Multi-Label Few-Shot Learning for World Music Audio TaggingCharilaos Papaioannou0https://orcid.org/0009-0008-8558-3255Emmanouil Benetos1https://orcid.org/0000-0002-6820-6764Alexandros Potamianos2https://orcid.org/0009-0007-1532-5288School of ECE, National Technical University of Athens, Zografou, GreeceCentre for Digital Music, Queen Mary University of London, London, U.K.School of ECE, National Technical University of Athens, Zografou, GreeceWe introduce Label-Combination Prototypical Networks (LC-Protonets) to address the problem of multi-label few-shot classification, where a model must generalize to new classes based on only a few available examples. Extending Prototypical Networks, LC-Protonets generate one prototype per label combination, derived from the power set of labels present in the limited training items, rather than one prototype per label. Our method is applied to automatic audio tagging across diverse music datasets, covering various cultures and including both modern and traditional music, and is evaluated against existing approaches in the literature. The results demonstrate a significant performance improvement in almost all domains and training setups when using LC-Protonets for multi-label classification. In addition to training a few-shot learning model from scratch, we explore the use of a pre-trained model, obtained via supervised learning, to embed items in the feature space. Fine-tuning improves the generalization ability of all methods, yet LC-Protonets achieve high-level performance even without fine-tuning, in contrast to the comparative approaches. We finally analyze the scalability of the proposed method, providing detailed quantitative metrics from our experiments. The implementation and experimental setup are made publicly available, offering a benchmark for future research.https://ieeexplore.ieee.org/document/10839319/Few-shot learningprototypical networksmulti-label classificationaudio taggingworld music datasets |
| spellingShingle | Charilaos Papaioannou Emmanouil Benetos Alexandros Potamianos LC-Protonets: Multi-Label Few-Shot Learning for World Music Audio Tagging IEEE Open Journal of Signal Processing Few-shot learning prototypical networks multi-label classification audio tagging world music datasets |
| title | LC-Protonets: Multi-Label Few-Shot Learning for World Music Audio Tagging |
| title_full | LC-Protonets: Multi-Label Few-Shot Learning for World Music Audio Tagging |
| title_fullStr | LC-Protonets: Multi-Label Few-Shot Learning for World Music Audio Tagging |
| title_full_unstemmed | LC-Protonets: Multi-Label Few-Shot Learning for World Music Audio Tagging |
| title_short | LC-Protonets: Multi-Label Few-Shot Learning for World Music Audio Tagging |
| title_sort | lc protonets multi label few shot learning for world music audio tagging |
| topic | Few-shot learning prototypical networks multi-label classification audio tagging world music datasets |
| url | https://ieeexplore.ieee.org/document/10839319/ |
| work_keys_str_mv | AT charilaospapaioannou lcprotonetsmultilabelfewshotlearningforworldmusicaudiotagging AT emmanouilbenetos lcprotonetsmultilabelfewshotlearningforworldmusicaudiotagging AT alexandrospotamianos lcprotonetsmultilabelfewshotlearningforworldmusicaudiotagging |