Hybrid Losses for Hierarchical Embedding Learning
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866912199326302208 |
|---|---|
| author | Tian, Haokun Lattner, Stefan McFee, Brian Saitis, Charalampos |
| author_facet | Tian, Haokun Lattner, Stefan McFee, Brian Saitis, Charalampos |
| contents | In traditional supervised learning, the cross-entropy loss treats all incorrect predictions equally, ignoring the relevance or proximity of wrong labels to the correct answer. By leveraging a tree hierarchy for fine-grained labels, we investigate hybrid losses, such as generalised triplet and cross-entropy losses, to enforce similarity between labels within a multi-task learning framework. We propose metrics to evaluate the embedding space structure and assess the model's ability to generalise to unseen classes, that is, to infer similar classes for data belonging to unseen categories. Our experiments on OrchideaSOL, a four-level hierarchical instrument sound dataset with nearly 200 detailed categories, demonstrate that the proposed hybrid losses outperform previous works in classification, retrieval, embedding space structure, and generalisation. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2501_12796 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Hybrid Losses for Hierarchical Embedding Learning Tian, Haokun Lattner, Stefan McFee, Brian Saitis, Charalampos Sound Information Retrieval Machine Learning Audio and Speech Processing In traditional supervised learning, the cross-entropy loss treats all incorrect predictions equally, ignoring the relevance or proximity of wrong labels to the correct answer. By leveraging a tree hierarchy for fine-grained labels, we investigate hybrid losses, such as generalised triplet and cross-entropy losses, to enforce similarity between labels within a multi-task learning framework. We propose metrics to evaluate the embedding space structure and assess the model's ability to generalise to unseen classes, that is, to infer similar classes for data belonging to unseen categories. Our experiments on OrchideaSOL, a four-level hierarchical instrument sound dataset with nearly 200 detailed categories, demonstrate that the proposed hybrid losses outperform previous works in classification, retrieval, embedding space structure, and generalisation. |
| title | Hybrid Losses for Hierarchical Embedding Learning |
| topic | Sound Information Retrieval Machine Learning Audio and Speech Processing |
| url | https://arxiv.org/abs/2501.12796 |