Bregman Conditional Random Fields: Sequence Labeling with Parallelizable Inference Algorithms
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866913869620838400 |
|---|---|
| author | Corro, Caio Lacroix, Mathieu Roux, Joseph Le |
| author_facet | Corro, Caio Lacroix, Mathieu Roux, Joseph Le |
| contents | We propose a novel discriminative model for sequence labeling called Bregman conditional random fields (BCRF). Contrary to standard linear-chain conditional random fields, BCRF allows fast parallelizable inference algorithms based on iterative Bregman projections. We show how such models can be learned using Fenchel-Young losses, including extension for learning from partial labels. Experimentally, our approach delivers comparable results to CRF while being faster, and achieves better results in highly constrained settings compared to mean field, another parallelizable alternative. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2506_00732 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Bregman Conditional Random Fields: Sequence Labeling with Parallelizable Inference Algorithms Corro, Caio Lacroix, Mathieu Roux, Joseph Le Machine Learning Computation and Language We propose a novel discriminative model for sequence labeling called Bregman conditional random fields (BCRF). Contrary to standard linear-chain conditional random fields, BCRF allows fast parallelizable inference algorithms based on iterative Bregman projections. We show how such models can be learned using Fenchel-Young losses, including extension for learning from partial labels. Experimentally, our approach delivers comparable results to CRF while being faster, and achieves better results in highly constrained settings compared to mean field, another parallelizable alternative. |
| title | Bregman Conditional Random Fields: Sequence Labeling with Parallelizable Inference Algorithms |
| topic | Machine Learning Computation and Language |
| url | https://arxiv.org/abs/2506.00732 |