Saved in:
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2021
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2104.11946 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866916389583847424 |
|---|---|
| author | Chorowski, Jan Ciesielski, Grzegorz Dzikowski, Jarosław Łańcucki, Adrian Marxer, Ricard Opala, Mateusz Pusz, Piotr Rychlikowski, Paweł Stypułkowski, Michał |
| author_facet | Chorowski, Jan Ciesielski, Grzegorz Dzikowski, Jarosław Łańcucki, Adrian Marxer, Ricard Opala, Mateusz Pusz, Piotr Rychlikowski, Paweł Stypułkowski, Michał |
| contents | We investigate the possibility of forcing a self-supervised model trained using a contrastive predictive loss to extract slowly varying latent representations. Rather than producing individual predictions for each of the future representations, the model emits a sequence of predictions shorter than that of the upcoming representations to which they will be aligned. In this way, the prediction network solves a simpler task of predicting the next symbols, but not their exact timing, while the encoding network is trained to produce piece-wise constant latent codes. We evaluate the model on a speech coding task and demonstrate that the proposed Aligned Contrastive Predictive Coding (ACPC) leads to higher linear phone prediction accuracy and lower ABX error rates, while being slightly faster to train due to the reduced number of prediction heads. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2104_11946 |
| institution | arXiv |
| publishDate | 2021 |
| record_format | arxiv |
| spellingShingle | Aligned Contrastive Predictive Coding Chorowski, Jan Ciesielski, Grzegorz Dzikowski, Jarosław Łańcucki, Adrian Marxer, Ricard Opala, Mateusz Pusz, Piotr Rychlikowski, Paweł Stypułkowski, Michał Machine Learning Sound Audio and Speech Processing We investigate the possibility of forcing a self-supervised model trained using a contrastive predictive loss to extract slowly varying latent representations. Rather than producing individual predictions for each of the future representations, the model emits a sequence of predictions shorter than that of the upcoming representations to which they will be aligned. In this way, the prediction network solves a simpler task of predicting the next symbols, but not their exact timing, while the encoding network is trained to produce piece-wise constant latent codes. We evaluate the model on a speech coding task and demonstrate that the proposed Aligned Contrastive Predictive Coding (ACPC) leads to higher linear phone prediction accuracy and lower ABX error rates, while being slightly faster to train due to the reduced number of prediction heads. |
| title | Aligned Contrastive Predictive Coding |
| topic | Machine Learning Sound Audio and Speech Processing |
| url | https://arxiv.org/abs/2104.11946 |