InfoNCE: Identifying the Gap Between Theory and Practice
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2024
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866912330139303936 |
|---|---|
| author | Rusak, Evgenia Reizinger, Patrik Juhos, Attila Bringmann, Oliver Zimmermann, Roland S. Brendel, Wieland |
| author_facet | Rusak, Evgenia Reizinger, Patrik Juhos, Attila Bringmann, Oliver Zimmermann, Roland S. Brendel, Wieland |
| contents | Prior theory work on Contrastive Learning via the InfoNCE loss showed that, under certain assumptions, the learned representations recover the ground-truth latent factors. We argue that these theories overlook crucial aspects of how CL is deployed in practice. Specifically, they either assume equal variance across all latents or that certain latents are kept invariant. However, in practice, positive pairs are often generated using augmentations such as strong cropping to just a few pixels. Hence, a more realistic assumption is that all latent factors change with a continuum of variability across all factors. We introduce AnInfoNCE, a generalization of InfoNCE that can provably uncover the latent factors in this anisotropic setting, broadly generalizing previous identifiability results in CL. We validate our identifiability results in controlled experiments and show that AnInfoNCE increases the recovery of previously collapsed information in CIFAR10 and ImageNet, albeit at the cost of downstream accuracy. Finally, we discuss the remaining mismatches between theoretical assumptions and practical implementations. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2407_00143 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | InfoNCE: Identifying the Gap Between Theory and Practice Rusak, Evgenia Reizinger, Patrik Juhos, Attila Bringmann, Oliver Zimmermann, Roland S. Brendel, Wieland Machine Learning Computer Vision and Pattern Recognition Prior theory work on Contrastive Learning via the InfoNCE loss showed that, under certain assumptions, the learned representations recover the ground-truth latent factors. We argue that these theories overlook crucial aspects of how CL is deployed in practice. Specifically, they either assume equal variance across all latents or that certain latents are kept invariant. However, in practice, positive pairs are often generated using augmentations such as strong cropping to just a few pixels. Hence, a more realistic assumption is that all latent factors change with a continuum of variability across all factors. We introduce AnInfoNCE, a generalization of InfoNCE that can provably uncover the latent factors in this anisotropic setting, broadly generalizing previous identifiability results in CL. We validate our identifiability results in controlled experiments and show that AnInfoNCE increases the recovery of previously collapsed information in CIFAR10 and ImageNet, albeit at the cost of downstream accuracy. Finally, we discuss the remaining mismatches between theoretical assumptions and practical implementations. |
| title | InfoNCE: Identifying the Gap Between Theory and Practice |
| topic | Machine Learning Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2407.00143 |