Abstract
The transfer learning paradigm of model pre-training and subsequent fine-tuning produces high- accuracy models. While most studies recommend scaling the pre-training size to benefit most from transfer learning, a question remains: what data and method should be used for pre-training? We investigate the impact of pre-training data distribution on the few-shot and full fine-tuning performance using 3 pre-training methods (supervised, contrastive language-image and image-image), 7 pre-training datasets, and 9 downstream datasets. Through extensive controlled experiments, we find that the choice of the pre-training data source is essential for the few-shot transfer, but its role decreases as more data is made available for fine-tuning. Additionally, we explore the role of data curation and examine the trade-offs between label noise and the size of the pre-training dataset. We find that using 2000× more pre-training data from LAION can match the performance of supervised ImageNet pre-training. Furthermore, we investigate the effect of pre-training methods, comparing language-image contrastive vs. image-image contrastive, and find that the latter leads to better downstream accuracy
Originalsprache | englisch |
---|---|
Seitenumfang | 38 |
Publikationsstatus | Veröffentlicht - 27 Feb. 2023 |
Veranstaltung | 1st Multimodal Representation Learning Workshop: ICLR 2023 - Virtual, Ruanda Dauer: 1 Mai 2023 → 5 Mai 2023 https://iclr.cc/virtual/2023/workshop/12836 |
Workshop
Workshop | 1st Multimodal Representation Learning Workshop |
---|---|
Kurztitel | ICLR 2023 |
Land/Gebiet | Ruanda |
Ort | Virtual |
Zeitraum | 1/05/23 → 5/05/23 |
Internetadresse |
Fields of Expertise
- Information, Communication & Computing