Few and Fewer: Learning Better from Few Examples Using Fewer Base Classes
Primary Area: transfer learning, meta learning, and lifelong learning
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Keywords: Few-Shot Learning, Transfer Learning, Data-centric
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.
TL;DR: We improve few-shot classification performance by finetuning on selected subsets of the base dataset or by selecting in a static library of specialist features extractors.
Abstract: When training data is scarce, it is common to make use of a feature extractor that has been pre-trained on a large “base” dataset, either by fine-tuning its parameters on the “target” dataset or by directly adopting its representation as features for a simple classifier. Fine-tuning is ineffective for few-shot learning, since the target dataset contains only a handful of examples. However, directly adopting the features without fine-tuning relies on the distribution of the base dataset being similar enough to that of the target dataset in order to achieve separability and generalization. This paper investigates whether better features for the target dataset can be obtained by training on fewer base classes, in an effort to bring the distribution of the base dataset closer to that of the target dataset. We consider cross-domain few-shot image classification in eight different domains from Meta-Dataset and entertain multiple real-world settings (domain-informed, task-informed and uninformed) where progressively less detail is known about the target task. To our knowledge, this is the first demonstration that fine-tuning on a subset of carefully selected base classes can significantly improve few-shot learning. Our contributions are simple and intuitive methods that can be implemented in any few-shot solution. We also give insights into the conditions in which these solutions are likely to provide a boost in accuracy. We release the code to reproduce all experiments from this paper on GitHub. https://anonymous.4open.science/r/Few-and-Fewer-C978
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 6940
Loading