Meta-Dataset: A Dataset of Datasets for Learning to Learn from Few Examples

TL;DR

Meta-Dataset introduces a large-scale, diverse benchmark for few-shot learning, enabling more realistic evaluation of models' ability to generalize across different datasets and tasks.

Contribution

The paper presents Meta-Dataset, a new benchmark with diverse datasets for evaluating few-shot learning models, along with analysis and new baselines to assess meta-learning benefits.

Findings

Meta-Dataset reveals key challenges in generalization across datasets.

Meta-learning shows varying benefits depending on dataset diversity.

Proposed baselines help quantify meta-learning advantages.

Abstract

Few-shot classification refers to learning a classifier for new classes given only a few examples. While a plethora of models have emerged to tackle it, we find the procedure and datasets that are used to assess their progress lacking. To address this limitation, we propose Meta-Dataset: a new benchmark for training and evaluating models that is large-scale, consists of diverse datasets, and presents more realistic tasks. We experiment with popular baselines and meta-learners on Meta-Dataset, along with a competitive method that we propose. We analyze performance as a function of various characteristics of test tasks and examine the models' ability to leverage diverse training sources for improving their generalization. We also propose a new set of baselines for quantifying the benefit of meta-learning in Meta-Dataset. Our extensive experimentation has uncovered important research challenges and we hope to inspire work in these directions.