The Curse of Low Task Diversity: On the Failure of Transfer Learning to Outperform MAML and Their Empirical Equivalence

TL;DR

This paper introduces a new metric called the diversity coefficient to measure task diversity in few-shot learning benchmarks, revealing that low diversity explains the empirical equivalence of transfer learning and MAML in such settings.

Contribution

The paper proposes the diversity coefficient metric and demonstrates its effectiveness in explaining the similarity between transfer learning and MAML solutions in low-diversity benchmarks.

Findings

Low diversity in MiniImageNet and CIFAR-FS correlates with similar solutions from transfer learning and MAML.

Transfer learning and MAML achieve comparable accuracy and feature similarity in low-diversity regimes.

Meta-test performance remains consistent across different model sizes in low-diversity settings.

Abstract

Recently, it has been observed that a transfer learning solution might be all we need to solve many few-shot learning benchmarks -- thus raising important questions about when and how meta-learning algorithms should be deployed. In this paper, we seek to clarify these questions by 1. proposing a novel metric -- the diversity coefficient -- to measure the diversity of tasks in a few-shot learning benchmark and 2. by comparing Model-Agnostic Meta-Learning (MAML) and transfer learning under fair conditions (same architecture, same optimizer, and all models trained to convergence). Using the diversity coefficient, we show that the popular MiniImageNet and CIFAR-FS few-shot learning benchmarks have low diversity. This novel insight contextualizes claims that transfer learning solutions are better than meta-learned solutions in the regime of low diversity under a fair comparison. Specifically, we empirically find that a low diversity coefficient correlates with a high similarity between transfer learning and MAML learned solutions in terms of accuracy at meta-test time and classification layer similarity (using feature based distance metrics like SVCCA, PWCCA, CKA, and OPD). To further support our claim, we find this meta-test accuracy holds even as the model size changes. Therefore, we conclude that in the low diversity regime, MAML and transfer learning have equivalent meta-test performance when both are compared fairly. We also hope our work inspires more thoughtful constructions and quantitative evaluations of meta-learning benchmarks in the future.