Adaptive Task Sampling for Meta-Learning

TL;DR

This paper introduces an adaptive task sampling method for meta-learning that selects more informative and challenging tasks, leading to improved generalization in few-shot classification benchmarks.

Contribution

The paper proposes a novel greedy class-pair based sampling approach to generate more informative tasks for meta-learning, enhancing performance over random sampling.

Findings

Achieves consistent improvements across multiple benchmarks

Enhances generalization in few-shot classification

Works with various feature backbones and meta-learning algorithms

Abstract

Meta-learning methods have been extensively studied and applied in computer vision, especially for few-shot classification tasks. The key idea of meta-learning for few-shot classification is to mimic the few-shot situations faced at test time by randomly sampling classes in meta-training data to construct few-shot tasks for episodic training. While a rich line of work focuses solely on how to extract meta-knowledge across tasks, we exploit the complementary problem on how to generate informative tasks. We argue that the randomly sampled tasks could be sub-optimal and uninformative (e.g., the task of classifying "dog" from "laptop" is often trivial) to the meta-learner. In this paper, we propose an adaptive task sampling method to improve the generalization performance. Unlike instance based sampling, task based sampling is much more challenging due to the implicit definition of the task in each episode. Therefore, we accordingly propose a greedy class-pair based sampling method, which selects difficult tasks according to class-pair potentials. We evaluate our adaptive task sampling method on two few-shot classification benchmarks, and it achieves consistent improvements across different feature backbones, meta-learning algorithms and datasets.