Interval Bound Interpolation for Few-shot Learning with Few Tasks

TL;DR

This paper introduces a novel interval bound interpolation method for few-shot learning that preserves task neighborhoods and generates artificial tasks, improving generalization in scenarios with limited tasks.

Contribution

It proposes using interval bounds to explicitly model task neighborhoods and interpolate new tasks, enhancing few-shot learning with few training tasks.

Findings

Improved performance on multiple datasets across domains.

Effective preservation of task neighborhoods during training.

Compatibility with meta-learning and metric-learning frameworks.

Abstract

Few-shot learning aims to transfer the knowledge acquired from training on a diverse set of tasks to unseen tasks from the same task distribution with a limited amount of labeled data. The underlying requirement for effective few-shot generalization is to learn a good representation of the task manifold. This becomes more difficult when only a limited number of tasks are available for training. In such a few-task few-shot setting, it is beneficial to explicitly preserve the local neighborhoods from the task manifold and exploit this to generate artificial tasks for training. To this end, we introduce the notion of interval bounds from the provably robust training literature to few-shot learning. The interval bounds are used to characterize neighborhoods around the training tasks. These neighborhoods can then be preserved by minimizing the distance between a task and its respective bounds. We then use a novel strategy to artificially form new tasks for training by interpolating between the available tasks and their respective interval bounds. We apply our framework to both model-agnostic meta-learning as well as prototype-based metric-learning paradigms. The efficacy of our proposed approach is evident from the improved performance on several datasets from diverse domains compared to current methods.