Improving Few-Shot Learning using Composite Rotation based Auxiliary Task

TL;DR

This paper introduces a novel composite rotation auxiliary task that enhances feature quality in neural networks, leading to improved few-shot classification performance across multiple benchmarks.

Contribution

It proposes a dual-level rotation-based self-supervised auxiliary task that improves feature generality for few-shot learning, outperforming existing methods.

Findings

Outperforms existing few-shot learning methods on benchmark datasets.

The composite rotation auxiliary task enhances feature quality and generality.

Improves classification accuracy in low-data regimes.

Abstract

In this paper, we propose an approach to improve few-shot classification performance using a composite rotation based auxiliary task. Few-shot classification methods aim to produce neural networks that perform well for classes with a large number of training samples and classes with less number of training samples. They employ techniques to enable the network to produce highly discriminative features that are also very generic. Generally, the better the quality and generic-nature of the features produced by the network, the better is the performance of the network on few-shot learning. Our approach aims to train networks to produce such features by using a self-supervised auxiliary task. Our proposed composite rotation based auxiliary task performs rotation at two levels, i.e., rotation of patches inside the image (inner rotation) and rotation of the whole image (outer rotation) and assigns one out of 16 rotation classes to the modified image. We then simultaneously train for the composite rotation prediction task along with the original classification task, which forces the network to learn high-quality generic features that help improve the few-shot classification performance. We experimentally show that our approach performs better than existing few-shot learning methods on multiple benchmark datasets.