Ortho-Shot: Low Displacement Rank Regularization with Data Augmentation for Few-Shot Learning

TL;DR

This paper introduces Ortho-Shot, a novel regularization method using low displacement rank and orthogonal constraints on convolutional layers, combined with data augmentation, to improve few-shot learning performance.

Contribution

It proposes a low displacement rank regularization strategy with orthogonal constraints and demonstrates its effectiveness combined with data augmentation for few-shot classification.

Findings

Significant performance improvement (~5%) over state-of-the-art methods.

Enhanced generalization and intra-class feature embeddings.

Effective reduction of overfitting in few-shot models.

Abstract

In few-shot classification, the primary goal is to learn representations from a few samples that generalize well for novel classes. In this paper, we propose an efficient low displacement rank (LDR) regularization strategy termed Ortho-Shot; a technique that imposes orthogonal regularization on the convolutional layers of a few-shot classifier, which is based on the doubly-block toeplitz (DBT) matrix structure. The regularized convolutional layers of the few-shot classifier enhances model generalization and intra-class feature embeddings that are crucial for few-shot learning. Overfitting is a typical issue for few-shot models, the lack of data diversity inhibits proper model inference which weakens the classification accuracy of few-shot learners to novel classes. In this regard, we broke down the pipeline of the few-shot classifier and established that the support, query and task data augmentation collectively alleviates overfitting in networks. With compelling results, we demonstrated that combining a DBT-based low-rank orthogonal regularizer with data augmentation strategies, significantly boosts the performance of a few-shot classifier. We perform our experiments on the miniImagenet, CIFAR-FS and Stanford datasets with performance values of about 5\% when compared to state-of-the-art