Adaptive Prototypical Networks

TL;DR

This paper introduces an adaptive approach to prototypical networks for few-shot learning, improving class separation in the embedding space by considering inter-class relationships during meta-testing.

Contribution

It proposes a novel method that pushes class embeddings apart during meta-testing, enhancing classification accuracy over standard prototypical networks.

Findings

Improved meta-testing accuracy on benchmark datasets.

Better separation of class embeddings in the feature space.

Outperforms standard prototypical networks and other few-shot models.

Abstract

Prototypical network for Few shot learning tries to learn an embedding function in the encoder that embeds images with similar features close to one another in the embedding space. However, in this process, the support set samples for a task are embedded independently of one other, and hence, the inter-class closeness is not taken into account. Thus, in the presence of similar-looking classes in a task, the embeddings will tend to be close to each other in the embedding space and even possibly overlap in some regions, which is not desirable for classification. In this paper, we propose an approach that intuitively pushes the embeddings of each of the classes away from the others in the meta-testing phase, thereby grouping them closely based on the distinct class labels rather than only the similarity of spatial features. This is achieved by training the encoder network for classification using the support set samples and labels of the new task. Extensive experiments conducted on benchmark data sets show improvements in meta-testing accuracy when compared with Prototypical Networks and also other standard few-shot learning models.