Mutual-Information Based Few-Shot Classification

TL;DR

This paper presents TIM, a transductive method for few-shot learning that maximizes mutual information between query features and labels, achieving significant improvements over state-of-the-art methods across various datasets.

Contribution

TIM introduces a mutual-information maximization approach with a fast alternating-direction solver for transductive few-shot classification, compatible with any feature extractor.

Findings

TIM outperforms state-of-the-art methods by 2-5% accuracy.

It is effective across multiple datasets and challenging scenarios.

The method is modular and can be applied on top of existing feature extractors.

Abstract

We introduce Transductive Infomation Maximization (TIM) for few-shot learning. Our method maximizes the mutual information between the query features and their label predictions for a given few-shot task, in conjunction with a supervision loss based on the support set. We motivate our transductive loss by deriving a formal relation between the classification accuracy and mutual-information maximization. Furthermore, we propose a new alternating-direction solver, which substantially speeds up transductive inference over gradient-based optimization, while yielding competitive accuracy. We also provide a convergence analysis of our solver based on Zangwill's theory and bound-optimization arguments. TIM inference is modular: it can be used on top of any base-training feature extractor. Following standard transductive few-shot settings, our comprehensive experiments demonstrate that TIM outperforms state-of-the-art methods significantly across various datasets and networks, while used on top of a fixed feature extractor trained with simple cross-entropy on the base classes, without resorting to complex meta-learning schemes. It consistently brings between 2 % and 5 % improvement in accuracy over the best performing method, not only on all the well-established few-shot benchmarks but also on more challenging scenarios, with random tasks, domain shift and larger numbers of classes, as in the recently introduced META-DATASET. Our code is publicly available at https://github.com/mboudiaf/TIM. We also publicly release a standalone PyTorch implementation of META-DATASET, along with additional benchmarking results, at https://github.com/mboudiaf/pytorch-meta-dataset.