Meta-Learner with Linear Nulling

TL;DR

This paper introduces a meta-learning algorithm that uses linear null-space projection to improve few-shot image classification by creating an alternative decision space, achieving state-of-the-art results on standard datasets.

Contribution

The paper presents a novel meta-learning method employing linear nulling to enhance classification accuracy in low-data scenarios, outperforming existing approaches.

Findings

Achieves top or near-top accuracy on Omniglot and miniImageNet datasets.

Uses null-space projection to quickly zero out errors during learning.

Demonstrates effectiveness of linear nulling in meta-learning contexts.

Abstract

We propose a meta-learning algorithm utilizing a linear transformer that carries out null-space projection of neural network outputs. The main idea is to construct an alternative classification space such that the error signals during few-shot learning are quickly zero-forced on that space so that reliable classification on low data is possible. The final decision on a query is obtained utilizing a null-space-projected distance measure between the network output and reference vectors, both of which have been trained in the initial learning phase. Among the known methods with a given model size, our meta-learner achieves the best or near-best image classification accuracies with Omniglot and miniImageNet datasets.