Gradient-Based Meta-Learning with Learned Layerwise Metric and Subspace

TL;DR

This paper introduces MT-net, a meta-learning framework that learns layerwise subspaces and distance metrics to improve task adaptation, achieving state-of-the-art results in few-shot learning.

Contribution

The paper proposes MT-net, which learns subspaces and metrics for each layer to enhance gradient-based meta-learning, making adaptation more efficient and less sensitive to initial learning rates.

Findings

Achieves state-of-the-art performance on few-shot tasks.

Learns subspaces that reflect task complexity.

Less sensitive to initial learning rate choices.

Abstract

Gradient-based meta-learning methods leverage gradient descent to learn the commonalities among various tasks. While previous such methods have been successful in meta-learning tasks, they resort to simple gradient descent during meta-testing. Our primary contribution is the {\em MT-net}, which enables the meta-learner to learn on each layer's activation space a subspace that the task-specific learner performs gradient descent on. Additionally, a task-specific learner of an {\em MT-net} performs gradient descent with respect to a meta-learned distance metric, which warps the activation space to be more sensitive to task identity. We demonstrate that the dimension of this learned subspace reflects the complexity of the task-specific learner's adaptation task, and also that our model is less sensitive to the choice of initial learning rates than previous gradient-based meta-learning methods. Our method achieves state-of-the-art or comparable performance on few-shot classification and regression tasks.