On the Subspace Structure of Gradient-Based Meta-Learning
This work provides a theoretical insight into the parameter updates in meta-learning, which could help improve efficiency and understanding for researchers in machine learning, though it appears incremental as it builds on prior observations.
The authors analyzed the distribution of post-adaptation parameters in Gradient-Based Meta-Learning (GBML), showing that updates occur in a low-dimensional subspace aligned with task-space dimensionality, extending this finding from image classification to regression. They also used this subspace structure to estimate the intrinsic dimension of tasks in few-shot learning datasets.
In this work we provide an analysis of the distribution of the post-adaptation parameters of Gradient-Based Meta-Learning (GBML) methods. Previous work has noticed how, for the case of image-classification, this adaptation only takes place on the last layers of the network. We propose the more general notion that parameters are updated over a low-dimensional \emph{subspace} of the same dimensionality as the task-space and show that this holds for regression as well. Furthermore, the induced subspace structure provides a method to estimate the intrinsic dimension of the space of tasks of common few-shot learning datasets.