Saved in:
Bibliographic Details
Main Authors: Zhang, Yilang, Li, Bingcong, Giannakis, Georgios B.
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2509.02418
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909766697091072
author Zhang, Yilang
Li, Bingcong
Giannakis, Georgios B.
author_facet Zhang, Yilang
Li, Bingcong
Giannakis, Georgios B.
contents Utilizing task-invariant knowledge acquired from related tasks as prior information, meta-learning offers a principled approach to learning a new task with limited data records. Sample-efficient adaptation of this prior information is a major challenge facing meta-learning, and plays an important role because it facilitates training the sought task-specific model with just a few optimization steps. Past works deal with this challenge through preconditioning that speeds up convergence of the per-task training. Though effective in representing locally quadratic loss curvatures, simple linear preconditioning can be hardly potent with complex loss geometries. Instead of relying on a quadratic distance metric, the present contribution copes with complex loss metrics by learning a versatile distance-generating function, which induces a nonlinear mirror map to effectively capture and optimize a wide range of loss geometries. With suitable parameterization, this generating function is effected by an expressive neural network that is provably a valid distance. Analytical results establish convergence of not only the proposed method, but also all meta-learning approaches based on preconditioning. To attain gradient norm less than $ε$, the convergence rate of $\mathcal{O}(ε^{-2})$ is on par with standard gradient-based meta-learning methods. Numerical tests on few-shot learning datasets demonstrate the superior empirical performance of the novel algorithm, as well as its rapid per-task convergence, which markedly reduces the number of adaptation steps, hence also accommodating large-scale meta-learning models.
format Preprint
id arxiv_https___arxiv_org_abs_2509_02418
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Learnable Loss Geometries with Mirror Descent for Scalable and Convergent Meta-Learning
Zhang, Yilang
Li, Bingcong
Giannakis, Georgios B.
Machine Learning
Utilizing task-invariant knowledge acquired from related tasks as prior information, meta-learning offers a principled approach to learning a new task with limited data records. Sample-efficient adaptation of this prior information is a major challenge facing meta-learning, and plays an important role because it facilitates training the sought task-specific model with just a few optimization steps. Past works deal with this challenge through preconditioning that speeds up convergence of the per-task training. Though effective in representing locally quadratic loss curvatures, simple linear preconditioning can be hardly potent with complex loss geometries. Instead of relying on a quadratic distance metric, the present contribution copes with complex loss metrics by learning a versatile distance-generating function, which induces a nonlinear mirror map to effectively capture and optimize a wide range of loss geometries. With suitable parameterization, this generating function is effected by an expressive neural network that is provably a valid distance. Analytical results establish convergence of not only the proposed method, but also all meta-learning approaches based on preconditioning. To attain gradient norm less than $ε$, the convergence rate of $\mathcal{O}(ε^{-2})$ is on par with standard gradient-based meta-learning methods. Numerical tests on few-shot learning datasets demonstrate the superior empirical performance of the novel algorithm, as well as its rapid per-task convergence, which markedly reduces the number of adaptation steps, hence also accommodating large-scale meta-learning models.
title Learnable Loss Geometries with Mirror Descent for Scalable and Convergent Meta-Learning
topic Machine Learning
url https://arxiv.org/abs/2509.02418