KerJEPA: Kernel Discrepancies for Euclidean Self-Supervised Learning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zimmermann, Eric, Wiltzer, Harley, Szeto, Justin, Alvarez-Melis, David, Mackey, Lester
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912783119941632
author Zimmermann, Eric
Wiltzer, Harley
Szeto, Justin
Alvarez-Melis, David
Mackey, Lester
author_facet Zimmermann, Eric
Wiltzer, Harley
Szeto, Justin
Alvarez-Melis, David
Mackey, Lester
contents Recent breakthroughs in self-supervised Joint-Embedding Predictive Architectures (JEPAs) have established that regularizing Euclidean representations toward isotropic Gaussian priors yields provable gains in training stability and downstream generalization. We introduce a new, flexible family of KerJEPAs, self-supervised learning algorithms with kernel-based regularizers. One instance of this family corresponds to the recently-introduced LeJEPA Epps-Pulley regularizer which approximates a sliced maximum mean discrepancy (MMD) with a Gaussian prior and Gaussian kernel. By expanding the class of viable kernels and priors and computing the closed-form high-dimensional limit of sliced MMDs, we develop alternative KerJEPAs with a number of favorable properties including improved training stability and design flexibility.
format Preprint
id arxiv_https___arxiv_org_abs_2512_19605
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle KerJEPA: Kernel Discrepancies for Euclidean Self-Supervised Learning
Zimmermann, Eric
Wiltzer, Harley
Szeto, Justin
Alvarez-Melis, David
Mackey, Lester
Machine Learning
Computer Vision and Pattern Recognition
Recent breakthroughs in self-supervised Joint-Embedding Predictive Architectures (JEPAs) have established that regularizing Euclidean representations toward isotropic Gaussian priors yields provable gains in training stability and downstream generalization. We introduce a new, flexible family of KerJEPAs, self-supervised learning algorithms with kernel-based regularizers. One instance of this family corresponds to the recently-introduced LeJEPA Epps-Pulley regularizer which approximates a sliced maximum mean discrepancy (MMD) with a Gaussian prior and Gaussian kernel. By expanding the class of viable kernels and priors and computing the closed-form high-dimensional limit of sliced MMDs, we develop alternative KerJEPAs with a number of favorable properties including improved training stability and design flexibility.
title KerJEPA: Kernel Discrepancies for Euclidean Self-Supervised Learning
topic Machine Learning
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2512.19605