Saved in:
Bibliographic Details
Main Authors: Adam, Maxwell, Furman, Zach, Hoogland, Jesse
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2509.26537
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911186005524480
author Adam, Maxwell
Furman, Zach
Hoogland, Jesse
author_facet Adam, Maxwell
Furman, Zach
Hoogland, Jesse
contents We introduce the loss kernel, an interpretability method for measuring similarity between data points according to a trained neural network. The kernel is the covariance matrix of per-sample losses computed under a distribution of low-loss-preserving parameter perturbations. We first validate our method on a synthetic multitask problem, showing it separates inputs by task as predicted by theory. We then apply this kernel to Inception-v1 to visualize the structure of ImageNet, and we show that the kernel's structure aligns with the WordNet semantic hierarchy. This establishes the loss kernel as a practical tool for interpretability and data attribution.
format Preprint
id arxiv_https___arxiv_org_abs_2509_26537
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle The Loss Kernel: A Geometric Probe for Deep Learning Interpretability
Adam, Maxwell
Furman, Zach
Hoogland, Jesse
Machine Learning
We introduce the loss kernel, an interpretability method for measuring similarity between data points according to a trained neural network. The kernel is the covariance matrix of per-sample losses computed under a distribution of low-loss-preserving parameter perturbations. We first validate our method on a synthetic multitask problem, showing it separates inputs by task as predicted by theory. We then apply this kernel to Inception-v1 to visualize the structure of ImageNet, and we show that the kernel's structure aligns with the WordNet semantic hierarchy. This establishes the loss kernel as a practical tool for interpretability and data attribution.
title The Loss Kernel: A Geometric Probe for Deep Learning Interpretability
topic Machine Learning
url https://arxiv.org/abs/2509.26537