Estimating the Local Learning Coefficient at Scale

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Furman, Zach, Lau, Edmund
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866909328135421952
author Furman, Zach
Lau, Edmund
author_facet Furman, Zach
Lau, Edmund
contents The \textit{local learning coefficient} (LLC) is a principled way of quantifying model complexity, originally derived in the context of Bayesian statistics using singular learning theory (SLT). Several methods are known for numerically estimating the local learning coefficient, but so far these methods have not been extended to the scale of modern deep learning architectures or data sets. Using a method developed in {\tt arXiv:2308.12108 [stat.ML]} we empirically show how the LLC may be measured accurately and self-consistently for deep linear networks (DLNs) up to 100M parameters. We also show that the estimated LLC has the rescaling invariance that holds for the theoretical quantity.
format Preprint
id arxiv_https___arxiv_org_abs_2402_03698
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Estimating the Local Learning Coefficient at Scale
Furman, Zach
Lau, Edmund
Machine Learning
68T07, 14B05, 62F15
The \textit{local learning coefficient} (LLC) is a principled way of quantifying model complexity, originally derived in the context of Bayesian statistics using singular learning theory (SLT). Several methods are known for numerically estimating the local learning coefficient, but so far these methods have not been extended to the scale of modern deep learning architectures or data sets. Using a method developed in {\tt arXiv:2308.12108 [stat.ML]} we empirically show how the LLC may be measured accurately and self-consistently for deep linear networks (DLNs) up to 100M parameters. We also show that the estimated LLC has the rescaling invariance that holds for the theoretical quantity.
title Estimating the Local Learning Coefficient at Scale
topic Machine Learning
68T07, 14B05, 62F15
url https://arxiv.org/abs/2402.03698