Salvato in:
Dettagli Bibliografici
Autori principali: Lau, Edmund, Furman, Zach, Wang, George, Murfet, Daniel, Wei, Susan
Natura: Preprint
Pubblicazione: 2023
Soggetti:
Accesso online:https://arxiv.org/abs/2308.12108
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866914960925261824
author Lau, Edmund
Furman, Zach
Wang, George
Murfet, Daniel
Wei, Susan
author_facet Lau, Edmund
Furman, Zach
Wang, George
Murfet, Daniel
Wei, Susan
contents The Local Learning Coefficient (LLC) is introduced as a novel complexity measure for deep neural networks (DNNs). Recognizing the limitations of traditional complexity measures, the LLC leverages Singular Learning Theory (SLT), which has long recognized the significance of singularities in the loss landscape geometry. This paper provides an extensive exploration of the LLC's theoretical underpinnings, offering both a clear definition and intuitive insights into its application. Moreover, we propose a new scalable estimator for the LLC, which is then effectively applied across diverse architectures including deep linear networks up to 100M parameters, ResNet image models, and transformer language models. Empirical evidence suggests that the LLC provides valuable insights into how training heuristics might influence the effective complexity of DNNs. Ultimately, the LLC emerges as a crucial tool for reconciling the apparent contradiction between deep learning's complexity and the principle of parsimony.
format Preprint
id arxiv_https___arxiv_org_abs_2308_12108
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle The Local Learning Coefficient: A Singularity-Aware Complexity Measure
Lau, Edmund
Furman, Zach
Wang, George
Murfet, Daniel
Wei, Susan
Machine Learning
Artificial Intelligence
62F15, 68T07, 14B05
The Local Learning Coefficient (LLC) is introduced as a novel complexity measure for deep neural networks (DNNs). Recognizing the limitations of traditional complexity measures, the LLC leverages Singular Learning Theory (SLT), which has long recognized the significance of singularities in the loss landscape geometry. This paper provides an extensive exploration of the LLC's theoretical underpinnings, offering both a clear definition and intuitive insights into its application. Moreover, we propose a new scalable estimator for the LLC, which is then effectively applied across diverse architectures including deep linear networks up to 100M parameters, ResNet image models, and transformer language models. Empirical evidence suggests that the LLC provides valuable insights into how training heuristics might influence the effective complexity of DNNs. Ultimately, the LLC emerges as a crucial tool for reconciling the apparent contradiction between deep learning's complexity and the principle of parsimony.
title The Local Learning Coefficient: A Singularity-Aware Complexity Measure
topic Machine Learning
Artificial Intelligence
62F15, 68T07, 14B05
url https://arxiv.org/abs/2308.12108