Bayesian Neural Scaling Law Extrapolation with Prior-Data Fitted Networks

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Lee, Dongwoo, Lee, Dong Bok, Adriaensen, Steven, Lee, Juho, Hwang, Sung Ju, Hutter, Frank, Kim, Seon Joo, Lee, Hae Beom
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912431472640000
author Lee, Dongwoo
Lee, Dong Bok
Adriaensen, Steven
Lee, Juho
Hwang, Sung Ju
Hutter, Frank
Kim, Seon Joo
Lee, Hae Beom
author_facet Lee, Dongwoo
Lee, Dong Bok
Adriaensen, Steven
Lee, Juho
Hwang, Sung Ju
Hutter, Frank
Kim, Seon Joo
Lee, Hae Beom
contents Scaling has been a major driver of recent advancements in deep learning. Numerous empirical studies have found that scaling laws often follow the power-law and proposed several variants of power-law functions to predict the scaling behavior at larger scales. However, existing methods mostly rely on point estimation and do not quantify uncertainty, which is crucial for real-world applications involving decision-making problems such as determining the expected performance improvements achievable by investing additional computational resources. In this work, we explore a Bayesian framework based on Prior-data Fitted Networks (PFNs) for neural scaling law extrapolation. Specifically, we design a prior distribution that enables the sampling of infinitely many synthetic functions resembling real-world neural scaling laws, allowing our PFN to meta-learn the extrapolation. We validate the effectiveness of our approach on real-world neural scaling laws, comparing it against both the existing point estimation methods and Bayesian approaches. Our method demonstrates superior performance, particularly in data-limited scenarios such as Bayesian active learning, underscoring its potential for reliable, uncertainty-aware extrapolation in practical applications.
format Preprint
id arxiv_https___arxiv_org_abs_2505_23032
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Bayesian Neural Scaling Law Extrapolation with Prior-Data Fitted Networks
Lee, Dongwoo
Lee, Dong Bok
Adriaensen, Steven
Lee, Juho
Hwang, Sung Ju
Hutter, Frank
Kim, Seon Joo
Lee, Hae Beom
Machine Learning
Artificial Intelligence
Scaling has been a major driver of recent advancements in deep learning. Numerous empirical studies have found that scaling laws often follow the power-law and proposed several variants of power-law functions to predict the scaling behavior at larger scales. However, existing methods mostly rely on point estimation and do not quantify uncertainty, which is crucial for real-world applications involving decision-making problems such as determining the expected performance improvements achievable by investing additional computational resources. In this work, we explore a Bayesian framework based on Prior-data Fitted Networks (PFNs) for neural scaling law extrapolation. Specifically, we design a prior distribution that enables the sampling of infinitely many synthetic functions resembling real-world neural scaling laws, allowing our PFN to meta-learn the extrapolation. We validate the effectiveness of our approach on real-world neural scaling laws, comparing it against both the existing point estimation methods and Bayesian approaches. Our method demonstrates superior performance, particularly in data-limited scenarios such as Bayesian active learning, underscoring its potential for reliable, uncertainty-aware extrapolation in practical applications.
title Bayesian Neural Scaling Law Extrapolation with Prior-Data Fitted Networks
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2505.23032