Saved in:
Bibliographic Details
Main Authors: Edwards, Thomas D. P., Alvey, James, Alsing, Justin, Nguyen, Nam H., Wandelt, Benjamin D.
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2405.13867
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915094847291392
author Edwards, Thomas D. P.
Alvey, James
Alsing, Justin
Nguyen, Nam H.
Wandelt, Benjamin D.
author_facet Edwards, Thomas D. P.
Alvey, James
Alsing, Justin
Nguyen, Nam H.
Wandelt, Benjamin D.
contents Scaling laws for large language models (LLMs) have provided useful guidance in training ever larger models for predictable performance gains. Time series forecasting shares a similar sequential structure to language, and is amenable to large-scale transformer architectures. Here we show that foundational decoder-only time series transformer models exhibit analogous scaling-behavior to LLMs, with architectural details (aspect ratio and number of heads) having a minimal effect over broad ranges. We assemble a large corpus of heterogenous time series data on which to train, and establish for the first time power-law scaling with parameter count, dataset size, and training compute, spanning five orders of magnitude.
format Preprint
id arxiv_https___arxiv_org_abs_2405_13867
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Scaling-laws for Large Time-series Models
Edwards, Thomas D. P.
Alvey, James
Alsing, Justin
Nguyen, Nam H.
Wandelt, Benjamin D.
Machine Learning
Artificial Intelligence
Scaling laws for large language models (LLMs) have provided useful guidance in training ever larger models for predictable performance gains. Time series forecasting shares a similar sequential structure to language, and is amenable to large-scale transformer architectures. Here we show that foundational decoder-only time series transformer models exhibit analogous scaling-behavior to LLMs, with architectural details (aspect ratio and number of heads) having a minimal effect over broad ranges. We assemble a large corpus of heterogenous time series data on which to train, and establish for the first time power-law scaling with parameter count, dataset size, and training compute, spanning five orders of magnitude.
title Scaling-laws for Large Time-series Models
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2405.13867