Loading [MathJax]/extensions/tex2jax.js

Optimizing Time Series Forecasting Architectures: A Hierarchical Neural Architecture Search Approach

Research output: Working paper/Preprint › Preprint

Authors

Research Organisations

Details

Original languageEnglish
Publication statusE-pub ahead of print - 10 Jun 2024

Abstract

The rapid development of time series forecasting research has brought many deep learning-based modules in this field. However, despite the increasing amount of new forecasting architectures, it is still unclear if we have leveraged the full potential of these existing modules within a properly designed architecture. In this work, we propose a novel hierarchical neural architecture search approach for time series forecasting tasks. With the design of a hierarchical search space, we incorporate many architecture types designed for forecasting tasks and allow for the efficient combination of different forecasting architecture modules. Results on long-term-time-series-forecasting tasks show that our approach can search for lightweight high-performing forecasting architectures across different forecasting tasks.

Keywords

    AutoML, Time Series Foreacsting, Neural Architecture Search

Cite this

Download
@techreport{dfbf322249834211a630d0785b24218a,
title = "Optimizing Time Series Forecasting Architectures: A Hierarchical Neural Architecture Search Approach",
abstract = " The rapid development of time series forecasting research has brought many deep learning-based modules in this field. However, despite the increasing amount of new forecasting architectures, it is still unclear if we have leveraged the full potential of these existing modules within a properly designed architecture. In this work, we propose a novel hierarchical neural architecture search approach for time series forecasting tasks. With the design of a hierarchical search space, we incorporate many architecture types designed for forecasting tasks and allow for the efficient combination of different forecasting architecture modules. Results on long-term-time-series-forecasting tasks show that our approach can search for lightweight high-performing forecasting architectures across different forecasting tasks. ",
keywords = "AutoML, Time Series Foreacsting, Neural Architecture Search",
author = "Difan Deng and Marius Lindauer",
year = "2024",
month = jun,
day = "10",
doi = "10.48550/arXiv.2406.05088",
language = "English",
type = "WorkingPaper",

}

Download

TY - UNPB

T1 - Optimizing Time Series Forecasting Architectures

T2 - A Hierarchical Neural Architecture Search Approach

AU - Deng, Difan

AU - Lindauer, Marius

PY - 2024/6/10

Y1 - 2024/6/10

N2 - The rapid development of time series forecasting research has brought many deep learning-based modules in this field. However, despite the increasing amount of new forecasting architectures, it is still unclear if we have leveraged the full potential of these existing modules within a properly designed architecture. In this work, we propose a novel hierarchical neural architecture search approach for time series forecasting tasks. With the design of a hierarchical search space, we incorporate many architecture types designed for forecasting tasks and allow for the efficient combination of different forecasting architecture modules. Results on long-term-time-series-forecasting tasks show that our approach can search for lightweight high-performing forecasting architectures across different forecasting tasks.

AB - The rapid development of time series forecasting research has brought many deep learning-based modules in this field. However, despite the increasing amount of new forecasting architectures, it is still unclear if we have leveraged the full potential of these existing modules within a properly designed architecture. In this work, we propose a novel hierarchical neural architecture search approach for time series forecasting tasks. With the design of a hierarchical search space, we incorporate many architecture types designed for forecasting tasks and allow for the efficient combination of different forecasting architecture modules. Results on long-term-time-series-forecasting tasks show that our approach can search for lightweight high-performing forecasting architectures across different forecasting tasks.

KW - AutoML

KW - Time Series Foreacsting

KW - Neural Architecture Search

U2 - 10.48550/arXiv.2406.05088

DO - 10.48550/arXiv.2406.05088

M3 - Preprint

BT - Optimizing Time Series Forecasting Architectures

ER -

By the same author(s)