Efficient Automated Deep Learning for Time Series Forecasting

TL;DR

This paper introduces an efficient AutoDL framework tailored for time series forecasting, combining a novel neural architecture search space with Bayesian and multi-fidelity optimization to outperform existing methods.

Contribution

It presents a new AutoDL approach specifically designed for time series forecasting, integrating a comprehensive search space and advanced optimization techniques.

Findings

Significantly outperforms baseline models across multiple datasets.

Efficiently searches large configuration spaces with multi-fidelity optimization.

Demonstrates the effectiveness of the proposed system, extsystem, in time series tasks.

Abstract

Recent years have witnessed tremendously improved efficiency of Automated Machine Learning (AutoML), especially Automated Deep Learning (AutoDL) systems, but recent work focuses on tabular, image, or NLP tasks. So far, little attention has been paid to general AutoDL frameworks for time series forecasting, despite the enormous success in applying different novel architectures to such tasks. In this paper, we propose an efficient approach for the joint optimization of neural architecture and hyperparameters of the entire data processing pipeline for time series forecasting. In contrast to common NAS search spaces, we designed a novel neural architecture search space covering various state-of-the-art architectures, allowing for an efficient macro-search over different DL approaches. To efficiently search in such a large configuration space, we use Bayesian optimization with multi-fidelity optimization. We empirically study several different budget types enabling efficient multi-fidelity optimization on different forecasting datasets. Furthermore, we compared our resulting system, dubbed \system, against several established baselines and show that it significantly outperforms all of them across several datasets.