Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Zhang, Juyuan, Zhu, Wei, Gao, Jiechao
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2502.13721
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866913698707144704
author Zhang, Juyuan
Zhu, Wei
Gao, Jiechao
author_facet Zhang, Juyuan
Zhu, Wei
Gao, Jiechao
contents Despite the success of Transformer-based models in the time-series prediction (TSP) tasks, the existing Transformer architecture still face limitations and the literature lacks comprehensive explorations into alternative architectures. To address these challenges, we propose AutoFormer-TS, a novel framework that leverages a comprehensive search space for Transformer architectures tailored to TSP tasks. Our framework introduces a differentiable neural architecture search (DNAS) method, AB-DARTS, which improves upon existing DNAS approaches by enhancing the identification of optimal operations within the architecture. AutoFormer-TS systematically explores alternative attention mechanisms, activation functions, and encoding operations, moving beyond the traditional Transformer design. Extensive experiments demonstrate that AutoFormer-TS consistently outperforms state-of-the-art baselines across various TSP benchmarks, achieving superior forecasting accuracy while maintaining reasonable training efficiency.
format Preprint
id arxiv_https___arxiv_org_abs_2502_13721
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Learning Novel Transformer Architecture for Time-series Forecasting
Zhang, Juyuan
Zhu, Wei
Gao, Jiechao
Machine Learning
Computation and Language
Despite the success of Transformer-based models in the time-series prediction (TSP) tasks, the existing Transformer architecture still face limitations and the literature lacks comprehensive explorations into alternative architectures. To address these challenges, we propose AutoFormer-TS, a novel framework that leverages a comprehensive search space for Transformer architectures tailored to TSP tasks. Our framework introduces a differentiable neural architecture search (DNAS) method, AB-DARTS, which improves upon existing DNAS approaches by enhancing the identification of optimal operations within the architecture. AutoFormer-TS systematically explores alternative attention mechanisms, activation functions, and encoding operations, moving beyond the traditional Transformer design. Extensive experiments demonstrate that AutoFormer-TS consistently outperforms state-of-the-art baselines across various TSP benchmarks, achieving superior forecasting accuracy while maintaining reasonable training efficiency.
title Learning Novel Transformer Architecture for Time-series Forecasting
topic Machine Learning
Computation and Language
url https://arxiv.org/abs/2502.13721