A decoder-only foundation model for time-series forecasting

Abstract

Motivated by recent advances in large language models for Natural Language Processing (NLP), we design a time-series foundation model for forecasting whose out-of-the-box zero-shot performance on a variety of public datasets comes close to the accuracy of state-of-the-art supervised forecasting models for each individual dataset. Our model is based on pretraining a patched-decoder style attention model on a large time-series corpus, and can work well across different forecasting history lengths, prediction lengths and temporal granularities.

Full paper on arXiv · Code

Details

arXiv identifier

2310.10688

Published

2023-10-14

Authors

Abhimanyu Das, Weihao Kong, Rajat Sen, Yichen Zhou

Models That Cite This Paper

• Described bytimesfm-2.5-200m-pytorch
• Described bytimesfm-3.0-pytorch
• Described bytimesfm-2.0-500m-pytorch
• Described bytimesfm-2.5-200m-transformers