This repository serves as a curated collection of papers and code related to the application of Large Language Models (LLMs) and Foundation Models (FMs) to Time Series (TS) analysis. It aims to provide researchers and practitioners with a comprehensive overview of the latest advancements, methodologies, and challenges in this rapidly evolving field, enabling better understanding and adoption of these powerful models for TS tasks.

How It Works

The project categorizes research into two main areas: LLMs for Time Series and Foundation Models for Time Series. LLM-based approaches typically involve fine-tuning existing LLMs on TS datasets or adapting them through prompt engineering to perform tasks like forecasting. Foundation Models for TS focus on learning generalizable representations from large TS datasets, independent of LLMs, for transfer learning to downstream applications.

Quick Start & Requirements

This project is a collection of papers and code pointers, not a single executable library. To utilize the code associated with specific papers, users will need to refer to the individual GitHub repositories linked within the README. Requirements will vary per project but generally include Python environments, deep learning frameworks (PyTorch, TensorFlow), and potentially specialized hardware like GPUs for training and inference.

Highlighted Details

• Extensive list of papers applying LLMs to time series forecasting, including prompt-based learning and fine-tuning strategies.
• Coverage of Foundation Models for Time Series, detailing approaches like LOTSA, Timeseries-PILE, and Chronos.
• Includes surveys and position papers offering a broader perspective on LLMs and FMs in time series analysis.
• Related fields section covers LLMs in recommendation systems and tabular data, as well as pre-trained time series models.

Maintenance & Community

The project is maintained by liaoyuhua. Links to specific GitHub repositories for individual papers are provided, allowing users to engage with the respective communities and track development. No central community channels (e.g., Discord, Slack) are listed for this collection itself.

Licensing & Compatibility

The licensing for individual code repositories varies, as indicated by the links to GitHub. Users must consult the license of each specific project they intend to use. Compatibility for commercial use or closed-source linking will depend on the individual licenses of the linked projects.

Limitations & Caveats

This repository is a curated list and does not provide a unified framework or API. Users must navigate individual project repositories for setup, dependencies, and usage. The rapid pace of research means that some linked papers or codebases may become outdated or superseded quickly.