Are Language Models Actually Useful for Time Series Forecasting？

thuml / AutoTimes

Official implementation for "AutoTimes: Autoregressive Time Series Forecasters via Large Language Models"

https://arxiv.org/abs/2402.02370

MIT License

111 stars 8 forks source link

Are Language Models Actually Useful for Time Series Forecasting？ #14

Closed nightforwar closed 3 weeks ago

nightforwar commented 1 month ago

Hello! I recently read the article Are Language Models Actually Useful for Time Series Forecasting. I also noticed that your experiments included ablation studies from this paper, which demonstrated that your LLM is effective. Could you please explain why the LLM is effective in your architecture? Thank you for your response!

WenWeiTHU commented 1 month ago

Hello, thank you for your interest in our work. In response to your question, we think the main difference of AutoTimes from previous work is the consistency of causality and autoregression. We have recently provided a detailed PPT to illustrate it, and some relevant slides are as follows. If you have more questions, feel free to join the discussion!

nightforwar commented 1 month ago

Thanks！！