search

KimMeen / Time-LLM

[ICLR 2024] Official implementation of " 🦙 Time-LLM: Time Series Forecasting by Reprogramming Large Language Models"
https://arxiv.org/abs/2310.01728
Apache License 2.0
1.46k stars 252 forks source link

LLMs for anomaly detection in the field of gravitational waves astronomy #128

Closed rruizdeaustri closed 3 months ago

rruizdeaustri commented 3 months ago

Hello,

I hope this message finds you well.

I am interested in exploring the application of Large Language Models (LLMs) technology for anomaly detection in the field of gravitational waves astronomy. Our work involves multimodal time series data, and we aim to enhance our detection capabilities by leveraging advanced machine learning techniques.

The proposed approach involves training a model on pure noise data with intentional gaps, allowing the model to learn how to accurately complete the series. Subsequently, we would introduce actual signal data with the expectation that the model would assign higher anomaly scores to these signals compared to the noise data.

I believe your framework could be instrumental in implementing this idea, providing a valuable proof of concept for the community. I have two questions:

Do you think this application could be a viable use case for your framework? Is there an existing script or set of tools within your framework that I could use to process my data accordingly? Thank you very much for your assistance.

Best regards, Roberto

kwuking commented 3 months ago

Do you think this application could be a viable use case for your framework? Is there an existing script or set of tools within your framework that I could use to process my data accordingly? Thank you very much for your assistance.

Thank you for your attention. I believe your idea is feasible. For Time-LLM, it’s enough to modify the output layer to turn it into a general framework capable of handling temporal comprehensive analysis tasks. You can refer to the relevant code in TSLib for the necessary modifications.