University of Cambridge > Talks.cam > Foundation AI > Enhancing Climate Prediction with Knowledge-Infused Deep Learning Models

  Browse A-Z

Log in

University Account External (via Google)

🔐 Log In

Information on

Subscribing to talksdetails Finding a talkdetails Adding a talkdetails Disseminating talksdetails Help and Documentationdetails

ℹ️ Information

Enhancing Climate Prediction with Knowledge-Infused Deep Learning Models

Download to your calendar using vCal

If you have a question about this talk, please contact Pietro Lio .

Modern deep learning models require large amounts of data and computational resources, often overlooking valuable domain knowledge that could enhance their accuracy and efficiency. In this talk, we present an innovative approach that integrates climate science knowledge into deep learning models to boost performance and uncover meaningful relationships between inputs and outputs. Specifically, we applied a cutting-edge GraphCast-like architecture for long-term climate prediction, enhanced by an attention mechanism that accounts for region-specific climate dynamics, such as the influence of the El Niño southern oscillator. This enables the model to capture non-local interactions with greater expressivity. The resulting hybrid model significantly improves prediction accuracy and highlights key teleconnection patterns, offering deeper insights into how oscillators influence global climate states. This method presents a promising avenue for developing more interpretable and precise climate models.

This talk is part of the Foundation AI series.

This talk is included in these lists:

Note that ex-directory lists are not shown.

Other lists

One O'Clock Research Spotlights (Cambridge Migration Research Network - CAMMIGRES) online via Zoom - contact organiser for details Lecture room 4, Engineering Department

Other talks

The elephant in the room: Machine Learning directly into physical models Title tbc Cambridge RNA Club - IN PERSON Tea and Coffee Break Afternoon Break TBC
 

© 2006-2025 Talks.cam, University of Cambridge. Contact Us | Help and Documentation | Privacy and Publicity