BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:ClimaX: A foundation model for weather and climate - Dr Tung Nguye
 n\, University of California Los Angeles
DTSTART:20230620T150000Z
DTEND:20230620T160000Z
UID:TALK199201@talks.cam.ac.uk
CONTACT:Annabelle Scott
DESCRIPTION:Most state-of-the-art approaches for weather and climate model
 ing are based on physics-informed numerical models of the atmosphere. Thes
 e approaches aim to model the non-linear dynamics and complex interactions
  between multiple variables\, which are challenging to approximate. Additi
 onally\, many such numerical models are computationally intensive\, especi
 ally when modeling the atmospheric phenomenon at a fine-grained spatial an
 d temporal resolution. Recent data-driven approaches based on machine lear
 ning instead aim to directly solve a downstream forecasting or projection 
 task by learning a data-driven functional mapping using deep neural networ
 ks. However\, these networks are trained using curated and homogeneous cli
 mate datasets for specific spatiotemporal tasks\, and thus lack the genera
 lity of numerical models. We develop and demonstrate ClimaX\, a flexible a
 nd generalizable deep learning model for weather and climate science that 
 can be trained using heterogeneous datasets spanning different variables\,
  spatio-temporal coverage\, and physical groundings. ClimaX extends the Tr
 ansformer architecture with novel encoding and aggregation blocks that all
 ow effective use of available compute while maintaining general utility. C
 limaX is pre-trained with a self-supervised learning objective on climate 
 datasets derived from CMIP6. The pre-trained ClimaX can then be fine-tuned
  to address a breadth of climate and weather tasks\, including those that 
 involve atmospheric variables and spatio-temporal scales unseen during pre
 training. Compared to existing data-driven baselines\, we show that this g
 enerality in ClimaX results in superior performance on benchmarks for weat
 her forecasting and climate projections\, even when pretrained at lower re
 solutions and compute budgets.
LOCATION:Drum Building\, Madingley Rise Site\, West Cambridge and on zoom:
   https://zoom.us/j/6708259482?pwd=Qk03U3hxZWNJZUZpT2pVZnFtU2RRUT09
END:VEVENT
END:VCALENDAR