![MILES team](/sites/default/files/styles/half_width/public/2024-06/Copy%20of%20Gagne_miles_ams_2024-01-30.png?itok=eyjJRZ39)
Artificial intelligence,
natural gains
Welcome to MILES, the Machine Integration and Learning for Earth Systems department at NSF NCAR.
Our mission is to better understand and predict Earth's weather and climate, and their associated impacts, by:
-
Integrating artificial intelligence (AI) and machine learning (ML) throughout Earth systems science.
-
Enhancing the Earth systems science community's AI/ML capabilities.
Our strength is working on problems where AI can be used for social good.
"We want to focus on ML domains that could spawn broad ecosystems of applied research and development."
Focusing on what matters
We are currently working on projects in the following core areas:
AI for Prediction—Generating more accurate weather and climate predictions and projections through AI.
AI for Efficiency—Replacing computationally expensive and slow modeling with AI components, leading to new potential uses for them—for example, in smartphones.
AI for Synthesis—Deploying AI to learn and summarize patterns and processes for a given physical system.
AI for Decision-Making—Using AI to combine information from across different domains, providing decision-makers with guidance and uncertainty information.
![Gabi Pfister](/sites/default/files/styles/half_width/public/2024-05/8C9A3800_Gabi_Pfister%20%281%29%20%281%29.jpg?itok=-bjh0oD-)
Major software packages
echo-opt: Earth Computer Hyperparameter Optimization—A Python package for distributed hyperparameter optimization built on the optuna package.
Hagelslag: Storm Tracking and Verification—A Python package for tracking storms in numerical model output, performing machine learning post-processing for severe weather, and performing probabilistic verification.
miles-guess: A Python package that enables both categorical and regression task evidential learning.
miles-bridgescaler: A Python package that enables turning the properties of a scikit-learn scaler object into a json file and define a new scaler object with those same properties.
Looking for AI help on CISL HPC?
Have a question on basic ML software support on CISL HPC systems? (For example, installing ML libraries or enabling GPU support.) If so, email help@ucar.edu to set up a ticket, or join the NHUG Slack Workspace.
For more detailed questions about AI/ML algorithms, email miles@ucar.edu.