Developing Recursive Feature Machines to Improve Interpretability and Scalability of Seasonal-to-Subseasonal Weather Prediction

Professor: Robert Webber

Description: The grand challenges for AI-driven weather and climate models are the lack of interpretability and the difficulty of optimizing hyperparameters. Popular AI models are powered by deep neural networks containing millions of parameters that produce output for inscrutable reason. The purpose and novelty of the proposed work is to develop and implement efficient AI weather and climate models that are inherently interpretable and easily optimized. 

Preferred Qualifications: Python programming experience