Intelligent Systems for Climate Research: When Will Deep Learners Meet Deep Knowledge?

Yolanda Gil, University of Southern California

Machine learning has set up a beachhead for AI in climate.  Although, learning algorithms are making new discoveries, using sophisticated techniques to exploit the data that is available, the phenomena under study are so complex that learning from data alone has limited possibilities.  We need to supplement data with knowledge about geoscience processes that can improve learning outcomes.  In this talk, I will describe how research in intelligent systems can advance geosciences research, not only by improving machine learning performance but also by making data collection more effective, data processing less time consuming, and data understanding more accessible.  I will also describe a new research project on automating machine learning through semantic representations.

No Recording Available