13 November (CLIM) Speiser, AI & Earth Systems

Michel Speiser, International Centre for Earth Simulation 

Title: Artificial Intelligence and Systems of the Earth: fundamentals and examples 

November 13 2024, 1:30pm 

Planetary Hall 224 and via Zoom (for link, email lortizur@gmu.edu) 

Host: Jagadesh Shukla 

This talk provides an overview of AI concepts with examples, based on the freshly published open access book Artificial Intelligence and Systems of the Earth. We begin with definitions of artificial intelligence, machine learning, and deep learning, explaining the building blocks of neural networks. We discuss the concepts of overfitting, regularization, dropout, stochastic gradient descent/backpropagation, and some important neural network architectures: convolutional neural networks (CNN) and transformers. Example applications include the recent deep learning models in weather prediction. The second part consists in a brief introduction to causal models, describing the concepts of causal inference and causal discovery, and how they compare to the deep learning paradigm. The illustrative example in this part involves inferring causation from time series in Earth system sciences.