Organized by: Michael R. Douglas, Sergei Gukov, Jim Halverson, Sven Krippendorf, Fabian Ruehle, Giancarlo La Camera, Luca Mazzucato, Jin WangAttendee List
Live video may be available, please take a look at http://scgp.stonybrook.edu/live.
The availability of very large datasets and the striking progress in artificial intelligence are revolutionizing the way scientists approach their disciplines. The deployment of state-of-the-art techniques in machine learning and statistical inference to study large datasets is leading to unprecedented discoveries and narrowing the gap between physics, mathematics, biology, computer science, and statistics. Despite this practical success, little is known about the general principles governing neural networks learning and dynamics and the geometry of data manifolds. The Simons Program on “Neural networks and the Data Science Revolution” will bring together researchers from the theoretical physics, artificial intelligence, and computational neuroscience communities to discuss foundational and theoretical aspects of neural networks, and highlight challenges and opportunities in their application to specific open problems along an axis of topics encompassing string theory, machine learning, big data science, and brain science. The program will begin with a workshop on the interface between theoretical physics, geometry and data science, with a focus on synthetic datasets arising in the classification of manifolds and knots, quantum and statistical field theories, and vacuum configurations of string theory. The program will end with a workshop on the physics of neural circuits, discussing ways to bridge the gap between neural network models and the large experimental datasets nowadays available in neuroscience.