The next C3.ai Digital Transformation Institute Colloquium on Digital Transformation Science Webinar will be Thursday, April 29 at 3:00 p.m. U.S. Central time. Presenting "Understanding Deep Learning through Optimization Bias" will be Nathan Srebro from the Toyota Technological Institute at Chicago.
Registration is required to attend this event.
Abstract: How and why are we succeeding in training huge non-convex deep networks? How can deep neural networks with billions of parameters generalize well, despite not having enough capacity to overfit any data? What is the true inductive bias of deep learning? And, does it all just boil down to a big fancy kernel machine? In this talk, I will highlight the central role the optimization geometry and optimization dynamics play in determining the inductive bias of deep learning, showing how specific optimization methods can allow generalization even in underdetermined overparameterized models.