Mathematics of Deep Learning Seminar: Lenka Zdeborova

Date & Time

Title: Gradient-based algorithms in high-dimensions under limited sample complexity

Abstract: I will overview our recent progress on analyzing gradient descent, Langevin algorithms and stochastic gradient descent in simple neural networks under the teacher-student setting. Our particular focus is on the high-dimensional regime where the number of samples is just a constant times the dimension.

Advancing Research in Basic Science and MathematicsSubscribe to Flatiron Institute announcements and other foundation updates