Foundations of Deep Learning Reading Group

Organizers: Michael Kapralov and Ola Svensson.

Recent advances in machine learning have led to breakthroughs in natural language processing, vision and more. Modern deep neural networks, while producing impressive results in practice, present many challenges: they use tremendous computational power, often produce results that are hard to interpret, and raise privacy concerns. The goal of our reading group is to cover some of the recent works that seek to establish a theoretical foundation for deep neural networks that will address these challenges.

Time and place: Wednesdays from 2.15pm to 4pm in BC 04.

Mailing list and calendar: There is a mailing list where the announcements of upcoming lectures will appear. To subscribe to this list, send an email (can be empty) to

List of topics: We will partially follow the excellent course by Boaz Barak on foundations of deep learning.

Tentative schedule: