Model Theory Seminar

March 21
Vince Guingona, Towson University
Statistical Learning and Model Theory

In this talk, I explore the connections between Statistical Learning Theory and Model Theory. This includes the connections between PAC-learning and NIP and the connections between differentially private PAC-learning and stability. Finally, I examine the work that my colleagues and I have started on improving the sample complexity of differentially private PAC-learning algorithms using techniques from stability theory. This work is joint with Alexei Kolesnikov, Miriam Parnes, and Natalie Piltoyan.