From DNN to 1-NN, and back again: Neyman seminar

Neyman Seminar
Oct 7, 2020 4:00pm to 5:00pm
Location: 
Zoom id: 97648161149. No passcode. Evans Hall
Status: 
Happening As Scheduled
Analyses based on the Empirical Risk Minimization (ERM) and uniform laws of large numbers have long served as the main theoretical foundation for supervised machine learning. Yet, even early on, they had to contend with phenomena such as lack of over-fitting in boosting. Perhaps even more troubling was the fact that one nearest neighbor, the simplest and most intuitive classifier, did not fit...
Misha Belkin, UC San Diego