Averaging Weights Leads to Wider Optima and Better Generalization

Averaging Weights Leads to Wider Optima and Better Generalization

This video is a recap of our April 2019 Americas TWiML Online Meetup: Averaging Weights Leads to Wider Optima and Better Generalization.

We had a bit of community discussion before the meetup started, so please skip to (03:25) if you’d like to start from the introduction.

In this month’s community segment, we discuss Compass, Qualcomm AI Day, O’Reilly AI Conference, distorted models for faster inference, and neural network compilers.

In our presentation segment, William Horton leads a discussion on deep learning and training procedures with the “Averaging Weights Leads to Wider Optima and Better Generalization” paper by Pavel Izmailov et al.

For links to the papers, podcasts, and more mentioned above or during this meetup, for more information on previous meetups, or to get registered for upcoming meetups, visit twimlai.com/meetup!

 

Topics mentioned in the community segment:

 

Paper mentioned in the presentation:

Topic: Averaging Weights Leads to Wider Optima and Better Generalization
Presenter: William Horton
Date: Wednesday 17th April 2019
Time: 17:00 PM US Pacific Time