Robert Batterman - Deep Learning, Correlations, and the Statistics of Natural Images
null, Robert W. Batterman (University of Pittsburgh)

October 18, 2024, 12:00pm - 1:00pm
The Center for Philosophy of Science, University of Pittsburgh

1117 Cathedral of Learning - 11th Floor
University of Pittsburgh, 4200 Fifth Avenue
Pittsburgh 15260
United States

This will be an accessible event, including organized related activities

This event is available both online and in-person

Organisers:

University of Pittsburgh

Details

The Center for Philosophy of Science at the University of Pittsburgh invites you to join us for our Lunch Time Talk. Attend in person at 1117 Cathedral of Learning or visit our live stream on YouTube at https://www.youtube.com/channel/UCrRp47ZMXD7NXO3a9Gyh2sg.

LTT:  Robert Batterman

Friday, October 18 @ 12:00 pm - 1:30 pm EDT

Title: Deep Learning, Correlations, and the Statistics of Natural Images

Abstract: In this paper I will highlight certain connections between Deep Neural Networks (DNNs) and  Renormalization Group methods in statistical physics.  I will suggest that some of these similarities can be exploited to understand why some DNNs can be so successful.  Specifically, I will suggest that DNNs generalize as well as they do because their training allows them to discover/recover certain remarkably robust statistical regularities present in their (diverse) training data sets. 

Can’t make it in-person? This talk will available online through the following:

Zoom:  https://pitt.zoom.us/s/94223992589

YouTube at https://www.youtube.com/channel/UCrRp47ZMXD7NXO3a9Gyh2sg.

Supporting material

Add supporting material (slides, programs, etc.)

Reminders

Registration

No

Who is attending?

No one has said they will attend yet.

Will you attend this event?


Let us know so we can notify you of any change of plan.