[PhilPhys] Friday - Bob Batterman - Deep Learning, Correlations, and the Statistics of Natural Images

[Center for Phil Sci]

The Center for Philosophy of Science invites you to join us for our Lunch
Time Talk.  Attend in person, Room 1117 on the 11th floor of the Cathedral
of Learning at the University of Pittsburgh or visit our live stream on
YouTube at *https://www.youtube.com/channel/UCrRp47ZMXD7NXO3a9Gyh2sg
<https://www.youtube.com/channel/UCrRp47ZMXD7NXO3a9Gyh2sg>*.





*LTT: **Bob Batterman <https://www.hps.pitt.edu/people/robert-batterman>*



Friday, October 18th @ 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/j/94223992589
<https://pitt.zoom.us/j/94223992589>*



YouTube at *https://www.youtube.com/channel/UCrRp47ZMXD7NXO3a9Gyh2sg
<https://www.youtube.com/channel/UCrRp47ZMXD7NXO3a9Gyh2sg>*.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listbox.elte.hu/pipermail/philphys/attachments/20241015/edb598e1/attachment-0001.html>