<div dir="ltr"><div class="gmail_quote"><div dir="ltr"><p style="margin:0in;color:rgb(36,36,36);font-family:Aptos,sans-serif;font-size:12pt"><span style="border:0px;font:inherit;margin:0px;padding:0px;vertical-align:baseline;color:black">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 </span><span style="border:0px;font:inherit;margin:0px;padding:0px;vertical-align:baseline;color:blue"><u><a title="https://www.youtube.com/channel/UCrRp47ZMXD7NXO3a9Gyh2sg" id="m_2669352837699294028gmail-OWA6b7528ff-70ed-d9f3-6489-e752d447f8ba" rel="noopener noreferrer" href="https://www.youtube.com/channel/UCrRp47ZMXD7NXO3a9Gyh2sg" style="border:0px;font:inherit;margin:0px;padding:0px;vertical-align:baseline;color:blue" target="_blank">https://www.youtube.com/channel/UCrRp47ZMXD7NXO3a9Gyh2sg</a></u></span><span style="border:0px;font:inherit;margin:0px;padding:0px;vertical-align:baseline;color:black">.    </span></p><p style="margin:0in;color:rgb(36,36,36);font-family:Aptos,sans-serif;font-size:12pt"><span style="border:0px;font:inherit;margin:0px;padding:0px;vertical-align:baseline;color:black"> </span></p><p style="margin:0in;color:rgb(36,36,36);font-family:Aptos,sans-serif;font-size:12pt"><span style="border:0px;font:inherit;margin:0px;padding:0px;vertical-align:baseline;color:black"> </span></p><p style="margin:0in;color:rgb(36,36,36);font-family:Aptos,sans-serif;font-size:12pt"><span style="border:0px;font:inherit;margin:0px;padding:0px;vertical-align:baseline;color:black"><b>LTT: </b></span><span style="border:0px;font-style:inherit;font-variant:inherit;font-weight:inherit;font-stretch:inherit;font-size:14pt;line-height:inherit;font-family:inherit;font-size-adjust:inherit;font-kerning:inherit;font-feature-settings:inherit;margin:0px;padding:0px;vertical-align:baseline;color:blue"><b><u><a title="https://sites.pitt.edu/~dmw121/" id="m_2669352837699294028gmail-OWA774f43cd-2025-8951-60cb-2f43a85b446d" rel="noopener noreferrer" href="https://www.hps.pitt.edu/people/robert-batterman" style="border:0px;font:inherit;margin:0px;padding:0px;vertical-align:baseline;color:blue" target="_blank">Bob Batterman</a></u></b></span></p><p style="margin:0in;color:rgb(36,36,36);font-family:Aptos,sans-serif;font-size:12pt"><span style="border:0px;font:inherit;margin:0px;padding:0px;vertical-align:baseline;color:black"> </span></p><p style="margin:0in;color:rgb(36,36,36);font-family:Aptos,sans-serif;font-size:12pt"><span style="border:0px;font:inherit;margin:0px;padding:0px;vertical-align:baseline;color:black">Friday, </span>October 18th<span style="border:0px;font:inherit;margin:0px;padding:0px;vertical-align:baseline;color:black"> @ 12:00 pm - 1:30 pm E</span>DT</p><p style="margin:0in;color:rgb(36,36,36);font-family:Aptos,sans-serif;font-size:12pt"><span style="border:0px;font:inherit;margin:0px;padding:0px;vertical-align:baseline;color:black"><b> </b></span></p><p style="margin:0in;color:rgb(36,36,36);font-family:Aptos,sans-serif;font-size:12pt"><span style="border:0px;font:inherit;margin:0px;padding:0px;vertical-align:baseline;color:black"><b>Title: </b></span><b>Deep Learning, Correlations, and the Statistics of Natural Images</b></p><p style="margin:0in;color:rgb(36,36,36);font-family:Aptos,sans-serif;font-size:12pt"><span style="border:0px;font:inherit;margin:0px;padding:0px;vertical-align:baseline;color:black"><b> </b></span></p><p style="margin:0in;color:rgb(36,36,36);font-family:Aptos,sans-serif;font-size:12pt"><span style="border:0px;font:inherit;margin:0px;padding:0px;vertical-align:baseline;color:black"><b>Abstract:</b></span></p><p style="margin:0in;color:rgb(36,36,36);font-family:Aptos,sans-serif;font-size:12pt">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. </p><p style="margin:0in;color:rgb(36,36,36);font-family:Aptos,sans-serif;font-size:12pt"><b> </b></p><p style="margin:0in;color:rgb(36,36,36);font-family:Aptos,sans-serif;font-size:12pt"><b>Can’t make it in-person? This talk will available online through the following:</b></p><p style="margin:0in;color:rgb(36,36,36);font-family:Aptos,sans-serif;font-size:12pt"><span style="border:0px;font:inherit;margin:0px;padding:0px;vertical-align:baseline;color:black"> </span></p><p style="margin:0in;color:rgb(36,36,36);font-family:Aptos,sans-serif;font-size:12pt">Zoom:  <span style="border:0px;font:inherit;margin:0px;padding:0px;vertical-align:baseline;color:blue"><u><a id="m_2669352837699294028gmail-OWAc32accef-1fc5-8732-4185-4451084ae3e9" rel="noopener noreferrer" href="https://pitt.zoom.us/j/94223992589" style="border:0px;font:inherit;margin:0px;padding:0px;vertical-align:baseline;color:blue" target="_blank">https://pitt.zoom.us/j/94223992589</a></u></span></p><p style="margin:0in;color:rgb(36,36,36);font-family:Aptos,sans-serif;font-size:12pt"><span style="border:0px;font-style:inherit;font-variant:inherit;font-weight:inherit;font-stretch:inherit;font-size:11pt;line-height:inherit;font-family:inherit;font-size-adjust:inherit;font-kerning:inherit;font-feature-settings:inherit;margin:0px;padding:0px;vertical-align:baseline;color:inherit"> </span></p><p style="margin:0in;color:rgb(36,36,36);font-family:Aptos,sans-serif;font-size:12pt"><span style="border:0px;font:inherit;margin:0px;padding:0px;vertical-align:baseline;color:black">YouTube at </span><span style="border:0px;font:inherit;margin:0px;padding:0px;vertical-align:baseline;color:blue"><u><a id="m_2669352837699294028gmail-OWA12d6fe41-bd77-89dc-09ef-9e13222ee5c5" rel="noopener noreferrer" href="https://www.youtube.com/channel/UCrRp47ZMXD7NXO3a9Gyh2sg" style="border:0px;font:inherit;margin:0px;padding:0px;vertical-align:baseline;color:blue" target="_blank">https://www.youtube.com/channel/UCrRp47ZMXD7NXO3a9Gyh2sg</a></u></span><span style="border:0px;font:inherit;margin:0px;padding:0px;vertical-align:baseline;color:black">.</span></p><p style="margin:0in;color:rgb(36,36,36);font-family:Aptos,sans-serif;font-size:12pt"><span style="border:0px;font-style:inherit;font-variant:inherit;font-weight:inherit;font-stretch:inherit;font-size:11pt;line-height:inherit;font-family:inherit;font-size-adjust:inherit;font-kerning:inherit;font-feature-settings:inherit;margin:0px;padding:0px;vertical-align:baseline;color:inherit"> </span></p></div>
</div></div>