<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1">
<meta name="viewport" content="width=device-width">
<meta content="IE=edge">
<meta content="text/html; charset=UTF-8">
<style type="text/css">
<!--
html
        {box-sizing:border-box;
        font-size:14px;
        margin:0}
table
        {border-spacing:0}
td
        {vertical-align:top;
        margin:0}
img
        {max-width:100%}
h1, h2, h3, h4, h5
        {font-weight:600;
        line-height:1.4}
.css-1om2ube
        {width:74px;
        height:18px;
        margin:0 0 0 5px;
        top:3px}
.css-13nbghr
        {width:100%;
        height:100%;
        line-height:1.6em;
        color:#333333;
        background-color:#fafafa;
        font-weight:400}
.css-18xjf65
        {margin:0 auto;
        clear:both;
        font-size:16px;
        padding:20px;
        width:720px}
.css-y9bb7n
        {background:#fff;
        font-size:16px;
        line-height:24px;
        word-wrap:break-word;
        padding:40px;
        border-radius:6px}
.css-40d1fs
        {font-size:12px;
        color:#888;
        margin:0;
        vertical-align:baseline;
        padding:20px 40px}
-->
</style>
</head>
<body itemscope="" itemtype="http://schema.org/EmailMessage" class="css-13nbghr" style="width:100%; height:100%; line-height:1.6em; color:#333333; background-color:#fafafa; font-weight:400">
<strong>
<div><font face="Tahoma" color="#000000" size="2"> </font></div>
</strong>
<hr tabindex="-1" style="display:inline-block; width:98%">
<font face="Tahoma" size="2"><b>From:</b> noreply+automations@airtableemail.com <noreply+automations@airtableemail.com>On Behalf OfTheoryBot (via Airtable) <noreply+automations@airtableemail.com><br>
<b>Sent:</b> Wednesday, April 12, 2023 12:00:53 AM (UTC-06:00) Central Time (US & Canada)<br>
<b>To:</b> Antares Chen <antaresc@uchicago.edu><br>
<b>Cc:</b> Christopher Kang <ctkang@uchicago.edu><br>
<b>Subject:</b> Theory Lunch 2023-04-12T17:30:00.000Z<br>
</font><br>
<div></div>
<div>
<div></div>
<table class="container css-18xjf65" style="border-spacing:0; margin:0 auto; clear:both; font-size:16px; padding:20px; width:720px">
<tbody>
<tr>
<td colspan="2" class="content  css-y9bb7n" style="vertical-align:top; margin:0; background:#fff; font-size:16px; line-height:24px; word-wrap:break-word; padding:40px; border-radius:6px">
<p style="margin-top:0"><span>Today's Theory Lunch talk:</span></p>
<p><em><span>Andrew Hands (University of Chicago): Permutation Equivariance in Higher Order Message Passing Networks and Beyond</span></em></p>
<p><span><a href="https://urldefense.com/v3/__https://uchicago.zoom.us/j/91616319229?pwd=dDdXQnFXeGNubFRkZy9hTDQrcWlXdz09__;!!BpyFHLRN4TMTrA!8iS5MioaE0jVsQX68yYWBAJ1ntFBPZ51Vdz2sOtKwv--OBq6P4wh7keNawXz4wEIX45gCJfKQtzLztw4WwfChUFe3zOW1xWenpY$">https://uchicago.zoom.us/j/91616319229?pwd=dDdXQnFXeGNubFRkZy9hTDQrcWlXdz09</a></span></p>
<p></p>
<p><span>Description: Message Passing Neural Networks (MPNNs) have proven to be a powerful tool for learning functions on graphs; however, they are inherently limited as they only propagate information across graphs in a purely local and isotropic fashion.
 Thus preventing them from being able to distinguish even simple structures such as k-regular graphs. To help address this issue, we introduce Permutation Equivariant Higher Graph Neural Networks, which are able to consider higher order message passing between
 not just vertices, but subgraphs as well. This allows us to generate a richer higher order representation, while still maintaining a low time complexity for lower density graphs, such as those arising from molecular learning.</span></p>
</td>
</tr>
<tr>
<td colspan="2" class="footer css-40d1fs" style="font-size:12px; color:#888; margin:0; vertical-align:baseline; padding:20px 40px">
<table style="border-spacing:0">
<tbody>
<tr>
<td style="vertical-align:top; margin:0">Sent via Automations on </td>
<td style="vertical-align:top; margin:0"><a href="https://urldefense.com/v3/__https://airtable.com?utm_medium=email&utm_source=product_team&utm_content=transactional-alerts__;!!BpyFHLRN4TMTrA!8iS5MioaE0jVsQX68yYWBAJ1ntFBPZ51Vdz2sOtKwv--OBq6P4wh7keNawXz4wEIX45gCJfKQtzLztw4WwfChUFe3zOWzlW_jPs$" style="color:#3276dc"><img src="https://static.airtable.com/images/type_logo@116h.png?v=3" width="74" height="18" alt="Airtable" class="css-1om2ube" style="max-width:100%; width:74px; height:18px; margin:0 0 0 5px; top:3px"></a></td>
</tr>
<tr>
<td colspan="2" style="vertical-align:top; margin:0">
<div>©2023 Airtable</div>
</td>
</tr>
</tbody>
</table>
</td>
</tr>
</tbody>
</table>
</div>
</body>
</html>