BEGIN:VCALENDAR
VERSION:2.0
PRODID:IEEE vTools.Events//EN
CALSCALE:GREGORIAN
BEGIN:VTIMEZONE
TZID:America/New_York
BEGIN:DAYLIGHT
DTSTART:20260308T030000
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3
TZNAME:EDT
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20251102T010000
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11
TZNAME:EST
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260313T160054Z
UID:57DCFA9C-9A3D-4F4B-94A4-6858CB104C1A
DTSTART;TZID=America/New_York:20260123T110000
DTEND;TZID=America/New_York:20260123T120000
DESCRIPTION:Title: A Stein Gradient Descent Approach for Doubly Intractable
  Distributions\n\nDr. Jaewoo Park\n\nAssistant Professor\, Department of A
 pplied Statistics\n\nYonsei University\n\n  \n\nFriday\, January 23rd\
 , 2026 \n\n11:00 A.M. – 12:00 P.M. Eastern Time \n\nNguyen Engineeri
 ng Building\, Jajodia Auditorium\, Room 1101 \n\n4511 Patriot Circle\, F
 airfax\, Virginia 22030\n\nThe seminar talk is also live-streamed. Ple
 ase [register](https://nam11.safelinks.protection.outlook.com/?url=https%3
 A%2F%2Fforms.office.com%2Fr%2FFdtuvURVnG&amp;data=05%7C02%7Ckhassan1%40gmu.edu
 %7C9d2413adcdf240ddb98d08de553b1ad5%7C9e857255df574c47a0c00546460380cb%7C0
 %7C0%7C639041909517347832%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUs
 IlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7
 C%7C%7C&amp;sdata=aOO5Tm3p7NOoO2TDVFDLi9NF%2BdtTiXudvBn68S2X%2FLU%3D&amp;reserved=
 0) to receive the link.\n\n \n\nAbstract: \n\nBayesian inference for d
 oubly intractable distributions is challenging because they include intrac
 table terms\, which are functions of parameters of interest. Although seve
 ral alternatives have been developed for such models\, they are computatio
 nally intensive due to repeated auxiliary variable simulations. We propose
  a novel Monte Carlo Stein variational gradient descent (MC-SVGD) approach
  for inference for doubly intractable distributions. Through an efficient 
 gradient approximation\, our MC-SVGD approach rapidly transforms an arbitr
 ary reference distribution to approximate the posterior distribution of in
 terest\, without necessitating any predefined variational distribution cla
 ss for the posterior. Such a transport map is obtained by minimizing Kullb
 ack-Leibler divergence between the transformed and posterior distributions
  in a reproducing kernel Hilbert space (RKHS). We also investigate the con
 vergence rate of the proposed method. We illustrate the application of the
  method to challenging examples\, including a Potts model\, an exponential
  random graph model\, and a Conway--Maxwell--Poisson regression model. The
  proposed method achieves substantial computational gains over existing al
 gorithms\, while providing comparable inferential performance for the post
 erior distributions.\n\nBio: \n\nJaewoo Park is an Assistant Professor i
 n the Department of Applied Statistics at Yonsei University. His research 
 focuses on computational methods for intractable likelihoods\, Bayesian mo
 deling for spatio-temporal data\, and spatial functional data analysis.\n\
 nRoom: Jajodia Auditorium\, Room 1101  \, Bldg: Nguyen Engineering Build
 ing\, \, George Mason University\, Fairfax\, Virginia\, United States\, 22
 030
LOCATION:Room: Jajodia Auditorium\, Room 1101  \, Bldg: Nguyen Engineerin
 g Building\, \, George Mason University\, Fairfax\, Virginia\, United Stat
 es\, 22030
ORGANIZER:kafi@ieee.org
SEQUENCE:19
SUMMARY:A Stein Gradient Descent Approach for Doubly Intractable Distributi
 ons
URL;VALUE=URI:https://events.vtools.ieee.org/m/533228
X-ALT-DESC:Description: &lt;br /&gt;&lt;p class=&quot;x_elementtoproof&quot;&gt;&lt;strong&gt;&lt;em&gt;&lt;span
  data-olk-copy-source=&quot;MessageBody&quot;&gt;Title: A Stein Gradient Descent Approa
 ch for Doubly Intractable Distributions&lt;/span&gt;&lt;/em&gt;&lt;/strong&gt;&lt;/p&gt;\n&lt;p class
 =&quot;x_elementtoproof&quot;&gt;&lt;strong&gt;Dr. Jaewoo Park&lt;/strong&gt;&lt;/p&gt;\n&lt;p class=&quot;x_MsoN
 ormal&quot;&gt;&lt;strong&gt;Assistant Professor\, Department of Applied Statistics&lt;/str
 ong&gt;&lt;/p&gt;\n&lt;p class=&quot;x_MsoNormal&quot;&gt;&lt;strong&gt;Yonsei University&lt;/strong&gt;&lt;/p&gt;\n&lt;
 p class=&quot;x_MsoNormal&quot;&gt;  &amp;nbsp\;&lt;/p&gt;\n&lt;p class=&quot;x_elementtoproof&quot;&gt;&lt;stro
 ng&gt;Friday\, January 23rd\, 2026&lt;/strong&gt; &amp;nbsp\;&lt;/p&gt;\n&lt;p class=&quot;x_elemen
 ttoproof&quot;&gt;&lt;strong&gt;11:00 A.M. &amp;ndash\; 12:00 P.M. Eastern Time&lt;/strong&gt; &amp;
 nbsp\;&lt;/p&gt;\n&lt;p class=&quot;x_elementtoproof&quot;&gt;&lt;strong&gt;Nguyen Engineering Buildin
 g\, Jajodia Auditorium\, Room 1101&lt;/strong&gt; &amp;nbsp\;&lt;/p&gt;\n&lt;p class=&quot;x_ele
 menttoproof&quot;&gt;&lt;strong&gt;4511 Patriot Circle\, Fairfax\, Virginia 22030&lt;/stron
 g&gt;&amp;nbsp\;&lt;/p&gt;\n&lt;p class=&quot;x_elementtoproof&quot;&gt;&lt;strong&gt;The seminar talk is
  also live-streamed. Please &lt;/strong&gt;&lt;a title=&quot;Original URL: https://forms
 .office.com/r/FdtuvURVnG. Click or tap if you trust this link.&quot; href=&quot;http
 s://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fforms.office
 .com%2Fr%2FFdtuvURVnG&amp;amp\;data=05%7C02%7Ckhassan1%40gmu.edu%7C9d2413adcdf
 240ddb98d08de553b1ad5%7C9e857255df574c47a0c00546460380cb%7C0%7C0%7C6390419
 09517347832%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMD
 AwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C&amp;amp\;s
 data=aOO5Tm3p7NOoO2TDVFDLi9NF%2BdtTiXudvBn68S2X%2FLU%3D&amp;amp\;reserved=0&quot; t
 arget=&quot;_blank&quot; rel=&quot;noopener noreferrer&quot; data-auth=&quot;NotApplicable&quot; data-li
 nkindex=&quot;0&quot;&gt;register&lt;/a&gt;&lt;strong&gt; to receive the link.&lt;/strong&gt;&amp;nbsp\;&lt;/p&gt;\
 n&lt;p class=&quot;x_elementtoproof&quot;&gt;&lt;strong&gt; &lt;/strong&gt;&amp;nbsp\;&lt;/p&gt;\n&lt;p class=&quot;x_
 elementtoproof&quot;&gt;&lt;strong&gt;Abstract:&lt;/strong&gt; &amp;nbsp\;&lt;/p&gt;\n&lt;p class=&quot;x_elem
 enttoproof&quot;&gt;Bayesian inference for doubly intractable distributions is cha
 llenging because they include intractable terms\, which are functions of p
 arameters of interest. Although several alternatives have been developed f
 or such models\, they are computationally intensive due to repeated auxili
 ary variable simulations. We propose a novel Monte Carlo Stein variational
  gradient descent (MC-SVGD) approach for inference for doubly intractable 
 distributions. Through an efficient gradient approximation\, our MC-SVGD a
 pproach rapidly transforms an arbitrary reference distribution to approxim
 ate the posterior distribution of interest\, without necessitating any pre
 defined variational distribution class for the posterior. Such a transport
  map is obtained by minimizing Kullback-Leibler divergence between the tra
 nsformed and posterior distributions in a reproducing kernel Hilbert space
  (RKHS). We also investigate the convergence rate of the proposed method. 
 We illustrate the application of the method to challenging examples\, incl
 uding a Potts model\, an exponential random graph model\, and a Conway--Ma
 xwell--Poisson regression model. The proposed method achieves substantial 
 computational gains over existing algorithms\, while providing comparable 
 inferential performance for the posterior distributions.&lt;/p&gt;\n&lt;p class=&quot;x_
 elementtoproof&quot;&gt;&amp;nbsp\;&lt;/p&gt;\n&lt;p class=&quot;x_elementtoproof&quot;&gt;&lt;strong&gt;Bio:&lt;/str
 ong&gt; &amp;nbsp\;&lt;/p&gt;\n&lt;p class=&quot;x_elementtoproof&quot;&gt;Jaewoo Park is an Assistan
 t Professor in the Department of Applied Statistics at Yonsei University. 
 His research focuses on computational methods for intractable likelihoods\
 , Bayesian modeling for spatio-temporal data\, and spatial functional data
  analysis.&lt;/p&gt;
END:VEVENT
END:VCALENDAR