<div dir="ltr"><br><div class="gmail_quote"><br><div dir="ltr"><br><div class="gmail_quote"><br><div dir="ltr"><div><br></div><div>



















<p class="MsoNormal" style="line-height:115%;font-size:11pt;font-family:"Calibri","sans-serif""><b><span style="font-size:12pt;line-height:115%" lang="EN-US">Probability Webinar -<span>   </span>IM-UFRJ <span></span></span></b></p>

<p class="MsoNormal" style="margin:0cm 0cm 10pt;line-height:115%;font-size:11pt;font-family:"Calibri","sans-serif""><span style="font-size:12pt;line-height:115%" lang="EN-US"><span> </span></span></p>

<p class="MsoNormal" style="margin:0cm 0cm 10pt;line-height:115%;font-size:11pt;font-family:"Calibri","sans-serif""><span style="font-size:12pt;line-height:115%" lang="EN-US">Dear colleagues,<span> 
</span><span></span></span></p>

<p class="MsoNormal" style="margin:0cm 0cm 10pt;line-height:115%;font-size:11pt;font-family:"Calibri","sans-serif""><span style="font-size:12pt;line-height:115%" lang="EN-US">Our next online seminar will be held on Monday, <b>November 30</b>, from <b>3 p.m. to 4 p.m</b>.
(Rio de Janeiro local time)<span></span></span></p>

<p class="MsoNormal" style="margin:0cm 0cm 10pt;line-height:115%;font-size:11pt;font-family:"Calibri","sans-serif""><span style="font-size:12pt;line-height:115%" lang="EN-US">The GoogleMeet link for the seminars is: </span><a href="https://meet.google.com/nxh-optr-wtq" style="color:blue;text-decoration:underline" target="_blank"><span style="font-size:12pt;line-height:115%" lang="EN-US">https://meet.google.com/nxh-optr-wtq</span></a><span style="font-size:12pt;line-height:115%" lang="EN-US"> 
<br>
<br>
<span></span></span></p>

<p class="MsoNormal" style="margin:0cm 0cm 10pt;line-height:115%;font-size:11pt;font-family:"Calibri","sans-serif""><span style="font-size:12pt;line-height:115%" lang="EN-US">Speaker: <span> </span><b>Paulo
Orenstein (IMPA) </b><span></span></span></p>

<p class="MsoNormal" style="margin:0cm 0cm 10pt;line-height:115%;font-size:11pt;font-family:"Calibri","sans-serif""><span style="font-size:12pt;line-height:115%" lang="EN-US">Title:</span><span style="font-size:12pt;line-height:115%" lang="EN-US"> <b>Importance
sampling with adaptive winsorization</b><span></span></span></p>

<p class="MsoNormal" style="line-height:115%;font-size:11pt;font-family:"Calibri","sans-serif""><span style="font-size:12pt;line-height:115%" lang="EN-US">Abstract:<b> </b></span><span style="font-size:12pt;line-height:115%" lang="EN-US"> </span><span lang="EN-US">Importance sampling is a widely used
technique to estimate the properties of a distribution. The resulting estimator
is always unbiased, but may sometimes incur huge or infinite variance. This
work investigates trading-off some bias for variance by winsorizing the
importance sampling estimator using an adaptive thresholding procedure based on
the Balancing Principle (also known as Lepskii's Method). This provides a
principled way to perform winsorization, with finite-sample optimality
guarantees and good empirical performance.<span></span></span></p>

<p class="MsoNormal" style="line-height:115%;font-size:11pt;font-family:"Calibri","sans-serif""><span style="font-size:12pt;line-height:115%" lang="EN-US"><span> </span></span></p>

<p class="MsoNormal" style="line-height:115%;font-size:11pt;font-family:"Calibri","sans-serif""><span style="font-size:12pt;line-height:115%" lang="EN-US">All the talks
are held in English. <span></span></span></p>

<p class="MsoNormal" style="margin:0cm 0cm 10pt;line-height:115%;font-size:11pt;font-family:"Calibri","sans-serif""><span style="font-size:12pt;line-height:115%" lang="EN-US">Thanks for circulating this information. <span></span></span></p>

<p class="MsoNormal" style="margin:0cm 0cm 10pt;line-height:115%;font-size:11pt;font-family:"Calibri","sans-serif""><span style="font-size:12pt;line-height:115%" lang="EN-US">Sincerely, </span></p><p class="MsoNormal" style="margin:0cm 0cm 10pt;line-height:115%;font-size:11pt;font-family:"Calibri","sans-serif""><span style="font-size:12pt;line-height:115%" lang="EN-US">
Organizers: Guilherme Ost and Maria Eulalia Vares<span></span></span></p>





</div></div>
</div></div>
</div></div>