<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=Windows-1252">
<style type="text/css" style="display:none;"> P {margin-top:0;margin-bottom:0;} </style>
</head>
<body dir="ltr">
<span style="font-family: Aptos, Aptos_EmbeddedFont, Aptos_MSFontService, Calibri, Helvetica, sans-serif; font-size: 12pt; color: rgb(0, 0, 0);">We are pleased to announce and invite you to the following joint talk in our ETH-FDS Seminar - Research Seminar
on Statistics:<br>
<br>
„Scaling and Generalizing Approximate Bayesian Inference“<br>
by David Blei, Columbia University<br>
<br>
Date and Time: Thursday, 27 February 2025, 16.15 - 17.15 (Zurich)<br>
Place: ETH Zurich, room tba<br>
<br>
Abstract: A core problem in statistics and machine learning is to approximate difficult-to-compute probability distributions. This problem is especially important in Bayesian statistics, which frames all inference about unknown quantities as a calculation about
a conditional distribution. In this talk I review and discuss innovations in variational inference (VI), a method that approximates probability distributions through optimization. VI has been used in myriad applications in machine learning and Bayesian statistics.
After quickly reviewing the basics, I will discuss two lines of research in VI. I first describe stochastic variational inference, an approximate inference algorithm for handling massive datasets, and demonstrate its application to probabilistic topic models
of millions of articles. Then I discuss black box variational inference, a more generic algorithm for approximating the posterior. Black box inference applies to many models but requires minimal mathematical work to implement. I will demonstrate black box
inference on deep exponential families---a method for Bayesian deep learning---and describe how it enables powerful tools for probabilistic programming. Finally, I will highlight some more recent results in variational inference, including statistical theory,
score-based objective functions, and interpolating between mean-field and fully dependent variational families.<br>
<br>
<br>
Seminar websites: <a href="https://math.ethz.ch/sfs/news-and-events/data-science-seminar.html" id="OWA1279a81a-89ce-b58b-32f1-bcfa049e0437" class="OWAAutoLink" style="text-align: left;">
https://math.ethz.ch/sfs/news-and-events/data-science-seminar.html</a>, <a href="https://math.ethz.ch/sfs/news-and-events/research-seminar.html" id="OWA0f5d5cac-45d8-5d2c-7d74-aadb94c32fa0" class="OWAAutoLink" style="text-align: left;">
https://math.ethz.ch/sfs/news-and-events/research-seminar.html</a><br>
<br>
<br>
<br>
ETH Zürich I Seminar für Statistik I Rämistrasse 101 I 8092 Zürich I Telefon +41 44 632 65 18 I
<a href="mailto:sekretariat@stat.math.ethz.ch" id="OWA56f1bf29-894e-3ebc-e282-12b38c92222a" class="OWAAutoLink" style="text-align: left;">
sekretariat@stat.math.ethz.ch</a></span>
<div style="font-family: Aptos, Aptos_EmbeddedFont, Aptos_MSFontService, Calibri, Helvetica, sans-serif; font-size: 12pt; color: rgb(0, 0, 0);">
<br>
</div>
<div class="elementToProof" style="font-family: Aptos, Aptos_EmbeddedFont, Aptos_MSFontService, Calibri, Helvetica, sans-serif; font-size: 12pt; color: rgb(0, 0, 0);">
<br>
</div>
</body>
</html>