<div dir="ltr">And, to provide icing on the cake of multiple excellent Auton presentations next week, please mark your calendars for Ifi's doctoral thesis proposal talk on Wednesday.<div><br></div><div>Cheers</div><div>Artur<br><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">---------- Forwarded message ---------<br>From: <strong class="gmail_sendername" dir="auto">Diane Stidle</strong> <span dir="auto"><<a href="mailto:stidle@andrew.cmu.edu">stidle@andrew.cmu.edu</a>></span><br>Date: Fri, Apr 28, 2023 at 2:38 PM<br>Subject: Thesis Proposal - May 3, 2023 - Ifigeneia Apostolopoulou - Decision-Making Under Latent Factors<br>To: <a href="mailto:ml-seminar@cs.cmu.edu">ml-seminar@cs.cmu.edu</a> <<a href="mailto:ML-SEMINAR@cs.cmu.edu">ML-SEMINAR@cs.cmu.edu</a>>, <<a href="mailto:fusi@microsoft.com">fusi@microsoft.com</a>>, <<a href="mailto:doucet@stats.ox.ac.uk">doucet@stats.ox.ac.uk</a>>, <<a href="mailto:arnaud-doucet@google.com">arnaud-doucet@google.com</a>><br></div><br><br>
<div>
<p><i><b>Thesis Proposal</b></i></p>
<p>Date: May 3, 2023<br>
Time: 3:30pm (EST) (Remote only)<br>
Speaker: Ifigeneia Apostolopoulou</p>
<p><b>Title: </b><b>Decision-Making Under Latent Factors</b></p>
<p>Abstract:<br>
Complex data arise in many fields including health care,
neuroscience, computer vision, and many others. However, they
often exhibit simple, yet unobserved patterns. These patterns can
be represented by latent variables that augment the observations
in the system of interest. This thesis surveys how to use latent
variable models to guide decision-making. We approach
decision-making as a workflow of three cognitive tasks: i)
Representation Learning, ii) Temporal Modeling, iii)
Uncertainty-Aware Reasoning. We propose that the processing
ability required for each task can be improved by the use of
latent variable models. To demonstrate that, we first discuss the
types of patterns that latent variables need to capture for each
task. Further, we develop inference tools that allow for
expressive posterior distributions. Finally, our empirical
analysis demonstrates the superior performance of latent variable
models on several machine learning problems compared to their
counterparts that operate purely in the observed data space.<br>
</p>
<div><br>
</div>
<div><b>Thesis Committee: </b></div>
<div>Artur Dubrawski (Chair)<br>
</div>
<div>
<div>Ruslan Salakhutdinov </div>
</div>
<div>
<div>Tom Mitchell</div>
</div>
<div>Nicolo Fusi (Microsoft) </div>
<div>Arnaud Doucet (Google Deepmind)</div>
<div><br>
</div>
<div>Zoom Link: <a href="https://cmu.zoom.us/j/96935140004?pwd=akNlbmU4TnRPWWlmWE5wWGIxaU5MZz09" target="_blank">https://cmu.zoom.us/j/96935140004?pwd=akNlbmU4TnRPWWlmWE5wWGIxaU5MZz09</a></div>
<div><br>
<br>
</div>
<pre cols="72">--
Diane Stidle
PhD Program Manager
Machine Learning Department
Carnegie Mellon University
<a href="mailto:stidle@andrew.cmu.edu" target="_blank">stidle@andrew.cmu.edu</a></pre>
</div>
</div></div></div>