<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=us-ascii">
</head>
<body>
<div class="" style="word-wrap:break-word; line-break:after-white-space">
<div class=""><span class="" style="font-size:14.666666984558105px">This is an announcement of Owen Melia's MS Presentation</span><br class="" style="font-size:14.666666984558105px">
<span class="" style="font-size:14.666666984558105px">===============================================</span><br class="" style="font-size:14.666666984558105px">
<span class="" style="font-size:14.666666984558105px">Candidate: Owen Melia</span><br class="" style="font-size:14.666666984558105px">
<br class="" style="font-size:14.666666984558105px">
<span class="" style="font-size:14.666666984558105px">Date: Friday, March 03, 2023</span><br class="" style="font-size:14.666666984558105px">
<br class="" style="font-size:14.666666984558105px">
<span class="" style="font-size:14.666666984558105px">Time: 3:45 pm CST</span><br class="" style="font-size:14.666666984558105px">
<br class="" style="font-size:14.666666984558105px">
<span class="" style="font-size:14.666666984558105px">Location: JCL 223</span><br class="" style="font-size:14.666666984558105px">
<br class="" style="font-size:14.666666984558105px">
<span class="" style="font-size:14.666666984558105px">M.S. Paper Title: ROTATION-INVARIANT RANDOM FEATURES PROVIDE A STRONG BASELINE FOR MACHINE LEARNING ON 3D POINT CLOUDS</span><br class="" style="font-size:14.666666984558105px">
<br class="" style="font-size:14.666666984558105px">
<span class="" style="font-size:14.666666984558105px">Abstract: Rotational invariance is a popular inductive bias used by many fields in machine learning,</span><br class="" style="font-size:14.666666984558105px">
<span class="" style="font-size:14.666666984558105px">such as computer vision and machine learning for quantum chemistry. Rotation-invariant</span><br class="" style="font-size:14.666666984558105px">
<span class="" style="font-size:14.666666984558105px">machine learning methods set the state of the art for many tasks, including molecular property</span><br class="" style="font-size:14.666666984558105px">
<span class="" style="font-size:14.666666984558105px">prediction and 3D shape classification. These methods generally either rely on taskspecific</span><br class="" style="font-size:14.666666984558105px">
<span class="" style="font-size:14.666666984558105px">rotation-invariant features, or they use general-purpose deep neural networks which</span><br class="" style="font-size:14.666666984558105px">
<span class="" style="font-size:14.666666984558105px">are complicated to design and train. We suggest a simple and general-purpose method for</span><br class="" style="font-size:14.666666984558105px">
<span class="" style="font-size:14.666666984558105px">learning rotation-invariant functions of three-dimensional point cloud data using a random</span><br class="" style="font-size:14.666666984558105px">
<span class="" style="font-size:14.666666984558105px">features approach. Specifically, we extend the random features method of Rahimi and Recht</span><br class="" style="font-size:14.666666984558105px">
<span class="" style="font-size:14.666666984558105px">[2007] by deriving a version that is invariant to three-dimensional rotations and showing</span><br class="" style="font-size:14.666666984558105px">
<span class="" style="font-size:14.666666984558105px">that it is fast to evaluate on point cloud data. We show through experiments that our</span><br class="" style="font-size:14.666666984558105px">
<span class="" style="font-size:14.666666984558105px">method matches or outperforms the performance of general-purpose invariant architectures</span><br class="" style="font-size:14.666666984558105px">
<span class="" style="font-size:14.666666984558105px">on standard molecular property prediction benchmark datasets QM7 and QM9. We also</span><br class="" style="font-size:14.666666984558105px">
<span class="" style="font-size:14.666666984558105px">show that our method is competitive with other general-purpose invariant architectures on</span><br class="" style="font-size:14.666666984558105px">
<span class="" style="font-size:14.666666984558105px">the ModelNet40 shape classification task. Finally, we show that our method is an order of</span><br class="" style="font-size:14.666666984558105px">
<span class="" style="font-size:14.666666984558105px">magnitude faster at prediction time than competing kernel methods.</span><br class="" style="font-size:14.666666984558105px">
<br class="" style="font-size:14.666666984558105px">
<span class="" style="font-size:14.666666984558105px">Advisors: Rebecca Willett</span><br class="" style="font-size:14.666666984558105px">
<br class="" style="font-size:14.666666984558105px">
<span class="" style="font-size:14.666666984558105px">Committee Members: Risi Kondor, Eric Jonas, and Rebecca Willett</span><br class="" style="font-size:14.666666984558105px">
<br class="" style="font-size:14.666666984558105px">
</div>
<div class=""></div>
</div>
<div class="" style="word-wrap:break-word; line-break:after-white-space">
<div class=""></div>
<div class=""><br class="">
</div>
<div class=""><br class="">
</div>
<div class=""><br class="">
</div>
<div class=""><br class="">
</div>
<br class="">
</div>
</body>
</html>