<html xmlns:v="urn:schemas-microsoft-com:vml" xmlns:o="urn:schemas-microsoft-com:office:office" xmlns:w="urn:schemas-microsoft-com:office:word" xmlns:m="http://schemas.microsoft.com/office/2004/12/omml" xmlns="http://www.w3.org/TR/REC-html40">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=us-ascii">
<meta name="Generator" content="Microsoft Word 15 (filtered medium)">
<!--[if !mso]><style>v\:* {behavior:url(#default#VML);}
o\:* {behavior:url(#default#VML);}
w\:* {behavior:url(#default#VML);}
.shape {behavior:url(#default#VML);}
</style><![endif]--><style><!--
/* Font Definitions */
@font-face
{font-family:"Cambria Math";
panose-1:2 4 5 3 5 4 6 3 2 4;}
@font-face
{font-family:Calibri;
panose-1:2 15 5 2 2 2 4 3 2 4;}
@font-face
{font-family:"Calibri Light";
panose-1:2 15 3 2 2 2 4 3 2 4;}
@font-face
{font-family:"Segoe UI";
panose-1:2 11 5 2 4 2 4 2 2 3;}
@font-face
{font-family:"Segoe UI Semibold";
panose-1:2 11 7 2 4 2 4 2 2 3;}
/* Style Definitions */
p.MsoNormal, li.MsoNormal, div.MsoNormal
{margin:0in;
font-size:11.0pt;
font-family:"Calibri",sans-serif;
mso-ligatures:standardcontextual;}
a:link, span.MsoHyperlink
{mso-style-priority:99;
color:#0563C1;
text-decoration:underline;}
span.EmailStyle18
{mso-style-type:personal-compose;
font-family:"Calibri Light",sans-serif;
font-weight:bold;}
.MsoChpDefault
{mso-style-type:export-only;
font-size:10.0pt;
mso-ligatures:none;}
@page WordSection1
{size:8.5in 11.0in;
margin:1.0in 1.0in 1.0in 1.0in;}
div.WordSection1
{page:WordSection1;}
--></style><!--[if gte mso 9]><xml>
<o:shapedefaults v:ext="edit" spidmax="1026" />
</xml><![endif]--><!--[if gte mso 9]><xml>
<o:shapelayout v:ext="edit">
<o:idmap v:ext="edit" data="1" />
</o:shapelayout></xml><![endif]-->
</head>
<body lang="EN-US" link="#0563C1" vlink="#954F72" style="word-wrap:break-word">
<div class="WordSection1">
<p class="MsoNormal" style="text-autospace:none"><b><span style="font-family:"Calibri Light",sans-serif">Title<o:p></o:p></span></b></p>
<p class="MsoNormal" style="text-autospace:none"><span style="font-family:"Calibri Light",sans-serif">Revolutionizing Edge AI: Improved and Automated Model Compression<o:p></o:p></span></p>
<p class="MsoNormal" style="text-autospace:none"><b><span style="font-family:"Calibri Light",sans-serif">Bio<o:p></o:p></span></b></p>
<p class="MsoNormal" style="text-autospace:none"><span style="font-family:"Calibri Light",sans-serif">Kaiqi Zhao is a final-year Ph.D. candidate advised by Prof. Ming Zhao in the School of<o:p></o:p></span></p>
<p class="MsoNormal" style="text-autospace:none"><span style="font-family:"Calibri Light",sans-serif">Computing and Augmented Intelligence (SCAI) at Arizona State University (ASU). Her research<o:p></o:p></span></p>
<p class="MsoNormal" style="text-autospace:none"><span style="font-family:"Calibri Light",sans-serif">interests are machine learning (ML) and deep learning (DL) model compression as well as cloud<o:p></o:p></span></p>
<p class="MsoNormal" style="text-autospace:none"><span style="font-family:"Calibri Light",sans-serif">and edge computing. Central to her research are innovating solutions to optimize large-scale<o:p></o:p></span></p>
<p class="MsoNormal" style="text-autospace:none"><span style="font-family:"Calibri Light",sans-serif">ML/DL models for Internet-of-Things (IoT) data-driven applications. She has published<o:p></o:p></span></p>
<p class="MsoNormal" style="text-autospace:none"><span style="font-family:"Calibri Light",sans-serif">first-authored papers at top-tier AI conferences, e.g., AISTATS (2023 and 2024), ICASSP (Oral),<o:p></o:p></span></p>
<p class="MsoNormal" style="text-autospace:none"><span style="font-family:"Calibri Light",sans-serif">and InterSpeech (Oral), and top-tier edge computing conferences, e.g., SEC, and won the Best<o:p></o:p></span></p>
<p class="MsoNormal" style="text-autospace:none"><span style="font-family:"Calibri Light",sans-serif">Poster Award at SEC’24. She did three research internships at Amazon Web Services as an<o:p></o:p></span></p>
<p class="MsoNormal" style="text-autospace:none"><span style="font-family:"Calibri Light",sans-serif">Applied Scientist. One of her internship works about compressing speech recognition models<o:p></o:p></span></p>
<p class="MsoNormal" style="text-autospace:none"><span style="font-family:"Calibri Light",sans-serif">has been integrated into the Amazon Alex library for production usage. Additionally, she was<o:p></o:p></span></p>
<p class="MsoNormal" style="text-autospace:none"><span style="font-family:"Calibri Light",sans-serif">awarded the Graduate College Completion Fellowship, the most prestigious award for graduate<o:p></o:p></span></p>
<p class="MsoNormal" style="text-autospace:none"><span style="font-family:"Calibri Light",sans-serif">students at ASU, and SCAI Doctoral Fellowships in 2022 and 2023.<o:p></o:p></span></p>
<p class="MsoNormal" style="text-autospace:none"><b><span style="font-family:"Calibri Light",sans-serif">Abstract<o:p></o:p></span></b></p>
<p class="MsoNormal" style="text-autospace:none"><span style="font-family:"Calibri Light",sans-serif">Machine learning models are increasingly employed by smart devices on the edge to support<o:p></o:p></span></p>
<p class="MsoNormal" style="text-autospace:none"><span style="font-family:"Calibri Light",sans-serif">important applications such as real-time virtual assistants and privacy-preserving healthcare.<o:p></o:p></span></p>
<p class="MsoNormal" style="text-autospace:none"><span style="font-family:"Calibri Light",sans-serif">However, deploying state-of-the-art (SOTA) deep learning models on devices faces multiple<o:p></o:p></span></p>
<p class="MsoNormal" style="text-autospace:none"><span style="font-family:"Calibri Light",sans-serif">serious challenges. First, it is infeasible to deploy large models on resource-constrained edge<o:p></o:p></span></p>
<p class="MsoNormal" style="text-autospace:none"><span style="font-family:"Calibri Light",sans-serif">devices whereas small models cannot achieve the SOTA accuracy. Second, it is difficult to<o:p></o:p></span></p>
<p class="MsoNormal" style="text-autospace:none"><span style="font-family:"Calibri Light",sans-serif">customize the models according to diverse application requirements in accuracy and speed and<o:p></o:p></span></p>
<p class="MsoNormal" style="text-autospace:none"><span style="font-family:"Calibri Light",sans-serif">diverse capabilities of edge devices. This talk presents several novel solutions to<o:p></o:p></span></p>
<p class="MsoNormal" style="text-autospace:none"><span style="font-family:"Calibri Light",sans-serif">comprehensively address the above challenges through automated and improved model<o:p></o:p></span></p>
<p class="MsoNormal" style="text-autospace:none"><span style="font-family:"Calibri Light",sans-serif">compression. First, it introduces Automatic Attention Pruning (AAP), an adaptive,<o:p></o:p></span></p>
<p class="MsoNormal" style="text-autospace:none"><span style="font-family:"Calibri Light",sans-serif">attention-based pruning approach to automatically reduce model parameters while meeting<o:p></o:p></span></p>
<p class="MsoNormal" style="text-autospace:none"><span style="font-family:"Calibri Light",sans-serif">diverse user objectives in model size, speed, and accuracy. AAP achieves an impressive<o:p></o:p></span></p>
<p class="MsoNormal" style="text-autospace:none"><span style="font-family:"Calibri Light",sans-serif">92.72% parameter reduction in ResNet-101 on Tiny-ImageNet without causing any accuracy<o:p></o:p></span></p>
<p class="MsoNormal" style="text-autospace:none"><span style="font-family:"Calibri Light",sans-serif">loss. Second, it presents Self-Supervised Quantization-Aware Knowledge Distillation, a<o:p></o:p></span></p>
<p class="MsoNormal" style="text-autospace:none"><span style="font-family:"Calibri Light",sans-serif">framework for reducing model precision without supervision from labeled training data. For<o:p></o:p></span></p>
<p class="MsoNormal" style="text-autospace:none"><span style="font-family:"Calibri Light",sans-serif">example, it quantizes VGG-8 to 2 bits on CIFAR-10 without any accuracy loss. Finally, the talk<o:p></o:p></span></p>
<p class="MsoNormal" style="text-autospace:none"><span style="font-family:"Calibri Light",sans-serif">explores two more works, Contrastive Knowledge Distillation and Module Replacing, for further<o:p></o:p></span></p>
<p class="MsoNormal" style="text-autospace:none"><span style="font-family:"Calibri Light",sans-serif">improving the performance of small models. All the works presented in this talk are designed to<o:p></o:p></span></p>
<p class="MsoNormal" style="text-autospace:none"><span style="font-family:"Calibri Light",sans-serif">address real-world challenges, with applications extending to production environments such as<o:p></o:p></span></p>
<p class="MsoNormal" style="text-autospace:none"><span style="font-family:"Calibri Light",sans-serif">Amazon Alexa, and have been successfully deployed on diverse hardware platforms, including<o:p></o:p></span></p>
<p class="MsoNormal"><span style="font-family:"Calibri Light",sans-serif">cloud instances and edge devices, catalyzing AI for the edge.<o:p></o:p></span></p>
<div>
<p class="MsoNormal"><span style="color:#5F5F5F">________________________________________________________________________________</span>
<span style="mso-ligatures:none"><o:p></o:p></span></p>
</div>
<div>
<div style="margin-top:.25in;margin-bottom:15.0pt">
<p class="MsoNormal"><span style="font-size:18.0pt;font-family:"Segoe UI",sans-serif;color:#252424">Microsoft Teams meeting</span><span style="font-family:"Segoe UI",sans-serif;color:#252424">
<o:p></o:p></span></p>
</div>
<div style="margin-bottom:15.0pt">
<div>
<p class="MsoNormal"><b><span style="font-size:10.5pt;font-family:"Segoe UI",sans-serif;color:#252424">Join on your computer, mobile app or room device</span></b><b><span style="font-family:"Segoe UI",sans-serif;color:#252424">
<o:p></o:p></span></b></p>
</div>
<p class="MsoNormal"><span style="font-family:"Segoe UI",sans-serif;color:#252424"><a href="https://teams.microsoft.com/l/meetup-join/19%3ameeting_MzIxZWVkM2MtMjg1MC00MjQ5LWIwY2MtZWM5ZmY2ZjdlNmY4%40thread.v2/0?context=%7b%22Tid%22%3a%22f3308007-477c-4a70-8889-34611817c55a%22%2c%22Oid%22%3a%22685c3f4f-29d5-4141-ada5-0fdeab8480e4%22%7d" target="_blank"><span style="font-size:10.5pt;font-family:"Segoe UI Semibold",sans-serif;color:#6264A7">Click
here to join the meeting</span></a> <o:p></o:p></span></p>
</div>
<div style="margin-top:15.0pt;margin-bottom:15.0pt">
<div style="margin-bottom:3.0pt">
<p class="MsoNormal"><span style="font-size:10.5pt;font-family:"Segoe UI",sans-serif;color:#252424">Meeting ID:
</span><span style="font-size:12.0pt;font-family:"Segoe UI",sans-serif;color:#252424">283 234 327 430</span><span style="font-size:10.5pt;font-family:"Segoe UI",sans-serif;color:#252424">
</span><span style="font-family:"Segoe UI",sans-serif;color:#252424"><br>
</span><span style="font-size:10.5pt;font-family:"Segoe UI",sans-serif;color:#252424">Passcode:
</span><span style="font-size:12.0pt;font-family:"Segoe UI",sans-serif;color:#252424">Sf3wbd
</span><span style="font-family:"Segoe UI",sans-serif;color:#252424"><o:p></o:p></span></p>
<div>
<p class="MsoNormal"><span style="font-size:10.5pt;font-family:"Segoe UI",sans-serif;color:#252424"><a href="https://www.microsoft.com/en-us/microsoft-teams/download-app" target="_blank"><span style="color:#6264A7">Download Teams</span></a> |
<a href="https://www.microsoft.com/microsoft-teams/join-a-meeting" target="_blank">
<span style="color:#6264A7">Join on the web</span></a><o:p></o:p></span></p>
</div>
</div>
</div>
<div style="margin-bottom:3.0pt">
<div style="margin-bottom:3.0pt">
<p class="MsoNormal"><b><span style="font-size:10.5pt;font-family:"Segoe UI",sans-serif;color:#252424">Or call in (audio only)</span></b><span style="font-family:"Segoe UI",sans-serif;color:#252424">
<o:p></o:p></span></p>
</div>
<div style="margin-bottom:3.0pt">
<p class="MsoNormal"><span style="font-family:"Segoe UI",sans-serif;color:#252424"><a href="tel:+16147066572,,80448131# "><span style="font-size:10.5pt;color:#6264A7">+1 614-706-6572,,80448131#</span></a>
</span><span style="font-size:10.5pt;font-family:"Segoe UI",sans-serif;color:#252424"> United States, Columbus
</span><span style="font-family:"Segoe UI",sans-serif;color:#252424"><o:p></o:p></span></p>
</div>
</div>
<p class="MsoNormal"><span style="font-size:10.5pt;font-family:"Segoe UI",sans-serif;color:#252424">Phone Conference ID:
</span><span style="font-size:12.0pt;font-family:"Segoe UI",sans-serif;color:#252424">804 481 31#
</span><span style="font-family:"Segoe UI",sans-serif;color:#252424"><o:p></o:p></span></p>
<div style="margin-bottom:15.0pt">
<p class="MsoNormal"><span style="font-family:"Segoe UI",sans-serif;color:#252424"><a href="https://dialin.teams.microsoft.com/8f5f7319-0053-4423-a154-4f8b6e7fb7dd?id=80448131" target="_blank"><span style="font-size:10.5pt;color:#6264A7">Find a local number</span></a>
| <a href="https://dialin.teams.microsoft.com/usp/pstnconferencing" target="_blank">
<span style="font-size:10.5pt;color:#6264A7">Reset PIN</span></a> <o:p></o:p></span></p>
</div>
<div style="margin-bottom:3.0pt">
<div>
<p class="MsoNormal"><span style="font-size:10.5pt;font-family:"Segoe UI",sans-serif;color:#252424;mso-ligatures:none"><img border="0" width="188" height="30" style="width:1.9583in;height:.3125in" id="_x0000_i1025" src="https://www.ohio.edu/sites/default/files/2018-11/invite_logo_teams.jpg"></span><span style="font-size:10.5pt;font-family:"Segoe UI",sans-serif;color:#252424"><o:p></o:p></span></p>
</div>
<p class="MsoNormal"><span style="font-size:10.5pt;font-family:"Segoe UI",sans-serif;color:#252424">If you encounter issues with this meeting, please visit the Help link. If you are not able to resolve the problems, please contact the meeting organizer to let
them know you are having difficulty. <o:p></o:p></span></p>
</div>
<div style="margin-top:15.0pt;margin-bottom:.25in">
<p class="MsoNormal"><span style="font-family:"Segoe UI",sans-serif;color:#252424"><a href="https://aka.ms/JoinTeamsMeeting" target="_blank"><span style="font-size:10.5pt;color:#6264A7">Learn More</span></a> |
<a href="https://www.ohio.edu/oit/services/collaboration/teams/help" target="_blank">
<span style="font-size:10.5pt;color:#6264A7">Help</span></a> | <a href="https://teams.microsoft.com/meetingOptions/?organizerId=685c3f4f-29d5-4141-ada5-0fdeab8480e4&tenantId=f3308007-477c-4a70-8889-34611817c55a&threadId=19_meeting_MzIxZWVkM2MtMjg1MC00MjQ5LWIwY2MtZWM5ZmY2ZjdlNmY4@thread.v2&messageId=0&language=en-US" target="_blank">
<span style="font-size:10.5pt;color:#6264A7">Meeting options</span></a> <o:p></o:p></span></p>
</div>
</div>
<div>
<p class="MsoNormal"><span style="color:#5F5F5F">________________________________________________________________________________</span>
<o:p></o:p></p>
</div>
<p class="MsoNormal"><o:p> </o:p></p>
</div>
</body>
</html>