BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:America/Los_Angeles BEGIN:DAYLIGHT DTSTART:20220313T030000 TZOFFSETFROM:-0800 TZOFFSETTO:-0700 RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3 TZNAME:PDT END:DAYLIGHT BEGIN:STANDARD DTSTART:20211107T010000 TZOFFSETFROM:-0700 TZOFFSETTO:-0800 RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11 TZNAME:PST END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20211206T172929Z UID:CBB46ACF-B597-469C-A76D-2AB0AEFD414F DTSTART;TZID=America/Los_Angeles:20211130T120000 DTEND;TZID=America/Los_Angeles:20211130T130000 DESCRIPTION:IEEE SPS chapter of the Oregon section will host an SPS disting uished lecture by Prof. Venkatesh Saligrama (Boston University). The lectu re title is "Personalizing Federated Learning to the Edge Device"\, and th e lecture is scheduled at noon-1pm PST on Tuesday\, November 30.\n\nThis e vent will be held via Zoom. When you register to the event\, please enter your email address. We will send Zoom information to the registrants via e mail one day before the event.\n\nThe talk abstract and the speaker bio ar e as follows:\nTitle: Personalizing Federated Learning to the Edge Device\ n\nAbstract: We propose a novel method for federated learning that is cust omized to the objective of a given edge device. In our proposed method\, a server trains a global meta-model by collaborating with devices without a ctually sharing data. The trained global meta-model is then customized loc ally by each device to meet its specific objective. Different from the con ventional federated learning setting\, training customized models for each device is hindered by both the inherent data biases of the various device s\, as well as the requirements imposed by the federated architecture. We present an algorithm that locally de-biases model updates\, while leveragi ng distributed data\, so that each device can be effectively customized to wards its objectives. Our method is fully agnostic to device heterogeneity and imbalanced data\, scalable to massive number of devices\, and allows for arbitrary partial participation. Our method has built-in convergence g uarantees\, and on benchmark datasets we demonstrate that it outperforms o ther state-of-art methods.\n\nBio: Venkatesh Saligrama is a faculty member in the Department of Electrical and Computer Engineering\, the Department of Computer Science (by courtesy)\, and a founding member of the Faculty of Computing and Data Sciences at Boston University. He holds a PhD from M IT. His research interests are broadly in the area of Artificial Intellige nce\, and his recent work has focused on machine learning with resource-co nstraints. He is an IEEE Fellow and recipient of several awards including Distinguished Lecturer for IEEE Signal Processing Society\, the Presidenti al Early Career Award (PECASE)\, ONR Young Investigator Award\, the NSF Ca reer Award. More information about his work is available at http://sites.b u.edu/data\n\nVirtual: https://events.vtools.ieee.org/m/291032 LOCATION:Virtual: https://events.vtools.ieee.org/m/291032 ORGANIZER:jinsub.kim@oregonstate.edu SEQUENCE:1 SUMMARY:IEEE Signal Processing Society Distinguished Lecture ("Personalizin g Federated Learning to the Edge Device") URL;VALUE=URI:https://events.vtools.ieee.org/m/291032 X-ALT-DESC:Description: <br /><p>IEEE SPS chapter of the Oregon section wil l host an SPS distinguished lecture by <strong>Prof. Venkatesh Saligrama < /strong>(Boston University). The lecture title is "<strong>Personalizing F ederated Learning to the Edge Device</strong>"\, and the lecture is schedu led at <strong>noon-1pm PST</strong> on <strong>Tuesday\, November 30</str ong>.</p>\n<p>This event will be <strong>held via Zoom</strong>. When you register to the event\, please <span style="text-decoration: underline\;"> enter your email address</span>. We will send Zoom information to the regi strants via email one day before the event.</p>\n<p>The talk abstract and the speaker bio are as follows:</p>\n<div><strong>Title</strong>: Personal izing Federated Learning to the Edge Device</div>\n<div>&nbsp\;</div>\n<di v><strong>Abstract</strong>: We propose a novel method for federated learn ing that is customized to the objective of a given edge device. In our pro posed method\, a server trains a global meta-model by collaborating with d evices without actually sharing data. The trained global meta-model is the n customized locally by each device to meet its specific objective.&nbsp\; Different from the conventional federated learning setting\, training cus tomized models for each device is hindered by both the inherent data biase s of the various devices\, as well as the requirements imposed by the fede rated architecture. We present an algorithm that locally de-biases model u pdates\, while leveraging distributed data\, so that each device can be ef fectively customized towards its objectives.&nbsp\; Our method is fully ag nostic to device heterogeneity and imbalanced data\, scalable to massive n umber of devices\, and allows for arbitrary partial participation. Our met hod has built-in convergence guarantees\, and on benchmark datasets we dem onstrate that it outperforms other state-of-art methods. &nbsp\;</div>\n<d iv>\n<div dir="ltr">\n<div dir="ltr">\n<div>&nbsp\;</div>\n<div><strong>Bi o</strong>: <span class="il">Venkatesh</span> Saligrama is a faculty membe r in the Department of Electrical and Computer Engineering\, the Departmen t of Computer Science (by courtesy)\, and a founding member of the Faculty of Computing and Data Sciences at Boston University. He holds a PhD from MIT. His research interests are broadly in the area of Artificial Intellig ence\, and his recent work has focused on machine learning with resource-c onstraints. He is an IEEE Fellow and recipient of several awards including Distinguished Lecturer for IEEE Signal Processing Society\, the President ial Early Career Award (PECASE)\, ONR Young Investigator Award\, the NSF C areer Award. More information about his work is available at&nbsp\;<a href ="http://sites.bu.edu/data" target="_blank" rel="noopener noreferrer" data -saferedirecturl="https://www.google.com/url?q=http://sites.bu.edu/data&am p\;source=gmail&amp\;ust=1637645921723000&amp\;usg=AOvVaw2h9Xpk4LfRTq5B754 sgccc">http://sites.bu.edu/data</a></div>\n</div>\n</div>\n</div>\n<p>&nbs p\;</p> END:VEVENT END:VCALENDAR