BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:US/Eastern BEGIN:DAYLIGHT DTSTART:20200308T030000 TZOFFSETFROM:-0500 TZOFFSETTO:-0400 RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3 TZNAME:EDT END:DAYLIGHT BEGIN:STANDARD DTSTART:20191103T010000 TZOFFSETFROM:-0400 TZOFFSETTO:-0500 RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11 TZNAME:EST END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20191114T200348Z UID:6202B406-4C63-439D-ADCB-62D39862D32D DTSTART;TZID=US/Eastern:20191113T150000 DTEND;TZID=US/Eastern:20191113T160000 DESCRIPTION:In today's world\, most data are collected in streams of contin uously flowing instances with some temporal nature. Many industries\, ther efore\, are faced with solving dynamic Machine Learning problems\; problem s where models must adapt to changing patterns\, or concept drift\, in ord er to remain useful. A common strategy is to predefine some arbitrary or d omain informed retrain cadence\, at which time a new model is trained from scratch on some amount of recent data. The trouble with this solution is that it is very uninformed (it has no information about when patterns actu ally change)\, prone to error (it has no information about what types of d rifts occurred or how to appropriately adapt) and can be inefficient or ev en completely impractical. Fear not\; there is a better way. Enter adaptiv e learning. Adaptive learning strategies learn and update incrementally on each instance\, or in batches\, with user in the loop or automatically wh ere appropriate. This keeps models up to date and relevant. Additionally\, concept drift detection mechanisms can inform the system when and how it should adapt\, depending on concept drift characteristics like magnitude a nd duration. These solutions ensure that models remain relevant through ti me\, reduce manual analysis during retrain and increase the longevity of s olutions. This talk will discuss data streams\, concept drift and concept drift detection methods. It will walk through various model adaption strat egies\, and finally\, online evaluation techniques.\n\nCo-sponsored by: A. Shami & OC2 Lab\n\nSpeaker(s): Sarah D'Ettorre \, \n\nRoom: 234\, Bldg: T EB\, ECE Dept.\, London\, Ontario\, Canada LOCATION:Room: 234\, Bldg: TEB\, ECE Dept.\, London\, Ontario\, Canada ORGANIZER:cgomezsu@uwo.ca SEQUENCE:2 SUMMARY:Introduction to Adaptive Learning URL;VALUE=URI:https://events.vtools.ieee.org/m/206904 X-ALT-DESC:Description: <br /><p>In today's world\, most data are collected in streams of continuously flowing instances with some temporal nature. M any industries\, therefore\, are faced with solving dynamic Machine Learni ng problems\; problems where models must adapt to changing patterns\, or c oncept drift\, in order to remain useful. A common strategy is to predefin e some arbitrary or domain informed retrain cadence\, at which time a new model is trained from scratch on some amount of recent data. The trouble w ith this solution is that it is very uninformed (it has no information abo ut when patterns actually change)\, prone to error (it has no information about what types of drifts occurred or how to appropriately adapt) and can be inefficient or even completely impractical. Fear not\; there is a bett er way. Enter adaptive learning. Adaptive learning strategies learn and up date incrementally on each instance\, or in batches\, with user in the loo p or automatically where appropriate. This keeps models up to date and rel evant. Additionally\, concept drift detection mechanisms can inform the sy stem when and how it should adapt\, depending on concept drift characteris tics like magnitude and duration. These solutions ensure that models remai n relevant through time\, reduce manual analysis during retrain and increa se the longevity of solutions. This talk will discuss data streams\, conce pt drift and concept drift detection methods. It will walk through various model adaption strategies\, and finally\, online evaluation techniques.</ p>\n<p>&nbsp\;</p> END:VEVENT END:VCALENDAR