BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:Asia/Shanghai BEGIN:STANDARD DTSTART:19910915T010000 TZOFFSETFROM:+0900 TZOFFSETTO:+0800 TZNAME:CST END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20251118T031808Z UID:E2AA9D21-17C7-4983-B7F6-C6C49EFE2A90 DTSTART;TZID=Asia/Shanghai:20251120T160000 DTEND;TZID=Asia/Shanghai:20251120T170000 DESCRIPTION:Continual learning is a vital area of deep learning that enable s trained models to acquire new tasks incrementally\, without retraining f rom scratch for each task. This talk introduces Analytic Continual Learnin g (ACL)\, a novel branch of continual learning that leverages classical re cursive least squares methods. By employing simple linear algebra\, ACL br ings closed-form recursive algorithms into deep continual learning\, achie ving an equivalence between continual learning and joint training. As a re sult\, deep models can be trained in a single epoch while maintaining high training speed and near-zero forgetting. This approach has been successfu lly applied to conventional continual learning settings as well as highly challenging scenarios involving large models.\n\nBinary Coffee\, Qingshui he Campus\, University of Electronic Science and Technology of China\, Che ngdu\, Sichuan\, China LOCATION:Binary Coffee\, Qingshuihe Campus\, University of Electronic Scie nce and Technology of China\, Chengdu\, Sichuan\, China ORGANIZER:windof47@gmail.com SEQUENCE:7 SUMMARY:Analytic Continual Learning with a Fast and Non-forgetting Closed-f orm Solution by Prof. Zhiping Lin URL;VALUE=URI:https://events.vtools.ieee.org/m/514731 X-ALT-DESC:Description: <br /><p><span lang="EN-US" style="font-size: 12.0p t\; mso-bidi-font-size: 11.0pt\; font-family: 'Times New Roman'\,serif\; m so-fareast-font-family: 等线\; mso-fareast-theme-font: minor-fareast\; m so-ansi-language: EN-US\; mso-fareast-language: EN-US\; mso-bidi-language: AR-SA\;">Continual learning is a vital area of deep learning that enables trained models to acquire new tasks incrementally\, without retraining fr om scratch for each task. This talk introduces Analytic Continual Learning (ACL)\, a novel branch of continual learning that leverages classical rec ursive least squares methods. By employing simple linear algebra\, ACL bri ngs closed-form recursive algorithms into deep continual learning\, achiev ing an equivalence between continual learning and joint training. As a res ult\, deep models can be trained in a single epoch while maintaining high training speed and near-zero forgetting. This approach has been successful ly applied to conventional continual learning settings as well as highly c hallenging scenarios involving large models.</span></p> END:VEVENT END:VCALENDAR