BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:Foundation models for wearable sensor data
DTSTART:20260527T070000Z
DTEND:20260527T130000Z
DTSTAMP:20260514T024000Z
UID:indico-event-3811@indico.ijs.si
CONTACT:events@slaif.si
DESCRIPTION:Speakers: Gašper Slapničar\, Mitja Lustrek (Department of In
 telligent Systems\, Jozef Stefan Institute)\n\nCourse provider: Jožef Ste
 fan Institute (JSI)Instructors: Gašper Slapničar (JSI)\, Mitja Luštrek 
 (JSI)\nLearning objectives: Gain practical knowledge on what FMs exist an
 d where they are useful.\nCourse content: The course will overview the mai
 n methodological ideas on how to build\, adapt\, and evaluate large self-s
 upervised models for biosignals (e.g.\, PPG\, ECG\, EEG …) and motion se
 nsors (accelerometers\, gyroscopes) under the messy constraints of real-wo
 rld wearables (noise\, motion artifacts\, missingness\, device/site shifts
 \, and limited labels). The training covers modern foundation-model backb
 ones for time series (CNN and ResNet encoders\, Transformers\, etc.)\, pre
 training paradigms (masked modeling\, contrastive/relative contrastive lea
 rning)\, and representation design choices that matter specifically for ph
 ysiology (beat-synchronous views\, morphology-aware learning). Moreover we
  will highlight state-of-the-art approaches on merging LLMs and sensor-bas
 ed models that result in hybrid approaches allowing for human-language int
 erpretation of wearable data (e.g.\, SensorLM).\nLearning outcomes: Practi
 cal knowledge on what FMs exist\, where they are useful\, and how to take 
 them beyond default setups to be useful in their domains and problems.\n\n
 https://indico.ijs.si/event/3811/
IMAGE;VALUE=URI:https://indico.ijs.si/event/3811/logo-1596601024.png
LOCATION:Jamova
URL:https://indico.ijs.si/event/3811/
END:VEVENT
END:VCALENDAR