BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Date iCal//NONSGML kigkonsult.se iCalcreator 2.20.2//
METHOD:PUBLISH
X-WR-CALNAME;VALUE=TEXT:TSE
BEGIN:VTIMEZONE
TZID:Europe/Paris
BEGIN:STANDARD
DTSTART:20251026T030000
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20260329T020000
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
UID:calendar.138343.field_date.0@www.tse-fr.eu
DTSTAMP:20260510T200820Z
CREATED:20251006T171001Z
DESCRIPTION:Jean Baptiste Fermanian (Université de Montpellier\;INRIA)\, “C
 lass conditional conformal prediction for multiple inputs by p-value aggre
 gation”\, MAD-Stat. Seminar\, Toulouse: TSE\, November 6\, 2025\, 11:00–12
 :15\, room Auditorium 3.\n\nConformal prediction methods are statistical t
 ools designed to quantify uncertainty\nand generate predictive sets with g
 uaranteed coverage probabilities. This work\nintroduces an innovative refi
 nement to these methods for classification tasks\, specifically\ntailored 
 for scenarios where multiple observations (multi-inputs) of a single\ninst
 ance are available at prediction time. Our approach is particularly motiva
 ted\nby applications in citizen science\, where multiple images of the sam
 e plant or\nanimal are captured by individuals. Our method integrates the 
 information from\neach observation into conformal prediction\, enabling a 
 reduction in the size of\nthe predicted label set while preserving the req
 uired class-conditional coverage\nguarantee. The approach is based on the 
 aggregation of conformal p-values computed\nfrom each observation of a mul
 ti-input. By exploiting the exact distribution\nof these p-values\, we pro
 pose a general aggregation framework using an abstract\nscoring function\,
  encompassing many classical statistical tools. Knowledge of this\ndistrib
 ution also enables refined versions of standard strategies\, such as major
 ity\nvoting. We evaluate our method on simulated and real data\, with a pa
 rticular focus\non Pl@ntNet\, a prominent citizen science platform that fa
 cilitates the collection\nand identification of plant species through user
 -submitted images.
DTSTART;TZID=Europe/Paris:20251106T110000
DTEND;TZID=Europe/Paris:20251106T121500
LAST-MODIFIED:20251028T011002Z
LOCATION:Toulouse: TSE\, November 6\, 2025\, 11:00–12:15\, room Auditorium 
 3
SUMMARY:MAD-Stat. Seminar
URL;TYPE=URI:https://www.tse-fr.eu/seminars/2025-class-conditional-conforma
 l-prediction-multiple-inputs-p-value-aggregation
END:VEVENT
END:VCALENDAR