BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//132.216.98.100//NONSGML kigkonsult.se iCalcreator 2.20.4//
BEGIN:VEVENT
UID:20251121T162542EST-1135ILU79p@132.216.98.100
DTSTAMP:20251121T212542Z
DESCRIPTION:Title: Learn then Test: Calibrating Predictive Algorithms to Ac
 hieve Risk Control.\n\n\n	Abstract:\n\n\nWe introduce Learn then Test\, a f
 ramework for calibrating machine learning models so that their predictions
  satisfy explicit\, finite-sample statistical guarantees regardless of the
  underlying model and (unknown) data-generating distribution. The framewor
 k addresses\, among other examples\, false discovery rate control in multi
 -label classification\, intersection-over-union control in instance segmen
 tation\, and the simultaneous control of the type-1 error of outlier detec
 tion and confidence set coverage in classification or regression. To accom
 plish this\, we solve a key technical challenge: the control of arbitrary 
 risks that are not necessarily monotonic. Our main insight is to reframe t
 he risk-control problem as multiple hypothesis testing\, enabling techniqu
 es and mathematical arguments different from those in the previous literat
 ure. We use our framework to provide new calibration methods for several c
 ore machine learning tasks with detailed worked examples in computer visio
 n.\n\nThis is joint work with Anastasios Angelopoulos\, Emmanuel Candès\, 
 Michael I. Jordan\, and Lihua Lei.\n\n\n	Speaker\n\n\nDr. Bates is a postdo
 ctoral researcher with Michael I. Jordan in the Statistics and EECS depart
 ments at UC Berkeley. He works on developing methods to analyze modern sci
 entific data sets\, leveraging sophisticated black box models while provid
 ing rigorous statistical guarantees. Specifically\, he works on problems i
 n high-dimensional statistics (especially false discovery rate control)\, 
 statistical machine learning\, conformal prediction and causal inference.
 \n\nPreviously\, he completed his Ph.D. in the Stanford Department of Stat
 istics advised by Emmanuel Candes. His thesis introduced methods for condi
 tional independence testing and false discovery rate control in genomics\,
  and he was honored to receive the Ric Weiland Graduate Fellowship and the
  Theodore W. Anderson Theory of Statistics Dissertation Award for this wor
 k. Before his Ph.D.\, he studied statistics and mathematics at Harvard Uni
 versity\, and spent a year teaching mathematics at NYU Shanghai. Outside r
 esearch\, I enjoy triathlons\, sailing\, hiking\, and reading speculative 
 fiction novels.\n\nhttps://mcgill.zoom.us/j/83436686293?pwd=b0RmWmlXRXE3OW
 R6NlNIcWF5d0dJQT09\n\nMeeting ID: 834 3668 6293\n\nPasscode: 12345\n\n \n
 \n \n
DTSTART:20220401T193000Z
DTEND:20220401T203000Z
SUMMARY:Stephen Bates (UC Berkeley)
URL:/mathstat/channels/event/stephen-bates-uc-berkeley
 -338799
END:VEVENT
END:VCALENDAR