BEGIN:VCALENDAR
PRODID:-//eluceo/ical//2.0/EN
VERSION:2.0
CALSCALE:GREGORIAN
BEGIN:VEVENT
UID:69c75f8fa33eff65108d747b304367f1
DTSTAMP:20260715T141051Z
SUMMARY:I-X Seminar Series: Understanding Data\, Noise\, and Uncertainty in
  Neural Networks with Kamil Ciosek
DESCRIPTION:\n\nAbstract:\n\n\n\n\n\nThis talk explores how the Neural Tang
 ent Kernel (NTK) can be used as a practical lens for understanding data in
 fluence\, noise\, and uncertainty in modern neural networks. Rather than f
 ocusing only on predictive accuracy\, we study how individual training exa
 mples affect predictions\, how models behave under noisy observations\, an
 d how to quantify when predictions should be trusted. Building on the well
 -established NTK connection between neural networks\, kernel methods\, and
  Gaussian processes\, we present three results: an information theoretic a
 pproach to data attribution\, a characterization of how regularization cor
 responds to observation noise in wide networks\, and an efficient method f
 or uncertainty estimation that captures more structure than standard last 
 layer approaches while remaining computationally practical. Together\, the
 se results show how the NTK perspective can provide simple and interpretab
 le tools for reasoning about model behaviour beyond accuracy.\n\nBio: \nK
 amil Ciosek is a researcher working on the theory and practice of machine 
 learning\, with interests spanning Neural Tangent Kernels\, Gaussian proce
 ss perspectives on neural networks\, Bayesian uncertainty estimation in de
 ep learning\, and calibration. His recent work studies the connections bet
 ween wide neural networks and Gaussian processes\, including the role of N
 TK features in Bayesian inference and uncertainty quantification. Separate
 ly\, he has also worked on calibration of predictive models\, including bo
 th first-order and second-order calibration methods. Kamil is currently a 
 Senior Research Scientist at Spotify and previously held research position
 s at Microsoft Research Cambridge and the University of Oxford. He receive
 d his PhD in Machine Learning from UCL.\n\n\n \n\n\nThere will be a netwo
 rking opportunity following this talk. If you would like to attend\, pleas
 e complete the registration form. Registration will close on Tuesday 16 Ju
 ne at 1200.\n\n
URL:https://www.imperial.ac.uk/events/209542/i-x-seminar-series-understandi
 ng-data-noise-and-uncertainty-in-neural-networks-with-kamil-ciosek/
DTSTART;TZID=Europe/London:20260623T140000
DTEND;TZID=Europe/London:20260623T150000
LOCATION:LRT 608 A+B\, I-X Level 6\, Translation and Innovation Hub (I-HUB)
 \, White City Campus\, Imperial College London\, London\, W12 0BZ\, United
  Kingdom
END:VEVENT
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
DTSTART:20260623T140000
TZNAME:BST
TZOFFSETTO:+0100
TZOFFSETFROM:+0100
END:DAYLIGHT
END:VTIMEZONE
END:VCALENDAR