BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:IC Colloquium : Safety Verification of Deep Neural Networks
DTSTART:20170925T161500
DTEND:20170925T173000
DTSTAMP:20260506T045016Z
UID:e37047d6d2797fc27a29d1264609557551549093c2a7d0b47c367866
CATEGORIES:Conferences - Seminars
DESCRIPTION:By: Marta Kwiatkowska - Oxford University\nVideo of her talk\n
 \nAbstract:\nDeep neural networks have achieved impressive experimental re
 sults in image classification\, but can surprisingly be unstable with resp
 ect to adversarial perturbations\, that is\, minimal changes to the input 
 image that cause the network to misclassify it. With potential application
 s including perception modules and end-to-end controllers for self-driving
  cars\, this raises concerns about their safety. This lecture will describ
 e progress with developing a novel automated verification framework for de
 ep neural networks to ensure safety of their classification decisions with
  respect to image manipulations\, for example scratches or changes to came
 ra angle or lighting conditions\, that should not affect the classificatio
 n. The techniques work directly with the network code and\, in contrast to
  existing methods\, can offer guarantees that adversarial examples are fou
 nd if they exist. We implement the techniques using Z3 and evaluate them o
 n state-of-the-art networks\, including regularised and deep learning netw
 orks. We also compare against existing techniques to search for adversaria
 l examples.\n\nBio:\nMarta Kwiatkowska is Professor of Computing Systems a
 nd Fellow of Trinity College\, University of Oxford. Prior to this she was
  Professor in the School of Computer Science at the University of Birmingh
 am\, Lecturer at the University of Leicester and Assistant Professor at th
 e Jagiellonian University in Cracow\, Poland. Kwiatkowska has made funda
 mental contributions to the theory and practice of model checking for prob
 abilistic systems\, focusing on automated techniques for verification and 
 synthesis from quantitative specifications. She led the development of the
  PRISM model checker (www.prismmodelchecker.org)\, the leading software to
 ol in the area and winner of the HVC Award 2016. Probabilistic model check
 ing has been adopted in many diverse fields\, including distributed comput
 ing\, wireless networks\, security\, robotics\, game theory\, systems biol
 ogy\, DNA computing and nanotechnology\, with genuine flaws found and corr
 ected in real-world protocols. Kwiatkowska awarded an honorary doctorate f
 rom KTH Royal Institute of Technology in Stockholm in 2014 and the Royal S
 ociety Milner Medal in 2018. Her recent work was supported by the ERC Adva
 nced Grant VERIWARE "From software verification to ‘everyware’ verific
 ation" and the EPSRC Programme Granton Mobile Autonomy. She is a Fellow 
 of ACM and Member of Academia Europea.\n \nMore information
LOCATION:BC 420 https://plan.epfl.ch/?room==BC%20420
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR