BEGIN:VCALENDAR
PRODID:-//Columba Systems Ltd//NONSGML CPNG/SpringViewer/ICal Output/3.3-
 M3//EN
VERSION:2.0
CALSCALE:GREGORIAN
METHOD:PUBLISH
BEGIN:VEVENT
DTSTAMP:20210422T143129Z
DTSTART:20210505T130000Z
DTEND:20210505T140000Z
SUMMARY:Computer Science Research Talk: : Exposing Previously Undetectabl
 e Faults in Deep Neural Networks
UID:{http://www.columbasystems.com/customers/uom/gpp/eventid/}o5-knszgl2t
 -qoipga
DESCRIPTION:You are welcome to the forthcoming Research Talk in Computer 
 Science (online)\n\nJoining details:\n\nJoin Zoom Meeting\nhttps://zoom.
 us/j/91635482399\n\nSpeaker: Isaac Dunn\nHost: Dr Lucas Cordeiro\n\nTitl
 e: Exposing Previously Undetectable Faults in Deep Neural Networks\n\nAb
 stract: Existing methods for testing DNNs solve the oracle problem by co
 nstraining the raw features (e.g. image pixel values) to be within a sma
 ll distance of a dataset example for which the desired DNN output is kno
 wn. But this limits the kinds of faults these approaches are able to det
 ect. In this paper\, we introduce a novel DNN testing method that is abl
 e to find faults in DNNs that other methods cannot. The crux is that\, b
 y leveraging generative machine learning\, we can generate fresh test ca
 ses that vary in their high-level features (for images\, these include o
 bject shape\, location\, texture\, and colour). We demonstrate that our 
 approach is capable of detecting deliberately injected faults as well as
  new faults in state-of-the-art DNNs\, and that in both cases\, existing
  methods are unable to find these faults.\n\nBio: Isaac Dunn is a PhD ca
 ndidate (third year) at the University of Oxford. Their research focuses
  on improving our understanding of limitations in current machine learni
 ng models\, with a view to making improvements and working towards genui
 nely trustworthy ML systems.\n\n
STATUS:TENTATIVE
TRANSP:TRANSPARENT
CLASS:PUBLIC
LOCATION:https://zoom.us/j/91635482399
END:VEVENT
END:VCALENDAR