BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Training and Tuning Deep Neural Networks: Faster\, Stronger and Be
 tter
DTSTART:20180823T093000
DTEND:20180823T113000
DTSTAMP:20260921T162047Z
UID:44dcf7542c2adeba8a13ab9d13ec273ce861e2592cb78c8d77da3742
CATEGORIES:Conferences - Seminars
DESCRIPTION:Chen Liu\nEDIC candidacy exam\nExam president: Prof. Pascal Fr
 ossard\nThesis advisor: Prof. Volkan Cevher\nCo-examiner: Prof. Alexandre 
 Alahi\n\nAbstract\nDeep neural networks dominate the state-of-the-art mode
 ls in tasks of computer vision\, speech\, bioinformatics etc. However\, tr
 aining modern deep neural networks is difficult\, because of the highly no
 n-convexity of the objective functions. On one hand\, deep neural networks
  usually have overwhelmingly large number of parameters\, they exhibit inh
 omogeneous curvatures along different directions. This place barriers for 
 training algorithms to escape saddle points. Numerous local minima in this
  ultra-high dimension space also poses a challenge for training algorithms
  to find a local minima which has good generalization property. On the oth
 er hand\, most modern neural network network models are vulnerable to some
  well-designed perturbation of the input. For most cases\, very small and 
 even imperceptible perturbations of input data can fool some sophisticated
  models. How to obtain a more robust model triggers more attention to the 
 curvature of training objective with respect to the input space. How to pr
 ovide some guarantee of the trained model against various input perturbati
 ons is also a open research question.\nMy research will mainly focus on th
 e training objective of deep neural networks in two aspects. The first foc
 us on the parameter space to design faster optimizers to find local minima
  with good generalization property faster. The second focus on the input s
 pace to find models robust to input perturbations.\n\nBackground papers\nT
 owards deep learning models resistant to adversarial attacks\, by  Madry\
 , A.\, et al\, arXiv preprint arXiv:1706.06083.\nCertifying some distribut
 ional robustness with principled adversarial training\, by  Aman\, S.\, e
 t al\, (2018).\nProvable defenses against adversarial examples via the con
 vex outer adversarial polytope\, by  Kolter J Z\, Wong E \,  arXiv prepr
 int arXiv:1711.00851\, 2017.\n\n 
LOCATION:ELD 120 https://plan.epfl.ch/?room=ELD120
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR