BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:IC Colloquium: Kernel distances for distinguishing and sampling fr
 om probability distributions
DTSTART:20190218T101500
DTEND:20190218T111500
DTSTAMP:20260407T163555Z
UID:886f2784f2ff18e99d1f7232d92150209c873b114049137fbdb2deeb
CATEGORIES:Conferences - Seminars
DESCRIPTION:By: Dougal Sutherland - University College London\nIC Faculty 
 candidate\n\nAbstract:\nProbability distributions are the core object of s
 tatistical machine learning\, and one of the core properties we can consid
 er is distances between them. In this talk\, we will consider using these 
 distances for two important tasks\, and show how to design distances which
  will be useful for each task. First\, we study the problem of two-sample 
 testing\, where we wish to determine whether two different datasets meanin
 gfully differ\, and if so how they differ. We then study this framework in
  the setting of training generative models\, such as generative adversaria
 l networks (GANs)\, which learn to sample from complex distributions such 
 as those of natural images.\n \nThe distances used are defined in terms o
 f kernels\, but we parameterize these kernels in as deep networks for flex
 ibility. This combination gives both theoretical and practical benefits ov
 er staying purely in either framework\, and we obtain state-of-the-art res
 ults for unsupervised image generation on CelebA and ImageNet with our nov
 el Scaled MMD GAN.\n\nBio:\nDougal Sutherland is a postdoctoral researcher
  at the Gatsby Computational Neuroscience Unit\, University College London
 \, working with Arthur Gretton. He received his PhD in 2016 from Carnegie 
 Mellon University\, advised by Jeff Schneider. His research focuses on pro
 blems of learning about distributions from samples\, including training im
 plicit generative models\, density estimation\, two-sample testing\, and d
 istribution regression. His work combines kernel frameworks with deep lear
 ning\, and aims for theoretical grounding of practical results.\n\nMore in
 formation
LOCATION:BC 420 https://plan.epfl.ch/?room==BC%20420
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR