BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Training and Understanding Deep Neural Networks for Robotics\, Des
 ign\, and Perception - Jason Yosinski (Cornell)
DTSTART:20150909T100000Z
DTEND:20150909T110000Z
UID:TALK60594@talks.cam.ac.uk
CONTACT:Zoubin Ghahramani
DESCRIPTION:Artificial Neural Networks (ANNs) form a powerful class of mod
 els with both theoretical and practical advantages. Networks with more tha
 n one hidden layer (deep neural networks) compute multiple functions on la
 ter layers that share the use of intermediate results computed on earlier 
 layers. This compositional\, hierarchical structure provides a strong bias
 \, or regularization\, toward solutions that seem to work well on a large 
 variety of real-world problems.\n\nIn this talk I will begin by showing a 
 few examples of how this general compositional bias can excel at such dive
 rse tasks as designing robot gaits and 3D objects. I will then discuss a f
 ew simple experiments that shed light on the inner workings of neural nets
  trained to classify images. The first study examines the computation perf
 ormed by the entire set of neurons on a layer in a network\, and subsequen
 t work illuminates the computation performed by individual units\, and fin
 ally the computation performed by the network as a whole. The experiments 
 taken together reveal some surprising behaviors of large networks and lead
  to a greater understanding and intuition for the computation performed by
  deep neural nets.
LOCATION:Engineering Department\, CBL Room BE-438
END:VEVENT
END:VCALENDAR