I did not attend the talk. It seems the slides are quite high-level and not necessarily explain a lot of details. I found the following to be very good(Learning architectures for AI; I read half of it):
http://www.iro.umontreal.ca/~lisa/pointeurs/TR1312.pdf
