News for hackers. Stuff that matters.
Re:What's Wrong with Deep Learning? by Houshalter on Monday June 15, 2015 @03:36PM (#9721486)
Hinton doesn't like that pooling loses track of the exact locations where features are located, and just tests if a feature occurs in some area.

The basic effect of this is to decrease the resolution, so it's more tractable to operate on. Without pooling you are stuck with a huge resolution at each layer.

There are no comments on this comment yet.