Max Pooling for Image Recognition

The goal is to get it all the way down through these layers to just a zero or a one a single number Yep The other thing That's interesting about max pooling is that and this is this is a thing that's kind of nice about it Uh It protects you a little bit from small changes in the image because imagine that your image gets a little bit distorted somehow. Then the max pooling means that Even if some of those individual outputs change as a result of the distortion There's a good chance that The max will still be the max even after the distortion.