Caffe Alexnet Model Understanding

Before looking at Caffenet, now look at this alexnet, mainly to help understand the paper.

Here is the main record of some and caffenet different places.

The first layer: mainly first normalized re-pooling

Second layer: deviation is 0.1. First normalized and then pooled

Layer Three: identical

Fourth floor: deviation is 0.1.

Fifth floor: Deviation is 0.1.

Sixth floor: Deviation of 0.1

Seventh floor: Deviation of 0.1

From, can also be seen, with Caffenet, is convolution and normalization of the order, as well as the difference between the size of the gap.

Caffe Alexnet Model Understanding