October 2018
Intermediate to advanced
246 pages
6h 26m
English
The logic behind keeping a rectified linear units (ReLUs) layer is very simple: it replaces all the negative values with 0. This helps us to make CNN mathematically healthier by avoiding negative values:

Here, in this layer, the size of the image is not altered. We will get the same size output as the input only when the negative values are replaced by 0.
Read now
Unlock full access