Week 1 Assignment 2 - BatchNormalization Axis=3

Course Q&A Deep Learning Specialization Convolutional Neural Networks
1

Week 1 Programming Assignment #2 Part 1 has a comment of “BatchNormalization for axis 3.” My question is why for axis 3?

Thanks in advance

2

Well, what is axis 3? Remember that we are doing 0-based indexing here. We are dealing with batches of images, so the dimensions are:

m x h x w x c

Where m is the number of samples (images), h and w are the vertical and horizontal coordinates of the pixels and c is the color channel values.

So it is normalizing over the color axis. Those are the actual “values” in this case and it would not make any sense to normalize them over any of the other axes, right?