What information does each dimension of W contain?


The screenshot is from the 1st assignment of Course 4… Just a bit confused about the dimension of W (filter)…

Hi Wz111,

The convolutions are over volume using filters. To convolve over the input volume the volume of a filter is filter size x filter size x depth of input volume. This is fxfxn__C_prev. Now the depth of the output volume is determined by the number of filters. So n_C indicates the number of filters used, which is equal to the depth of the output volume.

I hope this clarifies things.