In the introduction of the book, section “What Our Image Recognizer Learned”, there is this paragraph:
For each layer, the image part with the light gray background shows the reconstructed weights pictures,
We talk about layer 1 I believe, Is the first word supposed to be layer? I’m struggling to understand it, and to me for each node makes more sense.
They are talking about
layers. The visualizations are not just raw weights from nodes converted to images.
The paper refered in the book list the details of the processing done to get the visualizations (1).
From the paper: “We show the top 9 activations in a random subset of feature maps across the validation data, projected down to pixel space using our deconvolutional network approach.”
(1) Visualizing and Understanding Convolutional Networks
Thanks. it’s quite complex. Maybe will come back to it later on in the journey.