Include in introduction: deep learning networks as learning increasingly abstract representations of the input data

[svenvanderburg]

To give more intuition of how deep neural networks work, we could include in the introduction that you can view neural networks as learning increasingly abstract representations of the input data, for example using a picture like this:

https://www.researchgate.net/profile/Abhinav-Vishnu/publication/308883811/figure/fig2/AS:413666332692485@1475637148009/Pictorial-representation-of-features-in-3-layers-of-a-CNN-34-Notice-the-increasing.png

[svenvanderburg]

Unfortunately, I cannot find any CC-by licensed figures that visualize this well... And I think we need a figure to explain this well. The best way would be to find a figure in an open source journal, but I haven't found that yet.

[svenvanderburg]

This is what wikipedia offers us, but I don't really like it: https://en.wikipedia.org/wiki/File:Deep_Learning.jpg