Learning by de- and reconstruction – this is my motto when reading the book Understanding Deep Learning by Simon J.D. Prince, MIT Press, 2024. In 2012 he published a book titled Computer Vision: Models, Learning, and Inference. While that was based on Bayesian methodology the new book shifts the focus on a nonBayesian approach. I expect that further books of the same or other autors will combine both approaches and deepen the machine learning approach towards understanding and creating.

Learning by de- and reconstruction means that I take Prince’s Python notebooks, analyze and deconstruct the content and reconstruct the meaning in Julia/Pluto notebooks leaning on libraries such as FLUX.jl and LUX.jl.

1. Julia/Pluto-UDL-Notebook 1.1 -- Background Mathematics

2. Supervised Learning
   1. Julia/Pluto-UDL-Notebook 2.1
   2. Julia/Pluto-Notebook 2.1 with FLUX.jl
3. Shallow Neural Networks
   1. Shallow Neural Networks I
      1. Julia/Pluto-UDL-Notebook 3.1
      2. Julia/Pluto-Notebook 3.1 with FLUX.jl
      3. Julia/Pluto-Notebook 3.1 with LUX.jl
   2. Shallow Neural Networks II
      1. Julia/Pluto-UDL-Notebook 3.2
      2. Julia/Pluto-Notebook 3.2 with FLUX.jl
   3. Shallow Network Regions: Julia/Pluto-UDL-Notebook 3.3
   4. Activation functions
      1. Julia/Pluto-UDL-Notebook 3.4
      2. Julia/Pluto-UDL-Notebook 3.4 with FLUX.jl

-----------------------------------------------------------------------------------------

This is all draft for personal use; comments, bug reports, or proposals are welcome:

claus.moebus(at)uol.de

-----------------------------------------------------------------------------------------