My attempt for the verification of the hidden activations representing a 3-bit encoding of the input as shown on slide 15 lecture 9. You get better and worse results tweaking the learning rate. However, I did not manage to achieve a proper encoding. Maybe someone of you finds an error in my implementation or a better learning rate.
https://colab.research.google.com/drive/1NrNxlO43o9f50GsJubNoRzdvrJ7VElyI?usp=sharing