Some parts are hard to understand due to a mixture of audio quality and strong accent
YouTube auto-generated captions getting a lot of stuff wrong
potential chances for misunderstandings / student disconnects
Quality clears up a bit halfway though, maybe not as much of a problem as originally thought (terms appearing on the screen provide a lot of the clarification)
21st April 2025:
Move bonus parts to own Jupyter notebook
Update text, remove "cardinality" definition at the top and add it later
I think Section 4 also needs to be bonus section
Tidied up language and examples
Add in idea about PCA for dictionary learning from basis feature sets
The section "Identifying Sparse Contributions Across Frequencies" doesn't really add anything for me - it doesn't show anything above and beyond what the previous figure showed (and that's way more informative) and it's just a chance to show differences in frequency components, which is a style of plot that's already been presented in earlier sections, so I will remove it. It also felt like it needed a few sentences to introduce what's being shown, so making it effective would have just taken up even more space