More code presentation improvements: Factor Graphs, Generative Classification

[fonsp]

More upgraded code examples! This time:

• Factor Graphs
• Generative Classification (minor changes)
• Regression (tiny changes)

[fonsp]

@bertdv can you take a look at the new Factor Graphs lecture? In particular the "key concepts" that I added might need some tweaking.

[bertdv]

I reviewed only the Factor graphs.jl file. Looks great, brilliant work!! I edited the file a bit to my taste, see commit #0be46f9

Some notes:

• The issue with moving a two-column into a details box (in the exercises section) did not work yet. Please review.

• I want to be a bit conservative with "key concept" boxes, for now. They are essential, but they should not become a summary box. There are only a few key concepts in this course (maybe 10 to 15, about 1 or maximal 2 per lecture), and I prefer to really think about them at a higher level (like what is the main message of this course) when all the materials are there. E.g., perhaps it is better to finalize the lecture with 1 or two key concept boxes, rather than "hiding" them in the middle of a lecture. Not sure what is best, but I want to avoid having 3 or 4 key concepts per lecture.

[fonsp]

• The issue with moving a two-column into a details box (in the exercises section) did not work yet. Please review.

I had to fix this in Pluto (JuliaPluto/Pluto.jl#3321) and it was only released last week in Pluto 0.20.16. Can you update Pluto? I think you currently have 0.20.13.

• I want to be a bit conservative with "key concept" boxes, for now. They are essential, but they should not become a summary box. There are only a few key concepts in this course (maybe 10 to 15, about 1 or maximal 2 per lecture), and I prefer to really think about them at a higher level (like what is the main message of this course) when all the materials are there. E.g., perhaps it is better to finalize the lecture with 1 or two key concept boxes, rather than "hiding" them in the middle of a lecture. Not sure what is best, but I want to avoid having 3 or 4 key concepts per lecture.

Yep I agree! Let's use another type of box (maybe a block quote) for summaries that are not "key concepts".

[fonsp]

The way I see it, this math part does not need to change (estimating a linear function R^D -> R).

But our application is to estimate a polynomial R -> R. And then the trick is that you can see a polynomial of degree D-1 as a linear function R^D -> R, if you first map x to [0, x, x^2, ..., x^D].

I can write something about this in the lecture if you agree?

[fonsp]

I left my last comment as a TODO, I'll make an issue so we don't forget