Make SVM figures for non-geometrically motivated soft-margin SVM

[davidrosenberg]

i.e. adapt pictures from @brett1479 's SVM notes (https://davidrosenberg.github.io/mlcourse/Archive/2017/Labs/3-SVM-Notes_sol.pdf) to our treatment of SVMs without discussion of geometric margin. (Basically, if f(x) is our score function, draw the level sets (lines) at f(x)=0 and f(x) = +-1. Can label every training point with its score. Circle every point that's a support vector.

[davidrosenberg]

More ideas: one could draw a circle around each point, color-coded based on the coefficient (e.g. yellow for zero, red for the max value, and some version orange for everything in between). Maybe another circle around the point color-coded for the margin (not the geometric margin). Then lines corresponding to the target geometric margin (see Brett’s Ivanov reformulation).