TODOs

Code stuff

✅ Get rid of init stuff
✅ leading underscores for hidden methods

Text

✅ output/outcome/response
✅ Model set-up
✅ log likelihood
✅ sklearn
✅ multi-class
✅ non-linear
✅ underline vs. headers for "model set up" and other section labels
✅ In Generative classifiers, maybe change labels to prior and likelihood (rather than p(y) and p(x|y))
✅ Link vs. do this for dataset names

Notation/convention

"linear combination"
✅Add a thing to the conventions note about L(theta; data). Note that data might be represented as upper case or lower case
✅add a thing about { }_{n = 1}^N
✅ Make r.v.s upper case
✅ make matrices bold
✅ change losses to L

Content changes

✅ Gradient ascent vs. descent
✅ add mention of linear (no) activation function in NNs
✅ Neural nets make $\bx \in \R^{D_X}$ and $\by \in \R^{D_y}$ .
✅ Change wine dataset to cancer in c3 binary stuff

Things to review

Appendix

THINGS TO ADD

C2 ARIMA
C2 nonparametric regression (with kernels)
Unsupervised
SVMs