For gdb stuff I've just been working and got phase_2 = 1 2 4 8 16 32, started vaguely on phase 3, looked semi-vaguely at textbook, and spent the rest of my time on sklearn.
I have my github page updated or mostly updated with my sklearn stuff. I've gotten some weird behaviors, like that lin_reg was predicting fractions for binary labels, so it ended up having a massive rmse, while actually predicting better than tree regression. I started working with gold difference at 15 because using win/loss as the regressions were not predicting well for wins - better with gda i think.
