TODO

Mention that under supervised learning, we have classification and regression
Increase in v-measure with ground truth knowledge -- only tiny, so knowing ground truth doesn't help much because of poor data quality
Add material showing that, if we add more correctly labeled data, we still shift the line we fit in linear regression, which is silly for classification
Note that the exact form of the logistic regression function doesn't matter -- it's just convenient

Times:
  Intro: 30 mins
  Supervised: 90 mins
  Break: 20 minutes
  DR: 30 mins
  Clustering: 45 mins