Testing ML
- Create output that is easy to compare by computer (so numerical results are preferable over a graph)
- Fix randomizations to make the results repeatable
- Establish thresholds for results using different datasets (and also using different models)