Databases Reference
In-Depth Information
harder. It also establishes a band of accuracy around a problem that
you generally don't have—in other words, given no other information,
with nobody else working on the problem you're working on, you don't
know if your 75% accurate model is the best possible.
Figure 7-1. Chris Mulligan, a student in Rachel's class, created this
leapfrogging visualization to capture the competition in real time as it
progressed throughout the semester
This leapfrogging effect is good and bad. It encourages people to
squeeze out better performing models, possibly at the risk of overfit‐
ting, but it also tends to make models much more complicated as they
get better. One reason you don't want competitions lasting too long is
that, after a while, the only way to inch up performance is to make
things ridiculously complicated. For example, the original Netflix
Prize lasted two years and the final winning model was too compli‐
cated for them to actually put into production.
