Let's consider
one of the more interesting features in detail: race.
For
the purposes of this project, we wish to classify faces as either Asian,
Black, Hispanic,
or White.
Our work
on the race feature provides good examples of:
- Evolving
classification methodology
Adapting
the classification strategy in response to the characteristics of
the training data.
- Training
data augmentation
Strategies
for getting more from a sparse training set.
|