It is a disappointment that the boosting results are so bad,but a good thing that the Nearest Neighbor method does get
good results.
Last year when we ran the Nearest Neighbor algorithm, the results were not that great. This year, however,
they are better because we have a lot more training data and better quality. Last year we had 85 examples of
each letter and this year we have 675.
Furthermore, this year the examples are cropped better and are higher quality