Dimensional reduction can be helpful for making sense of data with
many dimensions. Even when removing several dimensions, a large
portion of the original variance can be maintained. Some dimensions
must be retained for the cosine-theta distance measurement to be
useful.
Kalin Jonas
