better to let object with both A and B characteristics be represented by both A and B active rather than learning something as AB. Better for generalization.
current focusing on image recognition
want to recognize variations of a picture of an object
starting with an end result, suppose I don't know the structure of the data
can try to build a heirarcy
Official GHC 07 Blogger You may comment on this blog
by visiting *GHC bulletin board*.