Enhancing NN

Perhaps not easy, but essential. For image recognition we need not just
software that works with pixels, but software that can add the intermediate
step of grouping pixels into eyeglasses, beards, and other features of
varying degrees of reliability and permanence. If a computer can play chess
as well as a human, it should be possible to spot the Mona Lisa with a
mustache.

Bill Silvert

"Dmitry A. Kazakov" <[EMAIL PROTECTED]> wrote in message
news:[EMAIL PROTECTED]
>  I asked whether the NN could be told to ignore
> >glasses, which would reduce the training set to 5,000 pictures, but was
told
> >that couldn't be done. Why not?
>
> That would not be easy for any system.