In article <1993May7.182337.18009 at reed.edu>, mfunk at reed.edu (Michele Funk) writes:
|> My scientific mind has been teased by a friend who claims to have heard
|> about a neural net that was trained to distinguish male and female
|> faces. The real curiosity is that this invoved an intermediate step
|> of having the face reduced from something like 25 nodes to 8 and then
|> the original input had to be recreated from the 8 nodes.
|>|> In anycase, my dear friend, who was so kind as to throw this potentially
|> thrilling bit-o-research out there doesn't remember the refernce. Have you
|> heard of any such thing, and if so, who did the work? Where can I get the
|> paper (if it's been written up)?
It was presented at the Nov 1990 NIPS conference, and published in the
proceedings (Golumb, Lawrence, and Sejnowski, in Advances in Neural
Information Processing 3, published by Morgan Kauffman 1991), page 573.