"These dimensions create a mental “face space” in which an infinite number of faces can be recognized. There is probably an average face, or something like it, at the origin, and the brain measures the deviation from this base.A newly encountered face might lie five units away from the average face in one dimension, seven units in another, and so forth. Each face cell reads the combined vector of about six of these dimensions. The signals from 200 face cells altogether serve to uniquely identify a face."
I don't understand this fully, not in the sense I understand "2+2 = 4", and the article doesn't go into the idea of an "average face". At least the first page of the Cell report doesn't go into an "average face" either, so I'm sticking my neck out when I write the rest of this post.
Assume there is an "average face" stored in our memories which serves as the baseline against which the coding of a new face takes place (like the Greenwich Meridian, serving as the starting point). I'd guess there's some innate biology we're born with, but the pump is primed by our early childhood experiences. So maybe by 1 week, 1 month, or 1 year we have an "average face" pretty well constructed.
Note the implications: we'll find faces more similar to our average face easier to recognize and likely more attractive. ( (It'd be a neat experience to test adults on facial recognition to see if they recognize faces similar to their mothers faster or as more pleasing than others.) That would account for the common idea that "all X's (insert race or ethnicity of your choice) look alike".
Now those implications aren't supported by the article or report--there's no implication that there's learning involved in recognizing faces. The way the biologists did the experiment they weren't likely to see it.