Neural Code for Faces Deciphered

akakaak writes: In a new paper published in Cell (http://dx.doi.org/10.1016/j.cell.2017.05.011), researchers Le Chang and Doris Tsao claim to have uncovered "The Code for Facial Identity in the Primate Brain". They develop a model representing each face as a vector in a 50-dimensional "face-space", and show that the firing rate for each face-sensitive neuron represents the location along a single axis through this space. This allows them to accurately predict the appearance of a viewed face from the collective recorded activity of the neurons. This work is a major advance in the decoding of complex neural representations, and refutes exemplar-based models of face recognition.